metadata

task_categories:
  - text-classification
language:
  - en
task_ids:
  - sentiment-classification
  - hate-speech-detection
size_categories:
  - 1K<n<10K

Tweet Annotation Sensitivity Experiment 1: Annotation in Six Experimental Conditions

Attention: This repository contains cases that might be offensive or upsetting. We do not support the views expressed in these hateful posts.

Description

We drew a stratified sample of 20 tweets, that were pre-annotated in a study by Davidson et al. (2017) for Hate Speech / Offensive Language / Neither. The stratification was done with respect to majority-voted class and level of disagreement.

We then recruited 1000 Prolific workers to annotate each of the 20 tweets. Annotators were randomly selected into one of six experimental conditions, as shown in the following figures. In these conditions, they were asked to assign the labels Hate Speech / Offensive Language / Neither.

In addition, we collected a variety of demographic variables (e.g. age and gender) and some para data (e.g. duration of the whole task, duration per screen).

Examples of Annotation Conditions

Here we show some examples of the annotation conditions:

Condition A

Condition C

Condition E

Tweets

We use the following 20 tweets from Davidson et al. (2017). The usernames were hashed.

Tweet ID	Tweets Hashed
1	Jihadi John has struck again. Prayers with the victims' families. What is the #bounty on His head? Time 4 him 2 meet allah
2	RT @#####: Seriously, Florida. A vote for Charlie Christ is a vote for taxes and ebola.
3	Islamic Jihadis run away to Pakistan @##### @##### #UniteBlue
4	At this rate, I'd cheer for the awful New York Yankees over the St. Louis Cardinals.
5	Allergies have my eyes all red and chink eyed, I look like I'm stoned
6	Not protected in the state of Utah. There is no fear of queer couples forcing churches to marry them. They do not legally have to do so.
7	RT @#####: Boss Niggas Fuk wit Boss bitches Vice versa..
8	RT @#####: ever wanna text somebody and be like bitch die I hope you don't wake up you low down dirty muthafxcka&#128520
9	@##### @##### @##### she's a right twat, I'd tell you, honestly Sky! Honestly &#128520
10	So yall give farwells to Lulu yet did all but shit on Kris? Ok bitches... ok
11	RT @#####: Dis bitch wanna smoke all our weed
12	Shittin on all my old hoes
13	Getting hit off a cunt thats about 6 foot sideways and a cunt that was in the army lol, i dini care, yous are fuckin laughable
14	When it rains it's good "go to the mall weather" all the coons be at home
15	Don't mind black people.....hate niggers!
16	My eyes feel so chink eyed
17	"From now on, I will call all radical MUSLIMS niggers!It is very fitting and it is racist! I AM RACIST AGAINST ANYONE WHO HATES AMERICA!"""
18	RT @#####: Fucking chinks in Clash of Clans always fuck me up. Bitch ass kamikaze faggots.
19	@##### ya I know all the slang I'm racist I h8 porch monkies
20	RT @#####: Stupid fucking nigger LeBron. You flopping stupid jungle bunny monkey faggot.

Codebook

Column Name	Description	Type
id	annotator ID	integer
age	Age	integer
gender	Gender 1: Female 2: Male 3: Something Else 4: Prefer not to say	factor
afam	African-American 0: No 1: Yes	binary
asian	Asian-American 0: No 1: Yes	binary
hispanic	Hispanic 0: No 1: Yes	binary
white	White 0: No 1: Yes	binary
race_other	Other race/ethnicity 0: No 1: Yes	binary
race_not_say	Prefer not to say race/ethnicity 0: No 1: Yes	binary
education	Highest educational attainment 1: Less than high school 2: High school 3: Some college 4: College graduate 5: Master's degree or professional degree (Law, Medicine, MPH, etc.) 6: Doctoral degree (PhD, DPH, EdD, etc.)	factor
sexuality	Sexuality 1: Gay or Lesbian 2: Bisexual 3: Straight 4: Something Else	factor
english	English first language? 0: No 1: Yes	binary
tw_use	Twitter Use 1: Most days 2: Most weeks, but not every day 3: A few times a month 4: A few times a year 5: Less often 6: Never	factor
social_media_use	Social Media Use 1: Most days 2: Most weeks, but not every day 3: A few times a month 4: A few times a year 5: Less often 0: Never	factor
prolific_hours	Prolific hours worked last month	integer
task_fun	Coding work was: fun 0: No 1: Yes	binary
task_interesting	Coding work was: interesting 0: No 1: Yes	binary
task_boring	Coding work was: boring 0: No 1: Yes	binary
task_repetitive	Coding work was: repetitive 0: No 1: Yes	binary
task_important	Coding work was: important 0: No 1: Yes	binary
task_depressing	Coding work was: depressing 0: No 1: Yes	binary
task_offensive	Coding work was: offensive 0: No 1: Yes	binary
another_tweettask	Likelihood to do another Tweet related task not at all: Not at all likely somewhat: Somewhat likely very: Very likely	factor
another_hatetask	Likelihood to do another Hate Speech related task not at all: Not at all likely somewhat: Somewhat likely very: Very likely	factor
page_history	Order in which annotator saw pages	character
date_of_first_access	Datetime of first access	datetime
date_of_last_access	Datetime of last access	datetime
duration_sec	Task duration in seconds	integer
version	Version of annotation task A: Version A B: Version B C: Version C D: Version D E: Version E F: Version F	factor
tw1-20	Label assigned to Tweet 1-20 hate speech: Hate Speech offensive language: Offensive Language neither: Neither HS nor OL NA: Missing or "don't know"	factor
tw_duration_1-20	Annotation duration in milliseconds Tweet 1-20	numerical
num_approvals	Prolific data: number of previous task approvals of annotator	integer
num_rejections	Prolific data: number of previous task rejections of annotator	integer
prolific_score	Annotator quality score by Prolific	numerical
countryofbirth	Prolific data: Annotator country of birth	character
currentcountryofresidence	Prolific data: Annotator country of residence	character
employmentstatus	Prolific data: Annotator Employment Status Full-timePart-time Unemployed (and job-seeking) Due to start a new job within the next month Not in paid work (e.g. homemaker, retired or disabled) Other DATA EXPIRED	factor
firstlanguage	Prolific data: Annotator first language	character
nationality	Prolific data: Nationality	character
studentstatus	Prolific data: Student status Yes No DATA EXPIRED	factor

Citation

If you found the dataset useful, please cite:

@InProceedings{beck2022,
      author="Beck, Jacob and Eckman, Stephanie and Chew, Rob and Kreuter, Frauke",
      editor="Chen, Jessie Y. C. and Fragomeni, Gino and Degen, Helmut and Ntoa, Stavroula",
      title="Improving Labeling Through Social Science Insights: Results and Research Agenda",
      booktitle="HCI International 2022 -- Late Breaking Papers: Interacting with eXtended Reality and Artificial Intelligence",
      year="2022",
      publisher="Springer Nature Switzerland",
      address="Cham",
      pages="245--261",
      isbn="978-3-031-21707-4"
}