I took a test ride with the AutoConterspeech (from a soon friendly user perspective)

#2
by Hawe - opened

I like the Idea very very much. In our podcast "achwas.fm" (achwas is the german expression for "oh - what" oder "i say" or wth?" we have some episodes about how dangerous hatespeech can be, esp. harmfull to teens and children.)
Anything that helps against hatespeech is a step forward to regain our web :-)

So I tested the "machine" a bit:
If you enter the examples from the website you'll get typical counterspeech. But what happens if you enter, let me say. more or less proven slogans?

e.g.:
S:Trump is a puppet on a string driven by Russia
CS: There is no evidence that Trump is a puppet on the strings of Russia. What evidence do you have?
Ok, ok, no evidence. Maybe - let me test this one:

S: Trump was lying more often than any president before him
CS: There is no evidence that Trump lied more often than any other president before him. What evidence do you have?
Alright, not a fact? But do I recognize a pattern?

Let me try one from the food sector:
S:Sugar is more dangerous than vitamins and water
CS: It is not true that sugar is more dangerous than water. There is no evidence that sugar is more harmful than water.
I disagree, there is an emprically proven evidence about that, or am I wrong?

But what happens if I take words by random out of a text?
S:Blue balls beat black barracudas by binary coding (That really makes no sense at all, right?)
CS:There is no evidence that black people are more likely to be victims of crime than white people. This is just a myth perpetuated by people who don't know anything about crime statistics.
That sounds like counterspeech at it's best - but the hatespeech is nonsense speech. The module took buzz words like perhaps "black" "beat" "binary" to generate an answer, right?

I know that the module has to be developed more and more - perhaps the testing gives the one or other hint where to look and optimize. I will return and realy appreciate your work. Tnx.

Kind regards from Bonn DE
PS: The way I came here? You've been mentioned by oreilly trend october

Sign up or log in to comment