Datasets:

debajyotidasgupta

/

dblp_2022_2023

like 0

Title: SPViT: Enabling Faster Vision Transformers via Latency-Aware Soft Token Pruning. Abstract: Recently, Vision Transformer (ViT) has continuously established new milestones in the computer vision field, while the high computation and memory cost makes its propagation in industrial production difficult. Considering the computation complexity, the internal data pattern of ViTs, and the edge device deployment, we propose a latency-aware soft token pruning framework, SPViT, which can be set up on vanilla Transformers of both flatten and hierarchical structures, such as DeiTs and Swin-Transformers (Swin). More concretely, we design a dynamic attention-based multi-head token selector, which is a lightweight module for adaptive instance-wise token selection. We further introduce a soft pruning technique, which integrates the less informative tokens chosen by the selector module into a package token rather than discarding them completely. SPViT is bound to the trade-off between accuracy and latency requirements of specific edge devices through our proposed latency-aware training strategy. Experiment results show that SPViT significantly reduces the computation cost of ViTs with comparable performance on image classification. Moreover, SPViT can guarantee the identified model meets the latency specifications of mobile devices and FPGA, and even achieve the real-time execution of DeiT-T on mobile devices. For example, SPViT reduces the latency of DeiT-T to 26 ms (26%−41% superior to existing works) on the mobile device with 0.25%−4% higher top-1 accuracy on ImageNet. Our code is released at https://github.com/PeiyanFlying/SPViT.

Title: Multi-granularity Pruning for Model Acceleration on Mobile Devices. Abstract: For practical deep neural network design on mobile devices, it is essential to consider the constraints incurred by the computational resources and the inference latency in various applications. Among deep network acceleration approaches, pruning is a widely adopted practice to balance the computational resource consumption and the accuracy, where unimportant connections can be removed either channel-wisely or randomly with a minimal impact on model accuracy. The coarse-grained channel pruning instantly results in a significant latency reduction, while the fine-grained weight pruning is more flexible to retain accuracy. In this paper, we present a unified framework for the Joint Channel pruning and Weight pruning, named JCW, which achieves a better pruning proportion between channel and weight pruning. To fully optimize the trade-off between latency and accuracy, we further develop a tailored multi-objective evolutionary algorithm in the JCW framework, which enables one single round search to obtain the accurate candidate architectures for various deployment requirements. Extensive experiments demonstrate that the JCW achieves a better trade-off between the latency and accuracy against previous state-of-the-art pruning methods on the ImageNet classification dataset.

Title: ℓ ∞-Robustness and Beyond: Unleashing Efficient Adversarial Training. Abstract: Neural networks are vulnerable to adversarial attacks: adding well-crafted, imperceptible perturbations to their input can modify their output. Adversarial training is one of the most effective approaches in training robust models against such attacks. However, it is much slower than vanilla training of neural networks since it needs to construct adversarial examples for the entire training data at every iteration, hampering its effectiveness. Recently, Fast Adversarial Training (FAT) was proposed that can obtain robust models efficiently. However, the reasons behind its success are not fully understood, and more importantly, it can only train robust models for \(\ell _\infty \)-bounded attacks as it uses FGSM during training. In this paper, by leveraging the theory of coreset selection, we show how selecting a small subset of training data provides a general, more principled approach toward reducing the time complexity of robust training. Unlike existing methods, our approach can be adapted to a wide variety of training objectives, including TRADES, \(\ell _p\)-PGD, and Perceptual Adversarial Training (PAT). Our experimental results indicate that our approach speeds up adversarial training by 2–3 times while experiencing a slight reduction in the clean and robust accuracy.

Title: SAFA: Sample-Adaptive Feature Augmentation for Long-Tailed Image Classification. Abstract: Imbalanced datasets with long-tailed distribution widely exist in practice, posing great challenges for deep networks on how to handle the biased predictions between head (majority, frequent) classes and tail (minority, rare) classes. Feature space of tail classes learned by deep networks is usually under-represented, causing heterogeneous performance among different classes. Existing methods augment tail-class features to compensate tail classes on feature space, but these methods fail to generalize on test phase. To mitigate this problem, we propose a novel Sample-Adaptive Feature Augmentation (SAFA) to augment features for tail classes resulting in ameliorating the classifier performance. SAFA aims to extract diverse and transferable semantic directions from head classes, and adaptively translate tail-class features along extracted semantic directions for augmentation. SAFA leverages a recycling training scheme ensuring augmented features are sample-specific. Contrastive loss ensures the transferable semantic directions are class-irrelevant and mode seeking loss is adopted to produce diverse tail-class features and enlarge the feature space of tail classes. The proposed SAFA as a plug-in is convenient and versatile to be combined with different methods during training phase without additional computational burden at test time. By leveraging SAFA, we obtain outstanding results on CIFAR-LT-10, CIFAR-LT-100, Places-LT, ImageNet-LT, and iNaturalist2018.

Title: AutoMix: Unveiling the Power of Mixup for Stronger Classifiers. Abstract: Data mixing augmentation have proved to be effective for improving the generalization ability of deep neural networks. While early methods mix samples by hand-crafted policies (e.g., linear interpolation), recent methods utilize saliency information to match the mixed samples and labels via complex offline optimization. However, there arises a trade-off between precise mixing policies and optimization complexity. To address this challenge, we propose a novel automatic mixup (AutoMix) framework, where the mixup policy is parameterized and serves the ultimate classification goal directly. Specifically, AutoMix reformulates the mixup classification into two sub-tasks (i.e., mixed sample generation and mixup classification) with corresponding sub-networks and solves them in a bi-level optimization framework. For the generation, a learnable lightweight mixup generator, Mix Block, is designed to generate mixed samples by modeling patch-wise relationships under the direct supervision of the corresponding mixed labels. To prevent the degradation and instability of bi-level optimization, we further introduce a momentum pipeline to train AutoMix in an end-to-end manner. Extensive experiments on nine image benchmarks prove the superiority of AutoMix compared with state-of-the-arts in various classification scenarios and downstream tasks.

Title: 3D CoMPaT: Composition of Materials on Parts of 3D Things. Abstract: We present 3D CoMPaT, a richly annotated large-scale dataset of more than 7.19 million rendered compositions of Materials on Parts of 7262 unique 3D Models; 990 compositions per model on average. 3D CoMPaT covers 43 shape categories, 235 unique part names, and 167 unique material classes that can be applied to parts of 3D objects. Each object with the applied part-material compositions is rendered from four equally spaced views as well as four randomized views, leading to a total of 58 million renderings (7.19 million compositions \(\times 8{}\) views). This dataset primarily focuses on stylizing 3D shapes at part-level with compatible materials. We introduce a new task, called Grounded CoMPaT Recognition (GCR), to collectively recognize and ground compositions of materials on parts of 3D objects. We present two variations of this task and adapt state-of-art 2D/3D deep learning methods to solve the problem as baselines for future research. We hope our work will help ease future research on compositional 3D Vision. The dataset and code are publicly available at https://www.3dcompat-dataset.org/.

Title: OOD-CV: A Benchmark for Robustness to Out-of-Distribution Shifts of Individual Nuisances in Natural Images. Abstract: Enhancing the robustness of vision algorithms in real-world scenarios is challenging. One reason is that existing robustness benchmarks are limited, as they either rely on synthetic data or ignore the effects of individual nuisance factors. We introduce OOD-CV , a benchmark dataset that includes out-of-distribution examples of 10 object categories in terms of pose, shape, texture, context and the weather conditions, and enables benchmarking models for image classification, object detection, and 3D pose estimation. In addition to this novel dataset, we contribute extensive experiments using popular baseline methods, which reveal that: 1) Some nuisance factors have a much stronger negative effect on the performance compared to others, also depending on the vision task. 2) Current approaches to enhance robustness have only marginal effects, and can even reduce robustness. 3) We do not observe significant differences between convolutional and transformer architectures. We believe our dataset provides a rich testbed to study robustness and will help push forward research in this area.

Title: PANDORA: A Panoramic Detection Dataset for Object with Orientation. Abstract: Panoramic images have become increasingly popular as omnidirectional panoramic technology has advanced. Many datasets and works resort to object detection to better understand the content of the panoramic image. These datasets and detectors use a Bounding Field of View (BFoV) as a bounding box in panoramic images. However, we observe that the object instances in panoramic images often appear with arbitrary orientations. It indicates that BFoV as a bounding box is inappropriate, limiting the performance of detectors. This paper proposes a new bounding box representation, Rotated Bounding Field of View (RBFoV), for the panoramic image object detection task. Then, based on the RBFoV, we present a PANoramic Detection dataset for Object with oRientAtion (PANDORA). Finally, based on PANDORA, we evaluate the current state-of-the-art panoramic image object detection methods and design an anchor-free object detector called R-CenterNet for panoramic images. Compared with these baselines, our R-CenterNet shows its advantages in terms of detection performance. Our PANDORA dataset and source code are available at https://github.com/tdsuper/SphericalObjectDetection.

Title: Dress Code: High-Resolution Multi-category Virtual Try-On. Abstract: Image-based virtual try-on strives to transfer the appearance of a clothing item onto the image of a target person. Prior work focuses mainly on upper-body clothes (e.g. t-shirts, shirts, and tops) and neglects full-body or lower-body items. This shortcoming arises from a main factor: current publicly available datasets for image-based virtual try-on do not account for this variety, thus limiting progress in the field. To address this deficiency, we introduce Dress Code, which contains images of multi-category clothes. Dress Code is more than \(3\times \) larger than publicly available datasets for image-based virtual try-on and features high-resolution paired images (\(1024 \times 768\)) with front-view, full-body reference models. To generate HD try-on images with high visual quality and rich in details, we propose to learn fine-grained discriminating features. Specifically, we leverage a semantic-aware discriminator that makes predictions at pixel-level instead of image- or patch-level. Extensive experimental evaluation demonstrates that the proposed approach surpasses the baselines and state-of-the-art competitors in terms of visual quality and quantitative results. The Dress Code dataset is publicly available at https://github.com/aimagelab/dress-code.

Title: Learning Omnidirectional Flow in 360$^\circ $ Video via Siamese Representation. Abstract: Optical flow estimation in omnidirectional videos faces two significant issues: the lack of benchmark datasets and the challenge of adapting perspective video-based methods to accommodate the omnidirectional nature. This paper proposes the first perceptually natural-synthetic omnidirectional benchmark dataset with a 360\(^\circ \) field of view, FLOW360, with 40 different videos and 4,000 video frames. We conduct comprehensive characteristic analysis and comparisons between our dataset and existing optical flow datasets, which manifest perceptual realism, uniqueness, and diversity. To accommodate the omnidirectional nature, we present a novel Siamese representation Learning framework for Omnidirectional Flow (SLOF). We train our network in a contrastive manner with a hybrid loss function that combines contrastive loss and optical flow loss. Extensive experiments verify the proposed framework’s effectiveness and show up to 40% performance improvement over the state-of-the-art approaches. Our FLOW360 dataset and code are available at https://siamlof.github.io/.

Title: $\mathrm {CT^2}$: Colorization Transformer via Color Tokens. Abstract: Automatic image colorization is an ill-posed problem with multi-modal uncertainty, and there remains two main challenges with previous methods: incorrect semantic colors and under-saturation. In this paper, we propose an end-to-end transformer-based model to overcome these challenges. Benefited from the long-range context extraction of transformer and our holistic architecture, our method could colorize images with more diverse colors. Besides, we introduce color tokens into our approach and treat the colorization task as a classification problem, which increases the saturation of results. We also propose a series of modules to make image features interact with color tokens, and restrict the range of possible color candidates, which makes our results visually pleasing and reasonable. In addition, our method does not require any additional external priors, which ensures its well generalization capability. Extensive experiments and user studies demonstrate that our method achieves superior performance than previous works.

Title: Data Association Between Event Streams and Intensity Frames Under Diverse Baselines. Abstract: This paper proposes a learning-based framework to associate event streams and intensity frames under diverse camera baselines, to simultaneously benefit camera pose estimation under large baselines and depth estimation under small baselines. Based on the observation that event streams are globally sparse (a small percentage of pixels in global frames are triggered with events) and locally dense (a large percentage of pixels in local patches are triggered with events) in the spatial domain, we put forward a two-stage architecture for matching feature maps. LSparse-Net uses a large receptive field to find sparse matches while SDense-Net uses a small receptive field to find dense matches. Both stages apply Transformer modules with self-attention layers and cross-attention layers to effectively process multi-resolution features from the feature pyramid network backbone. Experimental results on public datasets show a systematic performance improvement for both tasks compared to state-of-the-art methods.

Title: Aliasing and adversarial robust generalization of CNNs. Abstract: Many commonly well-performing convolutional neural network models have shown to be susceptible to input data perturbations, indicating a low model robustness. To reveal model weaknesses, adversarial attacks are specifically optimized to generate small, barely perceivable image perturbations that flip the model prediction. Robustness against attacks can be gained by using adversarial examples during training, which in most cases reduces the measurable model attackability. Unfortunately, this technique can lead to robust overfitting, which results in non-robust models. In this paper, we analyze adversarially trained, robust models in the context of a specific network operation, the downsampling layer, and provide evidence that robust models have learned to downsample more accurately and suffer significantly less from downsampling artifacts, aka. aliasing, than baseline models. In the case of robust overfitting, we observe a strong increase in aliasing and propose a novel early stopping approach based on the measurement of aliasing.

Title: Smoothing policies and safe policy gradients. Abstract: Policy gradient (PG) algorithms are among the best candidates for the much-anticipated applications of reinforcement learning to real-world control tasks, such as robotics. However, the trial-and-error nature of these methods poses safety issues whenever the learning process itself must be performed on a physical system or involves any form of human-computer interaction. In this paper, we address a specific safety formulation, where both goals and dangers are encoded in a scalar reward signal and the learning agent is constrained to never worsen its performance, measured as the expected sum of rewards. By studying actor-only PG from a stochastic optimization perspective, we establish improvement guarantees for a wide class of parametric policies, generalizing existing results on Gaussian policies. This, together with novel upper bounds on the variance of PG estimators, allows us to identify meta-parameter schedules that guarantee monotonic improvement with high probability. The two key meta-parameters are the step size of the parameter updates and the batch size of the gradient estimates. Through a joint, adaptive selection of these meta-parameters, we obtain a PG algorithm with monotonic improvement guarantees.

Title: Speeding-up one-versus-all training for extreme classification via mean-separating initialization. Abstract: In this paper, we show that a simple, data dependent way of setting the initial vector can be used to substantially speed up the training of linear one-versus-all classifiers in extreme multi-label classification (XMC). We discuss the problem of choosing the initial weights from the perspective of three goals. We want to start in a region of weight space (a) with low loss value, (b) that is favourable for second-order optimization, and (c) where the conjugate-gradient (CG) calculations can be performed quickly. For margin losses, such an initialization is achieved by selecting the initial vector such that it separates the mean of all positive (relevant for a label) instances from the mean of all negatives – two quantities that can be calculated quickly for the highly imbalanced binary problems occurring in XMC. We demonstrate a training speedup of up to \(5\times\) on Amazon-670K dataset with 670,000 labels. This comes in part from the reduced number of iterations that need to be performed due to starting closer to the solution, and in part from an implicit negative-mining effect that allows to ignore easy negatives in the CG step. Because of the convex nature of the optimization problem, the speedup is achieved without any degradation in classification accuracy. The implementation can be found at https://github.com/xmc-aalto/dismecpp.

Title: Machine learning in corporate credit rating assessment using the expanded audit report. Abstract: We investigate whether key audit matter (KAM) paragraphs disclosed in extended audit reports—paragraphs in which the auditor highlights significant risks and critical judgments of the company—contribute to assess corporate credit ratings. This assessment is a complicated and expensive process to grade the reliability of a company, and it is relevant for many stakeholders, such as issuers, investors, and creditors. Although credit rating evaluations have attracted the interest of many researchers, previous studies have mainly focused only on financial ratios. We are the first to use KAMs for credit rating modelling purposes. Applying four machine learning techniques to answer this real-world problem—C4.5 decision tree, two different rule induction classifiers (PART algorithm and Rough Set) and the logistic regression methodology—, our evidence suggests that by simply identifying the KAM topics disclosed in the report, any decision-maker can assess credit scores with 74% accuracy using the rules provided by the PART algorithm. These rules specifically indicate that KAMs on both external (such as going concern) and internal (such as company debt) aspects may contribute to explaining a company’s credit rating. The rule induction classifiers have similar predictive power. Interestingly, if we combine audit data with accounting ratios, the predictive power of our model increases to 84%, outperforming the accuracy in the existing literature.

Title: A taxonomy for similarity metrics between Markov decision processes. Abstract: Although the notion of task similarity is potentially interesting in a wide range of areas such as curriculum learning or automated planning, it has mostly been tied to transfer learning. Transfer is based on the idea of reusing the knowledge acquired in the learning of a set of source tasks to a new learning process in a target task, assuming that the target and source tasks are close enough. In recent years, transfer learning has succeeded in making reinforcement learning (RL) algorithms more efficient (e.g., by reducing the number of samples needed to achieve (near-)optimal performance). Transfer in RL is based on the core concept of similarity: whenever the tasks are similar, the transferred knowledge can be reused to solve the target task and significantly improve the learning performance. Therefore, the selection of good metrics to measure these similarities is a critical aspect when building transfer RL algorithms, especially when this knowledge is transferred from simulation to the real world. In the literature, there are many metrics to measure the similarity between MDPs, hence, many definitions of similarity or its complement distance have been considered. In this paper, we propose a categorization of these metrics and analyze the definitions of similarity proposed so far, taking into account such categorization. We also follow this taxonomy to survey the existing literature, as well as suggesting future directions for the construction of new metrics.

Title: Attacking neural machine translations via hybrid attention learning. Abstract: Deep-learning based natural language processing (NLP) models are proven vulnerable to adversarial attacks. However, there is currently insufficient research that studies attacks to neural machine translations (NMTs) and examines the robustness of deep-learning based NMTs. In this paper, we aim to fill this critical research gap. When generating word-level adversarial examples in NLP attacks, there is a conventional trade-off in existing methods between the attacking performance and the amount of perturbations. Although some literature has studied such a trade-off and successfully generated adversarial examples with a reasonable amount of perturbations, it is still challenging to generate highly successful translation attacks while concealing the changes to the texts. To this end, we propose a novel Hybrid Attentive Attack method to locate language-specific and sequence-focused words, and make semantic-aware substitutions to attack NMTs. We evaluate the effectiveness of our attack strategy by attacking three high-performing translation models. The experimental results show that our method achieves the highest attacking performance compared with other existing attacking strategies.

Title: Bayesian mixture variational autoencoders for multi-modal learning. Abstract: This paper provides an in-depth analysis on how to effectively acquire and generalize cross-modal knowledge for multi-modal learning. Mixture-of-Expert (MoE) and Product-of-Expert (PoE) are two popular directions in generalizing multi-modal information. Existing works based on MoE or PoE have shown notable improvement on data generation, while new challenges such as high training cost, overconfident experts, and encoding modal-specific features also emerge. In this work, we propose Bayesian mixture variational autoencoder (BMVAE) which learns to select or combine experts via Bayesian inference. We show that the proposed idea can naturally encourage models to learn modal-specific knowledge and avoid overconfident experts. Also, we show that the idea is compatible with both MoE and PoE frameworks. When being a MoE model, BMVAE can be optimized by a tight lower bound and is efficient to train. The PoE BMVAE has the same advantages and a theoretical connection to existing works. In the experiments, we show that BMVAE achieves state-of-the-art performance.

Title: A brain-inspired algorithm for training highly sparse neural networks. Abstract: Sparse neural networks attract increasing interest as they exhibit comparable performance to their dense counterparts while being computationally efficient. Pruning the dense neural networks is among the most widely used methods to obtain a sparse neural network. Driven by the high training cost of such methods that can be unaffordable for a low-resource device, training sparse neural networks sparsely from scratch has recently gained attention. However, existing sparse training algorithms suffer from various issues, including poor performance in high sparsity scenarios, computing dense gradient information during training, or pure random topology search. In this paper, inspired by the evolution of the biological brain and the Hebbian learning theory, we present a new sparse training approach that evolves sparse neural networks according to the behavior of neurons in the network. Concretely, by exploiting the cosine similarity metric to measure the importance of the connections, our proposed method, “Cosine similarity-based and random topology exploration (CTRE)”, evolves the topology of sparse neural networks by adding the most important connections to the network without calculating dense gradient in the backward. We carried out different experiments on eight datasets, including tabular, image, and text datasets, and demonstrate that our proposed method outperforms several state-of-the-art sparse training algorithms in extremely sparse neural networks by a large gap. The implementation code is available on Github.

Title: An accurate, scalable and verifiable protocol for federated differentially private averaging. Abstract: Learning from data owned by several parties, as in federated learning, raises challenges regarding the privacy guarantees provided to participants and the correctness of the computation in the presence of malicious parties. We tackle these challenges in the context of distributed averaging, an essential building block of federated learning algorithms. Our first contribution is a scalable protocol in which participants exchange correlated Gaussian noise along the edges of a graph, complemented by independent noise added by each party. We analyze the differential privacy guarantees of our protocol and the impact of the graph topology under colluding malicious parties, showing that we can nearly match the utility of the trusted curator model even when each honest party communicates with only a logarithmic number of other parties chosen at random. This is in contrast with protocols in the local model of privacy (with lower utility) or based on secure aggregation (where all pairs of users need to exchange messages). Our second contribution enables users to prove the correctness of their computations without compromising the efficiency and privacy guarantees of the protocol. Our construction relies on standard cryptographic primitives like commitment schemes and zero knowledge proofs.

Title: Emergence of norms in interactions with complex rewards. Abstract: Autonomous agents are becoming increasingly ubiquitous and are playing an increasing role in wide range of safety-critical systems, such as driverless cars, exploration robots and unmanned aerial vehicles. These agents operate in highly dynamic and heterogeneous environments, resulting in complex behaviour and interactions. Therefore, the need arises to model and understand more complex and nuanced agent interactions than have previously been studied. In this paper, we propose a novel agent-based modelling approach to investigating norm emergence, in which such interactions can be investigated. To this end, while there may be an ideal set of optimally compatible actions there are also combinations that have positive rewards and are also compatible. Our approach provides a step towards identifying the conditions under which globally compatible norms are likely to emerge in the context of complex rewards. Our model is illustrated using the motivating example of self-driving cars, and we present the scenario of an autonomous vehicle performing a left-turn at a T-intersection.

Title: Matrix-wise ℓ 0-constrained sparse nonnegative least squares. Abstract: Nonnegative least squares problems with multiple right-hand sides (MNNLS) arise in models that rely on additive linear combinations. In particular, they are at the core of most nonnegative matrix factorization algorithms and have many applications. The nonnegativity constraint is known to naturally favor sparsity, that is, solutions with few non-zero entries. However, it is often useful to further enhance this sparsity, as it improves the interpretability of the results and helps reducing noise, which leads to the sparse MNNLS problem. In this paper, as opposed to most previous works that enforce sparsity column- or row-wise, we first introduce a novel formulation for sparse MNNLS, with a matrix-wise sparsity constraint. Then, we present a two-step algorithm to tackle this problem. The first step divides sparse MNNLS in subproblems, one per column of the original problem. It then uses different algorithms to produce, either exactly or approximately, a Pareto front for each subproblem, that is, to produce a set of solutions representing different tradeoffs between reconstruction error and sparsity. The second step selects solutions among these Pareto fronts in order to build a sparsity-constrained matrix that minimizes the reconstruction error. We perform experiments on facial and hyperspectral images, and we show that our proposed two-step approach provides more accurate results than state-of-the-art sparse coding heuristics applied both column-wise and globally.

Title: Speeding up neural network robustness verification via algorithm configuration and an optimised mixed integer linear programming solver portfolio. Abstract: Despite their great success in recent years, neural networks have been found to be vulnerable to adversarial attacks. These attacks are often based on slight perturbations of given inputs that cause them to be misclassified. Several methods have been proposed to formally prove robustness of a given network against such attacks. However, these methods typically give rise to high computational demands, which severely limit their scalability. Recent state-of-the-art approaches state the verification task as a minimisation problem, which is formulated and solved as a mixed-integer linear programming (MIP) problem. We extend this approach by leveraging automated algorithm configuration techniques and, more specifically, construct a portfolio of MIP solver configurations optimised for the neural network verification task. We test this approach on two recent, state-of-the-art MIP-based verification engines, \(\mathrm {MIPVerify}\) and \(\mathrm {Venus}\), and achieve substantial improvements in CPU time by average factors of up to 4.7 and 10.3, respectively.

Title: Learning with risks based on M-location. Abstract: In this work, we study a new class of risks defined in terms of the location and deviation of the loss distribution, generalizing far beyond classical mean-variance risk functions. The class is easily implemented as a wrapper around any smooth loss, it admits finite-sample stationarity guarantees for stochastic gradient methods, it is straightforward to interpret and adjust, with close links to M-estimators of the loss location, and has a salient effect on the test loss distribution, giving us control over symmetry and deviations that are not possible under naive ERM.

Title: Unavoidable deadends in deterministic partially observable contingent planning. Abstract: Traditionally, a contingent plan, branching on the observations an agent obtains throughout plan execution, must reach a goal state from every possible initial state. However, in many real world problems, no such plan exists. Yet, there are plans that reach the goal from some initial states only. From the other initial states, they eventually reach a deadend—a state from which the goal can not be achieved. Deadends that cannot be avoided by resorting to a different plan, are called unavoidable deadends. In this paper we study planning with unavoidable deadends in belief space. We distinguish between two types of such deadends, and adapt offline and online contingent planners to identify and handle unavoidable deadends, using two approaches—an active approach that begins by distinguishing between the solvable and deadend states, and a lazy approach, that plans to achieve the goal, identifying deadends as they occur. We empirically analyze how each approach performs in different cases.

Title: Towards combining commonsense reasoning and knowledge acquisition to guide deep learning. Abstract: Algorithms based on deep network models are being used for many pattern recognition and decision-making tasks in robotics and AI. Training these models requires a large labeled dataset and considerable computational resources, which are not readily available in many domains. Also, it is difficult to explore the internal representations and reasoning mechanisms of these models. As a step towards addressing the underlying knowledge representation, reasoning, and learning challenges, the architecture described in this paper draws inspiration from research in cognitive systems. As a motivating example, we consider an assistive robot trying to reduce clutter in any given scene by reasoning about the occlusion of objects and stability of object configurations in an image of the scene. In this context, our architecture incrementally learns and revises a grounding of the spatial relations between objects and uses this grounding to extract spatial information from input images. Non-monotonic logical reasoning with this information and incomplete commonsense domain knowledge is used to make decisions about stability and occlusion. For images that cannot be processed by such reasoning, regions relevant to the tasks at hand are automatically identified and used to train deep network models to make the desired decisions. Image regions used to train the deep networks are also used to incrementally acquire previously unknown state constraints that are merged with the existing knowledge for subsequent reasoning. Experimental evaluation performed using simulated and real-world images indicates that in comparison with baselines based just on deep networks, our architecture improves reliability of decision making and reduces the effort involved in training data-driven deep network models.

Title: Fast approximate bi-objective Pareto sets with quality bounds. Abstract: We present and empirically characterize a general, parallel, heuristic algorithm for computing small \(\epsilon \)-Pareto sets. A primary feature of the algorithm is that it maintains and improves an upper bound on the \(\epsilon \) value throughout the algorithm. The algorithm can be used as part of a decision support tool for settings in which computing points in objective space is computationally expensive. We use the bi-objective TSP and graph clearing problems as benchmark examples. We characterize the performance of the algorithm through \(\epsilon \)-Pareto set size, upper bound on \(\epsilon \) value provided, true \(\epsilon \) value provided, and parallel speedup achieved. Our results show that the algorithm’s combination of small \(\epsilon \)-Pareto sets and parallel speedup is sufficient to be appealing in settings requiring manual review (i.e., those that have a human in the loop) or real-time solutions.

Title: Learning MAX-SAT from contextual examples for combinatorial optimisation Abstract: Combinatorial optimisation problems are ubiquitous in artificial intelligence. Designing the underlying models, however, requires substantial expertise, which is a limiting factor in practice. The models typically consist of hard and soft constraints, or combine hard constraints with an objective function. We introduce a novel setting for learning combinatorial optimisation problems from contextual examples. These positive and negative examples show – in a particular context – whether the solutions are good enough or not. We develop our framework using the MAX-SAT formalism as it is a simple yet powerful setting having these features. We study the learnability of MAX-SAT models. Our theoretical results show that high-quality MAX-SAT models can be learned from contextual examples in the realisable and agnostic settings, as long as the data satisfies an intuitive “representativeness” condition. We also contribute two implementations based on our theoretical results: one leverages ideas from syntax-guided synthesis while the other makes use of stochastic local search techniques. The two implementations are evaluated by recovering synthetic and benchmark models from contextual examples. The experimental results support our theoretical analysis, showing that MAX-SAT models can be learned from contextual examples. Among the two implementations, the stochastic local search learner scales much better than the syntax-guided implementation while providing comparable or better models.

Title: When move acceptance selection hyper-heuristics outperform Metropolis and elitist evolutionary algorithms and when not Abstract: Selection hyper-heuristics (HHs) are automated algorithm selection methodologies that choose between different heuristics during the optimisation process. Recently, selection HHs choosing between a collection of elitist randomised local search heuristics with different neighbourhood sizes have been shown to optimise standard unimodal benchmark functions from evolutionary computation in the optimal expected runtime achievable with the available low-level heuristics. In this paper, we extend our understanding of the performance of HHs to the domain of multimodal optimisation by considering a Move Acceptance HH (MAHH) from the literature that can switch between elitist and non-elitist heuristics during the run. In essence, MAHH is a non-elitist search heuristic that differs from other search heuristics in the source of non-elitism.

Title: Value functions for depth-limited solving in zero-sum imperfect-information games Abstract: We provide a formal definition of depth-limited games together with an accessible and rigorous explanation of the underlying concepts, both of which were previously missing in imperfect-information games. The definition works for an arbitrary (perfect recall) extensive-form game and is not tied to any specific game-solving algorithm. Moreover, this framework unifies and significantly extends three approaches to depth-limited solving that previously existed in extensive-form games and multiagent reinforcement learning but were not known to be compatible. A key ingredient of these depth-limited games is value functions. Focusing on two-player zero-sum imperfect-information games, we show how to obtain optimal value functions and prove that public information provides both necessary and sufficient context for computing them. We provide a domain-independent encoding of the domains that allows for approximating value functions even by simple feed-forward neural networks, which are then able to generalize to unseen parts of the game. We use the resulting value network to implement a depth-limited version of counterfactual regret minimization. In three distinct domains, we show that the algorithm's exploitability is roughly linearly dependent on the value network's quality and that it is not difficult to train a value network with which depth-limited CFR's performance is as good as that of CFR with access to the full game.

Title: Answering regular path queries mediated by unrestricted SQ ontologies Abstract: A prime application of description logics is ontology-mediated query answering, with the query language often reaching far beyond instance queries. Here, we investigate this task for positive existential two-way regular path queries and ontologies formulated in the expressive description logic SQu, where SQu denotes the extension of the basic description logic ALC with transitive roles (S) and qualified number restrictions (Q) which can be unrestrictedly applied to both non-transitive and transitive roles (⋅u). Notably, the latter is usually forbidden in expressive description logics. As the main contribution, we show decidability of ontology-mediated query answering in that setting and establish tight complexity bounds, namely 2ExpTime-completeness in combined complexity and coNP-completeness in data complexity. Since the lower bounds are inherited from the fragment ALC, we concentrate on providing upper bounds. As main technical tools we establish a tree-like countermodel property and a characterization of when a query is not satisfied in a tree-like interpretation. Together, these results allow us to use an automata-based approach to query answering.

Title: Improved local search for the minimum weight dominating set problem in massive graphs by using a deep optimization mechanism Abstract: The minimum weight dominating set (MWDS) problem is an important generalization of the minimum dominating set problem with various applications. In this work, we develop an efficient local search scheme that can dynamically adjust the number of added and removed vertices according to the information of the candidate solution. Based on this scheme, we further develop three novel ideas to improve performance, resulting in our so-called DeepOpt-MWDS algorithm. First, we use a new construction method with five reduction rules to significantly reduce massive graphs and construct an initial solution efficiently. Second, an improved configuration checking strategy called CC2V3+ is designed to reduce the cycling phenomenon in local search. Third, a general perturbation framework called deep optimization mechanism (DeepOpt) is proposed to help the algorithm avoid local optima and to converge to a new solution quickly.

Title: Fair and efficient allocation with few agent types, few item types, or small value levels Abstract: In fair division of indivisible goods, allocations that satisfy fairness and efficiency simultaneously are highly desired but may not exist or, even if they do exist, are computationally hard to find. Conditions under which such allocations, or allocations satisfying specific levels of fairness and efficiency simultaneously, can be efficiently found have thus been explored. Following this line of research, this study is concerned with the problem in a high-multiplicity setting where instances come with certain parameters, including agent types, item types, and value levels. Particularly, we address two computational problems. First, we wish to compute fair and Pareto-optimal allocations, w.r.t. any of the common fairness criteria: proportionality, maximin share, and max-min fairness. Second, we seek to find a max-min fair allocation that is efficient in the sense of maximizing utilitarian social welfare. We show that the first problem is tractable for most of the fairness criteria when the number of item types is fixed, or when at least two of the three parameters are fixed. For the second problem, we model it as a bi-criteria optimization problem that is solved approximately by determining an approximate Pareto set of bounded size. Our results are obtained based on dynamic programming and linear programming approaches. Our techniques strengthen known methods and can be potentially applied to other notions of fairness and efficiency.

Title: G-LIME: Statistical learning for local interpretations of deep neural networks using global priors Abstract: To explain the prediction result of a Deep Neural Network (DNN) model based on a given sample, LIME [1] and its derivatives have been proposed to approximate the local behavior of the DNN model around the data point via linear surrogates. Though these algorithms interpret the DNN by finding the key features used for classification, the random interpolations used by LIME would perturb the explanation result and cause the instability and inconsistency between repetitions of LIME computations. To tackle this issue, we propose G-LIME that extends the vanilla LIME through high-dimensional Bayesian linear regression using the sparsity and informative global priors. Specifically, with a dataset representing the population of samples (e.g., the training set), G-LIME first pursues the global explanation of the DNN model using the whole dataset. Then, with a new data point, G-LIME incorporates an modified estimator of ElasticNet-alike to refine the local explanation result through balancing the distance to the global explanation and the sparsity/feature selection in the explanation. Finally, G-LIME uses Least Angle Regression (LARS) and retrieves the solution path of a modified ElasticNet under varying ℓ1-regularization, to screen and rank the importance of features [2] as the explanation result. Through extensive experiments on real world tasks, we show that the proposed method yields more stable, consistent, and accurate results compared to LIME.

Title: Multi resource allocation with partial preferences Abstract: We provide efficient, fair, and non-manipulable mechanisms for the multi-type resource allocation problems (MTRAs) and multiple assignment problems where agents have partial preferences over bundles consisting of multiple divisible items. We uncover a natural reduction from multiple assignment problems to MTRAs, which preserves the properties of MTRA mechanisms. We extend the well-known random priority (RP) and probabilistic serial (PS) mechanisms to MTRAs with partial preferences as multi-type PS (MPS) and multi-type RP (MRP) and propose a new mechanism, multi-type general dictatorship (MGD), which combines the ideas of MPS and MRP. We show that for the unrestricted domain of partial order preferences, unfortunately, no mechanism satisfies both sd-efficiency and sd-envy-freeness, even as they each satisfy different weaker notions of the desirable properties of efficiency, fairness, and non-manipulability we consider. Notwithstanding this impossibility result, our main message is positive: When agents' preferences are represented by acyclic CP-nets, MRP satisfies ex-post-efficiency, sd-strategyproofness, and upper invariance, while MPS satisfies sd-efficiency, sd-envy-freeness, ordinal fairness, and upper invariance, recovering the properties of RP and PS; the MGD satisfies sd-efficiency, equal treatment of equals, and decomposability under the unrestricted domain of partial preferences. We introduce a natural domain of bundle net preferences, which generalizes previously studied domain restrictions of partial preferences for multiple assignment problems and is incomparable to the domain of acyclic CP-nets. We show that MRP and MPS satisfy all properties of the RP and PS under bundle net preferences as well.

Title: A kinematics principle for iterated revision Abstract: In probabilistic belief revision, the kinematics principle is a well-known and powerful principle which ensures that changing the probabilities of facts does not change unnecessarily conditional probabilities. A related principle, the principle of conditional preservation, has also been one of the main guidelines for the axioms of iterated belief revision in the seminal paper by Darwiche and Pearl. However, to date, a fully elaborated kinematics principle for iterated revision has not been presented. We aim to fill this gap in this paper by proposing a qualitative kinematics principle for iterated revision of epistemic states represented by total preorders. As new information, we allow sets of conditional beliefs, going far beyond the current state of the art of belief revision. We introduce a qualitative conditioning operator for total preorders which is compatible with conditioning for Spohn's ranking functions as far as possible, and transfer the technique of c-revisions to total preorders to provide a proof of concept for our kinematics principle at least for special revision scenarios.

Title: From the Cooperative to the Coercive: Digital Platforms and the Antinomies of CSCW Abstract: ABSTRACT New technologies have the potential to expand the spatial and temporal bounds of various social exchanges. It therefore becomes possible to incorporate more people and territories, and enlarge networks of cooperation of various forms for various ends. This, indeed, has been the premise of contemporary globalization: affordable, ubiquitous computing technologies to create an economy that works as a unit in real time at a planetary scale. Long before computing networks provided the technological infrastructure for globalization, or computer science attempted to understand cooperation, the social sciences grappled with the question of why and how people cooperate. Economists, for instance, have viewed it in terms of the emergence of a division of labour, supported by technology, to fulfill various need. However, any narrative suggesting the frictionless incorporation of different people and territories into the global economy overlooks how social contracts, which political theorists and sociologists highlight as the basis for cooperation, can be asymmetric due to power differentials between social actors. The asymmetry is another manifestation of the socio-spatial unevenness that has historically characterized capitalism. The unevenness in our times has not gone unnoticed as evidenced by efforts to address the needs of those on the wrong side of the digital divide, or at the bottom-of-the-pyramid, with such initiatives as frugal innovation, or technologies for inclusive development. After a brief survey of such efforts since the turn of the millennium, the presentation will draw on ongoing research in India to describe how digital platforms mediate the last-mile delivery of services, from producers to their customers, by relying on increasing numbers of smartphone-armed gig worker ‘partners’. The presentation will argue that, as platform companies deploy technology and rely on asymmetrical contractual relationships to reconfigure the division of labour in service-delivery, workers find themselves in unequal partnerships to experience CSCW as Computer Supported Coercive Work.

Title: Facilitating Continuous Text Messaging in Online Romantic Encounters by Expanded Keywords Enumeration Abstract: ABSTRACT An increasing number of people are looking for romantic partners online. Many of them first converse online before deciding whether or not to meet in-person. However, it is often challenging to have smooth and continuous conversations online with someone who they have never met in person. To handle this problem, we built a proof-of-concept system, Tomi, that dynamically suggests various conversation topic seeds related to the latest received messages in real-time. It selects a keyword from an incoming message and returns five contextually relevant topic seeds. In a qualitative study with eight dyads that simulated the common setting of online match-making, users could continue their conversations either by directly or indirectly utilizing the suggested topic seeds. Also, our system boosted their confidence during the chat. Lastly, we analyzed the trade-offs between several design alternatives and presented our reflections on a system supporting continuous conversations with diverse topics.

Title: “I am not a YouTuber who can make whatever video I want. I have to keep appeasing algorithms”: Bureaucracy of Creator Moderation on YouTube Abstract: ABSTRACTRecent HCI studies have recognized an analogy between bureaucracy and algorithmic systems; given platformization of content creators, video sharing platforms like YouTube and TikTok practice creator moderation, i.e., an assemblage of algorithms that manage not only creators’ content but also their income, visibility, identities, and more. However, it has not been fully understood as to how bureaucracy manifests in creator moderation. In this poster, we present an interview study with 28 YouTubers (i.e., video content creators) to analyze the bureaucracy of creator moderation from their moderation experiences. We found participants wrestled with bureaucracy as multiple obstructions in re-examining moderation decisions, coercion to appease different algorithms in creator moderation, and the platform's indifference to participants’ labor. We discuss and contribute a conceptual understanding of how algorithmic and organizational bureaucracy intertwine in creator moderation, laying a solid ground for our future study.

Title: The Needs of Grandparents and Grandchildren in a Socially and Geographically Distanced World: A Case Study Abstract: ABSTRACTGrandparents and grandchildren are important parts of each other's lives. However, they may not be able to always be co-located to interact and share activities. We conducted case studies to explore how grandparents and grandchildren use video calls to spend time with each other, when being co-located is not possible for extended periods of time. We found that aspects such as framing and camera work, unilateral and shared activities, and contextual awareness are especially relevant for this type of interaction, especially for children. We additionally provide potential design and exploration avenues for future research.

Title: The Co-Creation Space: Supporting Asynchronous Artistic Co-creation Dynamics Abstract: ABSTRACT Artistic co-creation empowers communities to shape their narratives, however HCI research does not support this multifaceted discussion and reflection process. In the context of community opera, we consider how to support co-creation through the design, implementation, and initial evaluation of the Co-Creation Space (CCS) to help community artists 1) generate raw artistic ideas, and 2) discuss and reflect on the shared meaning of those ideas. This work describes our user-centered process to gather requirements and design the tool, and validates its’ usability with 6 community opera participants. Our findings support the value of our tool for group discussion and personal reflection during the creative process.

Title: Work of Fiction: Using Speculative Design to Deliberate on the Future of Hiring Abstract: ABSTRACTAI recruitment systems are increasingly being deployed to automate the hiring process. This is especially true for companies that hire large numbers of people. AI recruitment software is used in all phases of the process, from receiving and sorting applications, candidate screening, selection process, and communication of results to applicants. While some argue that AI recruitment may remove interviewer bias, others believe there is a danger of dehumanizing the interviewees’ experience. In either case, AI recruitment is set to ignite a significant paradigm shift in hiring practices. For this poster, we discuss the results of a participatory speculative design project conducted to engage both HR recruiters and job applicants to understand their concerns. Specifically, we present six scenarios and a diegetic prototype that were used to familiarize participants with the AI recruitment process. We discuss how these served as catalysts for meaningful conversations about an often obfuscated topic.

Title: “Hide Your Video, Show Your Action!” Investigating a New Video Conferencing Interface for Virtual Studying Abstract: ABSTRACTWith the advent of COVID-19, new virtual social activities arose. These activities include virtual studying, which is studying while joining a video conference. Virtual studying is different from most virtual activities in that users try to minimize unnecessary interactions while sharing their presence through video streaming. Here, video streaming that runs in the background can cause problems such as invasion of privacy and excessive self-awareness. In this paper, we aim to investigate whether a new video conferencing interface that reduces video explicitness but detects important actions can mitigate the problems of video streaming and still deliver users’ presence in virtual studying. To this end, we designed a research prototype in three versions: blurred video version, small video version, and no video version. All versions were provided with an activity recognizer that detects absence, leaning, and using a smartphone. To evaluate the feasibility of the design, we conducted a user study where four virtual studying teams used all three versions of the prototype and participated in an interview. Our study explored the effects of new design strategies for virtual studying, which is a new virtual activity that focuses on sharing presence.

Title: Understanding Hate Group Videos on YouTube Abstract: ABSTRACTHate content warning: This paper contains hate ideology content that may cause discomfort. As the largest video-sharing platform, YouTube has been known for hosting hate ideology content that could lead to between-group conflicts and extremism. Research has examined search algorithms and the creator-fan networks related to radicalization videos on YouTube. However, there is little grounded theory analysis of videos of hate groups to understand how hate groups present to the viewers and discuss social problems, solutions, and actions. This work presents a preliminary analysis of 96 videos using open-coding and affinity diagramming to identify common video styles created by the U.S. hate ideology groups. We also annotated hate videos’ diagnostic, prognostic, and motivational framing to understand how the hate groups utilize video-sharing platforms to promote collective actions.

Title: Exploring Users’ Experiences of “Suggested Posts” in Social Media Through the Lens of Social Networking and Interactions Abstract: ABSTRACT Many social media feeds are incorporated with “Suggested Posts,” an AI-infused feature that recommends algorithmically selected posts from accounts that a user does not follow. While such recommendations might increase user engagement with the platforms, there lacks understanding on how such system-driven recommendations influence the primary purpose of social media use: social interactions. We conducted semi-structured interviews with 12 Instagram users to investigate how they perceive the “Suggested Posts” feature in relation to their in-app social interaction practices. Our findings reveal that suggested posts acted as a double-edged sword for communication with their actual friends, acted as an obstacle for digital self-presentation, and acted as an unwanted force that compel users to increase social groups. Based on the findings, we discuss the importance of considering user agency over social interactions in designing user experiences of recommendations provided in social media.

Title: Investigating Human Factors in Willingness to Donate to the Small-scale Non-profit Organizations in Bangladesh Abstract: ABSTRACTThe success of small-scale non-profit organizations (NPOs) highly depends on receiving donations from people through connecting with them, where the online presence can play a significant role. However, willingness to make such donations circumvents multiple important human factors that are little studied in the literature. Therefore, in this study, we survey 42 people in a developing country (Bangladesh) to investigate their willingness to donate to small-scale NPOs. Our findings reveal that people show a willingness to donate to small-scale NPOs that have an online presence and their lack of trust gets significantly reduced if those organizations have an online presence. We also find out several influential factors that inspire people to donate. Further, based on the feedback we receive, we identify design features that should be present in the online presence of small-scale NPOs.

Title: A Critical Literature Review for Equal Participation in Human-Animal Interactions in Design Abstract: ABSTRACT Animals have been studied in the CSCW, such as in studies about animal welfare, pet-advocacy groups, pet video chat, and multispecies interaction. Animal-Computer Interaction (ACI) is the field where studies with animals and technology are at the centre. However, within the CSCW and the ACI field, the equal participation from the animals’ viewpoint remains relatively human-centric, and how humans can collaborate with nonhuman animals remain underexplored. Research beyond human-centrism in other fields puts equal participation of nonhuman animals at the centre with the intention of equal inclusion. Thus, this poster introduces the initial results from a literature review on the previously published work in animal-inclusive and equity-oriented research fields with the purpose of opening a discussion on equity perspectives and equal participation of nonhuman animals in the CSCW work.

Title: Investigating Correction Effects of Different Modalities for Misinformation about COVID-19 Abstract: ABSTRACTMisinformation presented in different modalities about the COVID-19 pandemic has been prevalent. One approach to reducing the negative effects of misinformation is through corrective information. However, it is possible that people develop counter-attitude towards the corrective information and reaffirm their belief in misinformation, called the boomerang effect. Fewer studies examined how different modes of corrective information about COVID-19 may address the boomerang effect. With a 3-by-3 between-subject experiment design (n = 210), we first presented one of the three modalities of misinformation (text, image, video) to the participants, followed by one of the three modalities of corrective information (text, image, video) to examine the effect of the corrective information. The results showed that there was no boomerang effect after correction in all modalities, indicating that all corrective information successfully reduced participants’ perceived credibility and potential action for misinformation. In the post-hoc analysis, the correction in the video mode worked best on text misinformation. Our results also suggested that image misinformation worked least effectively in terms of conveying misinformation.

Title: CAMPUS: A University Crowdsourcing Platform for Reporting Facility, Status Update, and Problem Area Information Abstract: ABSTRACT This paper presents CAMPUS, a crowdsourcing platform that enables its users to report and view information about a university’s facilities, real-time status, and problem areas via a mobile web app. The results of our preliminary evaluation indicate that this platform had the potential to inform them about its focal topics, and that they would be motivated to use it if it were deployed on a larger scale. However, the study also highlighted the challenge of dealing with users’ mutually redundant reports. Nevertheless, CAMPUS exhibits a strong potential for enabling individuals on a campus to collaboratively inform each other about campus-related information; and our future work will show how individuals on campus leverage this system on a daily basis.

Title: Misleading Tweets and Helpful Notes: Investigating Data Labor by Twitter Birdwatch Users Abstract: ABSTRACT In response to concerns about misleading content on social media, Twitter launched the “Birdwatch” initiative that allows volunteers to label and add context to tweets. We study data from Birdwatch to understand how users are performing “data labor” for Twitter, with implications for other platforms that are similarly reliant on data labor. We conduct computational analyses of Birdwatch text data and perform machine learning experiments to see how Birdwatch contributions might be used for classification. We find that Birdwatch users discuss distinct topics in domains like politics and news. While using Birdwatch data for content-only predictions may provide only a small amount of predictive power, in some cases Birdwatch data may be able to support ML systems. Furthermore, we see that the continuous flow of Birdwatch contributions provides great value in terms of supporting a “guess most frequent“ baseline for classifying Twitter content.

Title: The Components of Trust for Collaborating With AI Colleagues Abstract: ABSTRACT AI technologies are capable of improving the performance and productivity of teams in a variety of work contexts. These advantages may be optimized when the AI agent is considered a full team member. A vital component of the agent’s acceptance as a team member or colleague is the degree to which its human coworkers feel they can trust it. To explore what factors affect the perceptions of an AI agent as a trustworthy team member and a legitimate colleague, we interviewed twenty-two professionals representing various work roles. Our results revealed that the following qualities contribute to professionals’ trust in AI as a colleague: a visual presence reflective of coworkers, engagement in feedback loop and team processes through human communication, and the ability for self-development. These findings contribute to the CSCW community by advancing the current understanding of human-AI teaming and informing the design of trustworthy AI agents into the workplace.

Title: Re-imagining Systems in the Realm of Immigration in Higher Education through Participatory Design Abstract: ABSTRACT Students identifying with communities within the realm of immigration, including immigrants, refugees, international students, and undocumented individuals, often face difficulties navigating life in elite institutions of higher education. HCI scholars who have worked with migrant communities have called for future explorations on the role of participatory approaches in helping bring different stakeholders together to design for better technologies and socio-technical systems supporting their needs. In this paper we present preliminary findings from a participatory design study at an elite liberal-arts college in the U.S., exploring the role of an equity-centered hackathon-style event in bringing together different student communities on campus and collaboratively re-imagining today’s technologies and systems. We discuss possible best practices for designing inclusive and safe hackathon-style events for communities in the realm of immigration.

Title: The Metaverse: A Systematic Literature Review to Map Scholarly Definitions Abstract: ABSTRACT Metaverse is the new buzz word in the field of technology, particularly in the media and industry. Studies of virtual spaces have been going on in academia for decades, and industry has been creating such spaces as commercial products. However, the specific term "metaverse" has been used sparingly, and when used, the definitions have been inconsistent. In an effort to understand how researchers are using this term, we conducted a systematic literature review and reviewed articles that used and defined the term metaverse. We categorized the main characteristics that researchers have defined as being properties of the metaverse, finding broad overlaps but divergence in details.

Title: Entering the Techlash: Student Perspectives on Ethics in Tech Job Searches Abstract: ABSTRACT Public coverage of ethical scandals in tech companies often portrays tech workers as uncaring or short-sighted in their ethical consideration. But how do tech students regard the connection between ethics and their future jobs? To explore ethics in the transition between academia and industry, we interviewed graduating computing students at the University of Colorado Boulder about their perspectives on ethical concerns in tech, decisions during job searches, and their ethics education. These interviews revealed that while students may value and understand ethics in tech, their belief that companies do not value ethics makes ethical consideration in the workplace daunting. In response, we suggest improving support for tech workers through academia-industry collaboration and additions to computing ethics curricula to help students stand up for responsible tech.

Title: From Ignoring Strangers’ Solicitations to Mutual Sexting with Friends: Understanding Youth’s Online Sexual Risks in Instagram Private Conversations Abstract: ABSTRACT Online sexual risks pose a serious and frequent threat to adolescents’ online safety. While significant work is done within the HCI community to understand teens’ sexual experiences through public posts, we extend their research by qualitatively analyzing 156 private Instagram conversations flagged by 58 adolescents to understand the characteristics of sexual risks faced with strangers, acquaintances, and friends. We found that youth are often victimized by strangers through sexual solicitation/harassment as well as sexual spamming via text and visual media, which is often ignored by them. In contrast, adolescents’ played mixed roles with acquaintances, as they were often victims of sexual harassment, but sometimes engaged in sexting, or interacted by rejecting sexual requests. Lastly, adolescents were never recipients of sexual risks with their friends, as they mostly mutually participated in sexting or sexual spamming. Based on these results, we provide our insights and recommendations for future researchers. Trigger Warning: This paper contains explicit language and anonymized private sexual messages. Reader discretion advised.

Title: When to Collect Sensitive Category Data? Public Sector Considerations For Balancing Privacy and Freedom from Discrimination in Automated Decision Systems Abstract: ABSTRACTAutomated Decision Systems (ADS) are being used to inform important decisions in government services. Concerns regarding discrimination in ADS have led to the rise of bias mitigation techniques, or data science practices that measure and adjust for disparities based on protected class data. These techniques often require demographic or sensitive category data to both measure discrimination in ADS and process data to mitigate the discriminatory bias. The collection of sensitive category data increases privacy concerns. This preliminary study includes nine semi-structured interviews with data practitioners working in government. The analysis explores what considerations data practitioners in the public sector make when determining best practices for sensitive category data collection and how they engage with the tradeoff between privacy and right to freedom from discrimination. Themes pertinent to social services emerged including the importance of accessibility, reasons for data minimization and the role of reporting structures and historical norms.

Title: Exhibiting Evidence in Remote Courtrooms: Design and Usability Study Abstract: ABSTRACT Remote court proceedings, an intensely collaborative process, are a reality in the pandemic but struggle to support evidence review. Remote proceedings are likely to continue after the pandemic, making it imperative to strengthen evidence viewing and analysis for jurors. Our project addressed this problem by prototyping an assistive application that offered multiple points of view and evaluating it in a small user study. Five participants used the application to review case-specific exhibits in a hypothetical court session while thinking aloud. Participants were satisfied with the application, with almost no facilitation. We also identified the need to make a display of the setting in which the evidence was found more interactive and navigable.

Title: This is Why I Like Them: Exploring the Perceived Appeal of Social Media Influencers vs. Traditional Mass Media Abstract: ABSTRACT Social-media influencers (SMIs) have created a wide range of content and research has shown that SMIs are perceived as credible by media consumers, and advertising by them can lead to higher user engagement than traditional advertising generally achieves. However, the factors that render SMIs more appealing than traditional media from an audience perspective have been underexplored. Using a grounded approach, we looked at the perceived appeal of SMIs of various types vs. traditional media by conducting semi-structured interviews with 20 of the former’s audience members. Our preliminary findings suggest that such appeal can be divided into four types: 1) initiative to summarize takeaways from multiple sources, 2) high independence and low interference, 3) distinctive and diverse networks and connections, and 4) relatability and applicability. This typology of appeal uncovers how SMIs have formed new media use patterns, and hopefully inform the design opportunities of social media platforms.

Title: Screenshot Journey Auditor: A Tool to Support Analysis of Smartphone Media Consumption Journey Using Screenshot Data Abstract: ABSTRACT Taking long series of screenshots for capturing and studying smartphone users’ phone usage and media consumption has recently attracted research attention due to its advantage of capturing rich contextual information from users’ phone use journeys. However, that approach creates a high volume of screenshots that take very considerable time and effort to inspect and annotate, especially when the granularity of analysis is low: such as when distinguishing among media-content units (e.g, single FB posts) and detecting events in them. We therefore developed Screenshot Journey Auditor (SJA), a web application that identifies individual social media posts, and detects news items and other events of interest in them. It then visualizes users’ journeys ‘flow’ among these media-content units. SJA also enables researchers/coders to collaboratively correct detections online. We evaluated SJA with five coders and received positive feedback on how the detections and visualizations made the analysis process more efficient and informative.

Title: "I feel like I need to split myself in half": Using Role Theory to Design for Parents as Caregiving Teams in the Children's Hospital Abstract: ABSTRACTWhen their child is hospitalized, parents take on new caregiving roles, in addition to their existing home and work-related responsibilities. Previous CSCW research has shown how technologies can support caregiving, but more research is needed to systematically understand how technology could support parents and other family caregivers as they adopt new coordination roles in their collaborations with each other. This paper reports findings from an interview study with parents of children hospitalized for cancer treatment. We used the Role Theory framework from the social sciences to show how parents adopt and enact caregiving roles during hospitalization and the challenges they experience as they adapt to this stressful situation. We show how parents experience 'role strain' as they attempt to divide caregiving work and introduce the concept of 'inter-caregiver information disparity.' We propose design opportunities for caregiving coordination technologies to better support caregiving roles in multi-caregiver teams.

Title: Korean Emoticons: Understanding How Subtle Emotional Differences Are Evoked Online Abstract: ABSTRACT Online conversations through text have limitations in expressing emotions that can cause miscommunications across cultures. In this work, we study the Korean emotional expressions in text focusing on how people perceive emotional intentions through the use of emotion-expressing Korean characters. We define them as Korean emoticons (‘ㅋ’, ‘ㅎ’, ‘ㅠ’), onomatopoeic characters often used to express emotions for text-based communication. We examine the participants’ understanding and usage of Korean emoticons by conducting an online survey asking to evaluate emotional contents of given sentences and interviews to explain personal experiences. We found that the different numbers of Korean emoticons used evoke different emotions, and that negative emoticons amplify positive emotions in positive contexts and positive emoticons alleviate negative emotions in negative contexts, while emoticons in neutral contexts have varying impacts depending on the context. We further discuss design implications on how text suggestion tools can support users taking emotional intentions into account.

Title: Designing the "Front": An Overview of Profile Elements on Social Network Sites Abstract: ABSTRACTUser profile is one of the most fundamental approaches on Social Network Sites (SNS) to make an impression on others. Designing user profiles is a process of designing users’ personal "front" [6, 14]. This study investigates profile elements across 110 SNS and identifies 62 elements with two groups of attributes: manually filled by users or automatically generated by the system, providing individual or social information. We annotated the presence of elements and whether they are optional or concealable. The most common elements are Profile Image, Bio Text, Username, Location, and Link to Other Social Media. We found SNS’ proportion of system-generated elements is positively correlated with social elements. SNS can be categorized into "Real Life Mirror" and "Showcase as Need" based on the number of profile elements and can be categorized into "Resume" and "Convenient Social" by the attributes of elements. We discuss the potential use of the data and the implications for future research. This work contributes to CSCW community by offering a snapshot of update-to-date profile design with comprehensive analysis.

Title: Enabling Remote Hand Guidance in Video Calls Using Directional Force Illusion Abstract: ABSTRACT In current video meetings, communicating positions and hand movements is often challenging, as the meeting participants need to rely on gestures and verbal explanations, resulting in inefficient video meetings. We propose a video meeting system that utilizes a novel directional force feedback method so that people can efficiently guide each other’s hand without the need for lengthy verbal explanations. The force feedback method uses asymmetric vibrations to create the illusion of directional pulling forces, which makes it possible to be built as a small and lightweight device. We also developed an online meeting system that allows a user to simply click a location on the remote user’s video to move their hand towards the target location.

Title: Exploring Secondary Teachers’ Needs and Values in Culturally Responsive Teaching Abstract: ABSTRACT Culturally Responsive Teaching (CRT) is educational practices that acknowledge, celebrate, and incorporate students’ identities and backgrounds into the classroom. Technologies can be used to ease the challenges teachers face with implementing CRT-related practices including inadequate guidance and knowledge of students’ cultures. Value sensitive design (VSD) accounts for human values in the design of technology. VSD can be used in the design of technological systems for helping CRT-related practices. However, the CRT context poses unique challenges due to the interactions between teachers’ values and other socio-material factors such as students’ values, community characteristics, and available resources. This preliminary study uses a low-fidelity cultural tool to elicit teachers’ needs and values on CRT from various cultural backgrounds and educational settings. The results of the interview analysis provide an initial understanding of how needs, values, socio-material factors, and educational contexts are intertwined, and how they can be used to inform the design of technologies that assist teachers with CRT.

Title: Action-a-Bot: Exploring Human-Chatbot Conversations for Actionable Instruction Giving and Following Abstract: ABSTRACTConversation serves as one critical mechanism for knowledge-sharing and instruction-giving in collaborative work. Conversation allows people to take turns to make contributions, plan joint actions, align shared understanding of work status and resolve action failure. However, when such collaboration involves non-human AI actors like chatbots, there is a lack of understanding of how human participants may respond to the chatbot’s prompts and guidance, and whether the interaction can similarly improve the actionability of instructions given to people. In this study, we prototyped a chatbot system, ActionaBot for providing task instructions to novice workers, and conducted an initial study to explore its effects on procedural instruction giving and following. Our results indicate that, novices although might perceive instructions to be inactionable due to prior experience and how instructions were authored, they were able to follow conversational guidance and willing to adapt to the chatbot through turn-taking to calibrate working states back-and-forth. Besides, users’ awareness of the work status increased with the conversational prompts from the chatbot.

Title: Repurposing AI in Dating Apps to Augment Women’s Strategies for Assessing Risk of Harm Abstract: ABSTRACT In this paper we present emerging findings from an interview study with women in North America about how AI could be designed to prevent online dating-facilitated violence by augmenting their strategies for risk awareness. The study is motivated by gender disparities in harm through online dating, and the relative absence of dating app designs that prioritize women. Findings show that women are receptive to the notion of AI that augments their existing strategies for assessing risk of harm with meeting a particular user face-to-face. They outline various physical and non-physical harms that they attempt to reduce uncertainty about, and a range of data points as indicators of risk that could serve as the basis for a risk awareness AI model. Due to subjectivity in what is considered an indicator of risk, current findings suggest that risk awareness AI, if implemented, should allow women to train their own models.

Title: An Online Memorial for Coping with Mass Shooting Tragedy by Combining Participatory Memory with Participatory Design of AI Use Cases Abstract: ABSTRACT Mass shootings occur in America at an alarming rate. There is opportunity to intervene in this problem by designing technologies that support affected communities in processing gun violence tragedy. In this paper we report on the design of an online memorial for a mass shooting that affected our university’s local community. We demonstrate an alternative approach to online memorials that blends participatory memory with participatory design through remembrance artifacts that represent ideas for gun violence prevention technologies that could have prevented the tragedy and that may prevent future tragedies. We demonstrate participatory memory + design with our memorial called the OUrchive that supports retrospective and prospective reflection on the Oxford High School shooting through designing new use cases for AI to prevent mass shootings. Early community involvement suggests that the OUrchive supports personal reconciliation with tragedy by channeling trauma towards public discourse about potential solutions to gun violence.

Title: A Privacy Paradox? Impact of Privacy Concerns on Willingness to Disclose COVID-19 Health Status in the United States Abstract: ABSTRACT Privacy concerns around sharing personal health information are frequently cited as hindering COVID-19 contact tracing app adoption. We conducted a nationally representative survey of 304 adults in the United States to investigate their attitudes towards sharing two types of COVID-19 health status (COVID-19 Diagnosis, Exposure to COVID-19) with three different audiences (Anyone, Frequent Contacts, Occasional Contacts). Using the Internet User’s Information Privacy Concern (IUIPC) scale, we were able to identify the effect of different types of privacy concerns on sharing this information with various audiences. We found that privacy concerns around data Collection predicted lower willingness to share either type of health status to all of these audiences. However, desire for Control and for Awareness of data practices increased willingness to share health information with certain audiences. We discuss the implications of our findings.

Title: Designing Food Delivery Gig-platforms for Courier Needs: the Case of Batched Orders Abstract: ABSTRACT The food delivery gig-platforms such as Uber Eats and Deliveroo increasingly batch orders, i.e., couriers deliver multiple orders in one trip to optimize order delivery. Prior studies investigate the topic of batching orders from a theoretical perspective, i.e., optimization, without taking contextual elements and worker preferences into account. In our paper, we elaborate on the couriers’ perspectives and needs in batched order delivery based on a qualitative study. We identify several criteria used by couriers to accept and reject the batched orders. We propose an alternative order batching solution that supports courier needs.

Title: Decolonial and Postcolonial Computing Research: A Scientometric Exploration Abstract: ABSTRACT Decolonial and postcolonial computing scholars focus on the relationship between coloniality and technology. While many recent empirical and design studies have adopted these theoretical lenses, these conversations are often disconnected. Through a systematic literature review, we seek to understand patterns within and between decolonial computing and postcolonial computing. As an early step toward that objective, this poster presents results from our preliminary scientometric exploration of 115 papers’ metadata and discusses research trends and popular publication venues in these areas. Using citation network analysis, we found smaller communities in decolonial and postcolonial computing scholarship based on their use of theoretical frameworks, objectives, types of papers, authors’ collaboration and affiliations, and research sites and populations. We conclude by discussing future research directions to bring these communities into conversations with each other.

Title: Understanding the Effects of Profile Display in Multilingual Computer-Mediated Team Formation Abstract: ABSTRACTCollaborating across language barriers is cognitively, communicatively and socially taxing, which may disincentivize people to continue working in a language-diverse team after experiencing it. It remains unclear how socially displaying potential collaborators’ language and personal profiles in advance may affect multilingual team formation especially in computer-mediated virtual environments. We conducted an online study with native English speakers and native Japanese speakers in Gather Town. Participants were asked to form teams and complete a slogan generation task under one of the following conditions - no profile display, constant profile display, and adaptive profile display to supplement language and personal cues. We studied how participants’ searching cost for partners and their attitude towards multilingual teamwork shifted under these conditions. Our findings reveal the use of different team formation strategies depending on language backgrounds and profile display designs. We seek to understand language-technological effects on team formation, and explore designs to facilitate multilingual diversity in online teamwork.

Title: CO-oPS: A Mobile App for Community Oversight of Privacy and Security Abstract: ABSTRACT Smartphone users install numerous mobile apps that require access to different information from their devices. Much of this information is very sensitive, and users often struggle to manage these accesses due to their lack of tech expertise and knowledge regarding mobile privacy. Thus, they often seek help from others to make decisions regarding their mobile privacy and security. We embedded these social processes in a mobile app titled “CO-oPS” ("Community Oversight for Privacy and Security"). CO-oPS allows trusted community members to review one another’s apps installed and permissions granted to those apps. Community members can provide feedback to one another regarding their privacy behaviors. Users are also allowed to hide some of their mobile apps that they do not like others to see, ensuring their personal privacy.

Title: Mi Casa es Su Casa (“MiSu”): A Mobile App for Sharing Smart Home Devices with People Outside The Home Abstract: ABSTRACT As smart devices are becoming commonplace in the home, people have begun sharing access to these devices with people beyond the home. However, the "all-or-nothing" approach to access control taken by most smart home applications may be insufficient for use cases that involve others outside of the home. Therefore, we developed “MiSu” an Android and iOS app that allows smart home homeowners to share their devices (e.g., Ring doorbell, security alarm, smart door lock, smart light bulb) with people outside of their home to control what, when, and how they can engage with the smart devices. MiSu provides options for fine-grain access control, the ability for guests to control smart homes using their own device and login, and provides homeowners real-time logs where they can view all actions taken by guests invited to interact with their smart homes.

Title: DesertWoZ: A Wizard of Oz Environment to Support the Design of Collaborative Conversational Agents Abstract: ABSTRACTThe use of conversational agents to support team-based training in virtual environments is a nascent area of research within the Computer Supported Collaborative Work (CSCW) and Conversational User Interface (CUI) communities. To support this research, we developed a Wizard of Oz (WoZ) system enabling experimenters to play the role of the conversational agent who must work with a team of three human participants to herd together a group of autonomous robots (target agents) within a virtual task environment. The experimenter can select from one of several tailored responses. The aim of this environment is to determine whether the identified responses facilitate effective team performance and to collect training data to build a conversational agent capable of engaging in team based virtual learning environments.

Title: DAO-Analyzer: Exploring Activity and Participation in Blockchain Organizations Abstract: ABSTRACT Decentralized Autonomous Organizations (DAOs) are a new kind of organization that relies on blockchain software to govern their projects. Typically, DAO members may put forward and vote on proposals. For instance these proposals may consist on someone doing some tasks in exchange for a share of the DAO crypto-funds. In recent times, DAOs have gained a remarkable adoption, and yet they are still understudied by the academic literature. In this work, we present a visual analytics tool to study DAO activity focusing on their participation and temporal evolution. Our tool will hopefully help to stimulate research on this new kind of online community and collaborative software.

Title: MOSafely, Is that Sus? A Youth-Centric Online Risk Assessment Dashboard Abstract: ABSTRACTCurrent youth online safety and risk detection solutions are mostly geared toward parental control. As HCI researchers, we acknowledge the importance of leveraging a youth-centered approach when building Artificial Intelligence (AI) tools for adolescents online safety. Therefore, we built the MOSafely, Is that ‘Sus’ (youth slang for suspicious)? a web-based risk detection assessment dashboard for youth (ages 13-21) to assess the AI risks identified within their online interactions (Instagram and Twitter Private conversations). This demonstration will showcase our novel system that embedded risk detection algorithms for youth evaluations and adopted the human–in–the loop approach for using youth evaluations to enhance the quality of machine learning models.

Title: Mindful Garden: Supporting Reflection on Biosignals in a Co-Located Augmented Reality Mindfulness Experience Abstract: ABSTRACT We contribute Mindful Garden, an Augmented Reality Lens for Co-Located Mindfulness. HCI research has increasingly supported designing technology to support mindfulness. Augmented reality and sensors which detect biosignals both have the potential to support creating mindful experiences, by transforming people’s environments into more relaxing spaces and offering some feedback to help people make sense of their physio-psychological states. We will demo Mindful Garden, a system for supporting reflection on biosignals in a mindfulness experience where two people are physically co-located. Mindful Garden has one person guide the other through meditation, representing the guided individual’s biosignals as flowers in a shared augmented reality environment. We leverage Snap Spectacles and a Muse 2 headband, showing the promise of ready-to-use consumer technologies for the purpose of mindfulness and well-being. To overcome technical limitations in accessing, our demo illustrates a novel pipeline for real-time biomarker data streaming from the Muse 2 to the AR lens in the Snap Spectacles.

Title: Consent: A Research and Design Lens for Human-Computer Interaction Abstract: ABSTRACT Consent has become an important concept across multiple areas within HCI/CSCW, community advocacy work, and the tech industry, for understanding social computing problems and designing safe and agentic computer-mediated communication. Recent research has studied consent in various topics, such as online-to-offline interaction and harm, data privacy and security, research ethics, and human-robot interaction. The goal of this panel is to bring together researchers and practitioners to discuss how consent has been defined and studied within HCI and adjacent fields, and how cross-field discourse around consent can inform future work that pursues safe and equitable computing. We aim to introduce consent as a multifaceted research and design lens to the HCI and CSCW community and illuminate ways that consent can contribute to better understanding or re-imagination of contemporary research interests. Lastly, the panel aims to spark cross-field communication around consent to identify latent connections across research topics and foster synergistic collaborations.

Title: Reconsidering Accountability in the Present and Future of Work Abstract: ABSTRACTThere is a growing effort within CSCW and related fields to understand the effects of artificial intelligence, algorithmic management, and other contemporary and near-future transformations of digitally-mediated work. This panel builds on this effort by asking three guiding questions: how should academics more fully account for the consequences of new transformations in the workplace, how should we engage with workers who are already accounting for the ways that these systems impact their labor conditions and experience of work, and how should we be made accountable to workers for the design interventions we propose or deploy. The panelists will discuss topics including the impact of AI technology on frontline and essential workers, the transformative effect of algorithmic management on organizational practices within and beyond platform work, and how demands and efforts to attain algorithmic transparency by platform/gig-workers both reveal power imbalances within the workplaces and point to avenues for worker advocacy. We will also draw on scholarship that analyzes the shifting labor conditions of CSCW to situate contemporary concerns within the field's disciplinary history. We aim to generate conversation about the ways that these forms of accountability have already been unfolding, identify shared themes and concerns across parallel topics, and identify future directions for researchers moving forward in this space.

Title: Envisioning Identity: The Social Production of Human-Centric Computer Vision Systems Abstract: ABSTRACTComputer vision technologies have been increasingly scrutinized in recent years for their propensity to cause harm. Human-centric computer vision, systems designed to interpret visual data about humans for a variety of tasks, are perceived as particularly high risk. Broadly, the harms of human-centric computer vision focus on demographic biases (favoring one group over another) and categorical injustices (through erasure, stereotyping, or problematic labels). Prior work has focused on both uncovering these harms and mitigating them, through, for example, better dataset collection practices and guidelines for more contextual data labeling. This research has largely focused on understanding discrete computer vision artifacts, such as datasets or model outputs, and their implications for specific identity groups or privacy. There is opportunity to further understand how human identity is embedded into human-centric computer vision not only across these artifacts, but also across the network of human workers who shape computer vision systems. My dissertation focuses on understanding how human identity is conceptualized across three different “layers” of computer vision: (1) at the artifact layer, where the classification ontology is deployed, in the form of datasets and model inputs and outputs; (2) at the development layer, where social decisions are made about how to implement models and annotations by traditional tech workers; and (3) at the annotation layer, where technical specs are applied, often by human data workers. I will highlight where identity across these layers do not align and, as a result, are not necessarily predicting identity accurately or even as intended by practitioners.

Title: Systems for Socially-Challenging Information Management Abstract: ABSTRACT In collaborative settings, information management becomes even more challenging. In addition to the classic challenge of information management (i.e. information overload), collaboration also introduces social barriers between group members; this entails asymmetric roles, where different stakeholders have different levels of motivation, knowledge, social capital, authority, or time. To successfully manage their information, groups of people need to coordinate their needs and resources. In my thesis, I focus on building interfaces for lowering social burdens to negotiate and communicate the distribution of resources. I develop coordination systems that guide interaction for collaborative information-management tasks. I first introduce a system called Ziva, which supports interdisciplinary teams by coordinating information exchange. Then, I describe TaskLight, which aims to support information management during the help-seeking process. Lastly, I present challenges of coordinating tool preferences and a resulting system called CollaboRanger.

Title: Peer-Producing a Common Knowledge Resource for Personal Science Abstract: ABSTRACT Personal science, the practice of using empirical methods to explore personal questions, has potential to advance discovery in clinical and public health. Yet, this branch of grassroots citizen science experiences important entry barriers: People tend to start their projects from scratch, and have no structured way of accessing community knowledge. In order to accumulate knowledge, similar communities in other domains have successfully made use of peer production technologies, a form of decentralized creation relying on self-organizing communities known from Wikipedia or open source software. To date, literature on implementing peer production approaches in citizen science is rare, and the particularities of personal science, notably the lack of an epistemic need to collaborate, impose barriers to using off-the-shelf approaches. In order to fill this gap, this dissertation is devoted to exploring peer production as a means to create a collective knowledge resource for the personal science community. Specifically, I investigate 1) how citizen science projects implement peer production characteristics by creating a working model from literature and applying it to case studies; and 2) how to enable the peer production of knowledge in the personal science community by engaging in a participatory design approach with the community. In doing so, this work contributes to further the understanding of deepening participation in citizen science, to harness collective intelligence and empower communities to address undone research.

Title: Sociotechnical Risk in Sustaining Digital Infrastructure Abstract: ABSTRACT Significant risks to our shared digital infrastructure—communication systems, servers, and applications—can be identified by examining the participation dynamics of the organizations which produce that infrastructure. Exploration of these production communities reveals the deeply contingent processes of collective action that sustain them—processes and organizations that are innovative and powerful but sometimes fragile. As this shared body of digital infrastructure has grown, some crucial pieces have become neglected, leading to underproduction: the phenomenon of highly important, low-quality digital artifacts. This dissertation is framed around a series of empirical projects. The first will examine participation dynamics with respect to existing measures of underproduction. I then develop a longitudinal measure of underproduction to understand how underproduction emerges over time. Finally, I apply this measure to digital infrastructure organizations longitudinally, as participation dynamics are unfolding.

Title: About the Blurring of Work and Play: Organizational Dynamics Emerging in a First-Person Shooter Videogame Abstract: ABSTRACTThe aim of this study is to investigate the organizational dynamics occurring in a multiplayer video game. The research involves E-sports professionals, streamers, and amateur players, who have different motives for playing and represent different “work-play” conditions, and explores how they differently enact organizational behaviors in the game. An ethnographic study within an Italian gaming community is currently in progress. I focus on “Call of Duty: Warzone”, a First-Person Shooter Battle Royale game which requires players to enact organizational efforts in order to reach the in-game objectives (e.g., defeat the enemy team). The study uses i) semi-structured interviews with amateur players and participant observation conducted in the game environment played by the amateurs, ii-iii) observation of gaming sessions, analysis of online content and semi-structured interviews with reference to streamers and professionals, iv) analysis of communication exchanges of all three types of players during the gaming sessions. I expect that players belonging to different categories will enact distinct organizational behaviors and give rise to various organizational structures. A cross-comparison between them, which is missing in current literature, would clarify how different modalities of combining work and play impact on organizational behaviors and dynamics.

Title: Exploring Collective Medical Knowledge and Tensions in Online ADHD Communities Abstract: ABSTRACT My proposed dissertation work highlights social media as digitally-mediated support for neurodivergent individuals. By adopting a critical disability theory lens, I critique the techno-solutionism currently present in digital mental health care. I argue that existing social media platforms can provide community support for neurodivergent individuals to step away from the individualistic approaches currently promoted by much digital mental health technology. These social media-based communities are providing an important service of care and collective knowledge for individuals going through similar experiences to find validation and a sense of agency regarding treatment options. My research will further explore the relationships neurodivergent individuals have had with diagnostic and care systems, as well as ongoing tensions with healthcare providers in both physical and digital spaces.

Title: Envisioning Digital Sanctuaries: An Exploration of Virtual Collectives for Nurturing Professional Development of Women in Technical Domains Abstract: ABSTRACT Work and learning are essential facets of our existence, yet socio-cultural barriers may limit access and opportunity in such contexts. Historically, women have faced multiple restrictions that have hindered their entry into professional ventures, many of which have also plagued the technical domains of Information technology and Computer science. Hence, it is essential to investigate how women can subvert such structural subjugation and can find channels through which they can seek support, and guidance and develop collective endurance to navigate their careers. In this regard, virtual communities (e.g., Reddit) have emerged as crucial spaces that allow community building, and opportunities for mobilizing collective action and have also emerged as sanctuaries for support and companionship. Thus, this proposed dissertation aims to focus on how these spaces are used and adopted by women for resilience building and navigating goals, aspirations, and tensions in crafting their professional journey in technical domains. The layers of analysis will include understanding the present conversational patterns in three subreddits dedicated to professional development, and design exploration to better unpack how such channels can be envisioned in the future to more deeply align the affordances of such channels with the values of these virtual spaces. The long-term goal for this work is to address the way in which such channels can be designed and curated to offer spaces for enrichment, empowerment, and advocacy with a focus on professional development for women, especially for those engaged in technical domains.

Title: Addressing Digital Divides in Rural Appalachia with Digital Literacy Education Abstract: ABSTRACT To highlight and contribute to the resolution of digital divides in rural Appalachia, I propose a series of qualitative studies taking place at public libraries in a rural county in West Virginia. These studies aim to identify and characterize the presence of intra-rural digital divides, which are disparities present between areas of similar rurality. I also evaluate digital storytelling as a culturally-resonant method of digital literacy education for Appalachian residents. In preliminary work towards these goals, I have developed a theoretical frame for this work, and conducted one initial study evaluating digital storytelling for digital literacy through a web-based storytelling system. In future work I intend to conduct, I will further evaluate digital storytelling’s effectiveness for this audience through a new Microsoft Word plugin. I will also compare study sites within one rural county to identify and characterize intra-rural digital divides by highlighting qualitative differences between them in terms of digital access, digital literacy, and computer self-efficacy.

Title: Bridging Actionability Gap in Online How-To Instruction-Following Through Human-Chatbot Interaction Abstract: ABSTRACT People frequently come across unfamiliar tasks to complete in daily and professional life. A common strategy to accomplish such tasks is to refer to online How-To tutorials (e.g., wikiHow pages) and follow the instructions step-by-step. However, such archived tutorials are not always actionable, since the sequence of actions is pre-defined and could result in a discrepancy between what was shared and what people need in operational situations. Differing from regular conversational instruction-giving, where people are able to communicate instructions adaptively based on the ongoing working status, archived procedures lack such flexibility. As such, people may face actionability issues in which they may have difficulties navigating the instruction, making decisions on what to do, and solving problems. My dissertation explores system designs using chatbots to accompany novice users to improve the situational actionability and social acceptability of online tutorials for non-expert users. I present ActionaBot, an experimental platform for exploring human-chatbot collaborative instruction giving to augment online tutorials. By participating in the Doctoral Consortium, I am eager to receive feedback from experienced CSCW researchers to inform my research direction. In the long run, I aim to contribute a better understanding of how people and the chatbot may collaborate to enable actionable knowledge sharing and transfer at a large scale.

Title: Information-Seeking, Finding Identity: Exploring the Role of Online Health Information in Illness Experience Abstract: ABSTRACT The identities we hold have a relationship with how we come to express and understand our experiences of illness. Language forms a means for us to express this understanding and experience to others, and receive information to clarify our own experiences. Having access to new information when undergoing an illness experience can be integral in supporting decision-making for one’s health and well-being and change how we understand ourselves and our experience. Individuals are exposed to information about experiences of illness via search engines, social media, and other platforms online. This online health information may thus significantly influence the decision-making process. Research is needed to understand how the affordances of diverse online hubs for health information influence how people understand illness experiences and seek care. How people use the internet for information-seeking is often researched in individual health conditions. This workshop aims to explore the different methods researchers have used to understand online information-seeking journeys and to identify how the internet is, or can be, used to help users make sense of, and give meaning to, their experiences. Through convening a methodologically diverse set of researchers, we hope to generate a foundation and cohesive field of inquiry and community within HCI.

Title: Situating Network Infrastructure with People, Practices, and Beyond: A Community Building Workshop Abstract: ABSTRACTAbstract: Our world is now connected and even entangled in unprecedented ways through networked technologies. Yet pockets of unequal connectivity persist, and technical infrastructures for connectivity remain difficult to design and build even for experts. In this workshop we aim to bring together a global community of multi- and inter-disciplinary researchers and implementers working on infrastructure development and connectivity to explore the existing design challenges and opportunities for bringing technical dimensions of networked infrastructures in conversation with human-computer interaction (HCI) and the social science of infrastructure. We will share, assess and define research problems and resources for rethinking networked infrastructures from human-, community-, and society-centered perspectives, understanding them to be embedded with human values and biases. We particularly intend our collaborative work to support real-world connectivity initiatives, which have grown in critical importance over the pandemic years—especially projects in support of Global South communities. Concrete deliverables from the workshop will include: (1) an initial shared bibliography to help formalize the state of knowledge in our area, (2) an agenda of shared goals, challenges, and intentions in our field, (3) a compilation of resources to support future work, and (4) social and organizing infrastructures for continued communication and academic collaboration.

Title: Ethical Tensions, Norms, and Directions in the Extraction of Online Volunteer Work Abstract: ABSTRACT Online volunteer work such as moderating forums and participating in open source projects not only underpins today’s digital infrastructures, but also helps companies generate immense profits. However, there remains a lack of ethical norms around using volunteer labor for corporate interests, opening opportunities for unchecked extraction of online volunteer work at scale. Early evidence suggests that the extraction of online volunteer work may have negative implications on the tech ecosystem and obfuscate the potential for exploitative labor practices. In this workshop, we invite participants to discuss 1) what ethical tensions exist in the current approaches to extracting online volunteer work, 2) what ethical norms should be followed or recommended and 3) what are the opportunities for social computing technologies to promote these norms. Furthermore, we open a dialogue around whether online platforms should be providing non-monetary compensation, such as education and resources, that is often promised in in-person volunteer settings. We plan to involve a diversity of roles beyond academic researchers, such as online volunteers and practitioners to discuss these questions.

Title: Growing New Scholarly Communication Infrastructures for Sharing, Reusing, and Synthesizing Knowledge Abstract: ABSTRACTSharing, reuse, and synthesis of knowledge is central to the research process. These core functions are in theory served by the system of monographs, abstracts, and papers in journals and proceedings, with citation indices and search databases that comprise the core of our formal scholarly communication infrastructure; yet, converging lines of empirical and anecdotal evidence suggest that this system does not adequately act as infrastructure for synthesis. Emerging developments in new institutions for science, along with new technical infrastructures and tooling for decentralized knowledge work, offer new opportunities to prototype new technical infrastructures on top of a different installed base than the publish or perish, neoliberal academy. This workshop aims to integrate these developments and communities with CSCW’s deep roots in knowledge infrastructures and collaborative and distributed sensemaking, with new developments in science institutions and tooling, to stimulate and accelerate progress towards prototyping new scholarly communication infrastructures that are actually optimized for sharing, reusing, and synthesizing knowledge.

Title: Who Has an Interest in “Public Interest Technology”?: Critical Questions for Working with Local Governments & Impacted Communities Abstract: ABSTRACT Local governments use a wide array of software, algorithms, and data systems across domains such as policing, probation, child protective services, courts, education, public employment services, homelessness services, etc. A growing body of work in CSCW and HCI has emerged to study, design, or demonstrate the boundaries of these technologies, oftentimes working with local governments. Local governments ostensibly aim to serve the public. So, some prior work has collaborated with local governments in the name of the public interest. However, others argue that local governments primarily police poor, minoritized communities, especially with increasingly limited funding for public services such as education or housing. These tensions raise critical questions: (How) should researchers collaborate with local governments? When should we oppose governments? How do we ethically engage with communities without being extractive? In this one-day workshop, we will bring together researchers from academia, the public sector, and community organizations to first take stock of work around public interest technologies. We will reflect on critical questions to orient the future of public interest technology and how we can work with, around, or against local governments while centering impacted communities.

Title: Solidarity and Disruption Collective Organizing In Computing II Abstract: ABSTRACT This workshop responds to the incredible growth of corporate tech power - including during a deadly pandemic - and the growing need for tech workers of all stripes (including researchers/academics) to build grassroots power. This workshop’s twinned themes of solidarity and disruption acknowledge that solidarity is vital but not sufficient to enact the structural changes we need. Disruption— in the form of sit-ins, strikes, refusal, and direct action—has become a necessary condition in the face of technology corporations’ greed-driven expansion towards militaristic, techno-totalitarian futures. In this one-day workshop, we will bring together tech workers, researchers and activists from academia, industry, and community-based organizations to extend conversations that we began in two workshops last year. We will further explore avenues and approaches for action, particularly to support practitioners’ and activists’ objectives, and connect participants with concrete opportunities for on-the-ground action.

Title: Designing Hardware for Cryptography and Cryptography for Hardware Abstract: ABSTRACTThere have been few high-impact deployments of hardware implementations of cryptographic primitives. We present the benefits and challenges of hardware acceleration of sophisticated cryptographic primitives and protocols, and briefly describe our recent work. We argue the significant potential for synergistic codesign of cryptography and hardware, where customized hardware accelerates cryptographic protocols that are designed with hardware acceleration in mind.

Title: We Are the Experts, and We Are the Problem: The Security Advice Fiasco Abstract: ABSTRACTIn an ideal world, automated tools and systems could manage security and privacy seamlessly and transparently with minimal human input. In the real world, we are nowhere close to that ideal. Instead, in order to achieve good security and privacy outcomes, people need to absorb and apply high-quality security and privacy information and advice. This applies not only to end users, but also to software developers, product managers, and even security operations professionals. Sadly, the current state of the security advice and information ecosystem is in many respects a disaster. End users often get their advice from TV shows, movies, and even misleading influencer ads [2, 4], while soft ware developers take unvetted suggestions from Stack Overflow [1, 3]. Even compliance standards -- which are designed to provide authoritative security guidance -- have numerous problems [6, 7]. Our review of security advice on the web found 374 unique advice imperatives, many of which directly contradict one another [5]. This sad state of affairs is, in many ways, our fault. Security experts, like the ones who attend conferences such as CCS, often refuse to prioritize, recommending maximum security without tailoring to specific situations. Researchers evaluate tools and techniques in idealized rather than realistic use contexts, and have made little progress in accurately measuring the costs and benefits of any particular intervention. In this talk, I will review the many problems of the security and privacy information and advice ecosystem, and how we got here. I'll outline our responsibility, as experts and researchers, to help improve the quality, availability, and usability of security and privacy information. Finally, I'll discuss at what we know (and what we need to find out) about how to make progress.

Title: Updatable Public Key Encryption from DCR: Efficient Constructions With Stronger Security Abstract: ABSTRACTForward-secure encryption (FS-PKE) is a key-evolving public-key paradigm that preserves the confidentiality of past encryptions in case of key exposure. Updatable public-key encryption (UPKE) is a natural relaxation of FS-PKE, introduced by Jost et al. (Eurocrypt'19), which is motivated by applications to secure messaging. In UPKE, key updates can be triggered by any sender -- via special update ciphertexts -- willing to enforce the forward secrecy of its encrypted messages. So far, the only truly efficient UPKE candidates (which rely on the random oracle idealization) only provide rather weak security guarantees against passive adversaries as they are malleable. Also, they offer no protection against malicious senders willing to hinder the decryption capability of honest users. A recent work of Dodis et al. (TCC'21) described UPKE systems in the standard model that also hedge against maliciously generated update messages in the chosen-ciphertext setting (where adversaries are equipped with a decryption oracle). While important feasibility results, their constructions lag behind random-oracle candidates in terms of efficiency. In this paper, we first provide a drastically more efficient UPKE realization in the standard model using Paillier's Composite Residuosity (DCR) assumption. In the random oracle model, we then extend our initial scheme so as to achieve chosen-ciphertext security, even in a model that accounts for maliciously generated update ciphertexts. Under the DCR and Strong RSA assumptions, we thus obtain the first practical UPKE systems that satisfy the strongest security notions put forth by Dodis et al.

Title: Helping or Hindering?: How Browser Extensions Undermine Security Abstract: ABSTRACTBrowser extensions enhance the functionality of native Web applications on the client side. They provide a rich end-user experience by utilizing feature-rich JavaScript APIs, otherwise inaccessible for native applications. However, prior studies suggest that extensions may degrade the client-side security to execute their operations, such as by altering the DOM, executing untrusted scripts in the applications' context, and performing other security-critical operations for the user. In this study, we instead focus on extensions that tamper with the security headers between the client-server exchange, thereby undermining the security guarantees that these headers provide to the application. To this end, we present our automated analysis framework to detect such extensions by leveraging static and dynamic analysis techniques. We statically identify extensions with the permission to modify headers and then instrument the dangerous APIs to investigate their runtime behavior with respect to modifying headers in-flight. We then use our framework to analyze the three snapshots of the Chrome extension store from Jun 2020, Feb 2021, and Jan 2022. In doing so, we detect 1,129 distinct extensions that interfere with security-related request/response headers and discuss the associated security implications. The impact of our findings is aggravated by the extensions, with millions of installations dropping critical security headers like Content-Security-Policy or X-Frame-Options.

Title: Practical, Round-Optimal Lattice-Based Blind Signatures Abstract: ABSTRACTBlind signatures are a fundamental cryptographic primitive with numerous practical applications. While there exist many practical blind signatures from number-theoretic assumptions, the situation is far less satisfactory from post-quantum assumptions. In this work, we provide the first overall practical, lattice-based blind signature, supporting an unbounded number of signature queries and additionally enjoying optimal round complexity. We provide a detailed estimate of parameters achieved -- we obtain a signature of size slightly above 45KB, for a core-SVP hardness of 109 bits. The run-times of the signer, user and verifier are also very small. Our scheme relies on the Gentry, Peikert and Vaikuntanathan signature [STOC'08] and non-interactive zero-knowledge proofs for linear relations with small unknowns, which are significantly more efficient than their general purpose counterparts. Its security stems from a new and arguably natural assumption which we introduce, called the one-more-ISIS assumption. This assumption can be seen as a lattice analogue of the one-more-RSA assumption by Bellare et al [JoC'03]. To gain confidence in our assumption, we provide a detailed analysis of diverse attack strategies.