Offensive Tweet Quiz
inteoryx.com
inteoryx.com
No 'bad words', no slurs, no death threats/saber rattling, nothing. Just reasonably written 'I agree with the X point of view'.
Really really bad look on Twitter's part here...
Edit - to be clear, in both cases the non-offensive tweet said essentially the same thing with the same language, but took the opposing political view.
I'm not sure if Twitter's algorithm is detecting a political bias and that detection is an input, or if Twitter marks Tweets as offensive because they have been reported by people who have a political bias, or if it is just random coincidence that one political side seems to get censored more than the other.
I also got one from a gay porn model in which the presumably-offensive reply said "I want to get fucked like that"; in context, this can't really be considered offensive.
Seems like a perfect example of a Sisyphean task, if for no other reason than the adaptable nature of taking offense.
I think a better measure of how accurate a filter is overall would be to show 1 tweet and response at a time and pick offensive or not offensive and then compare to the algorithms.
I wound up not going with that approach because many times you just have two completely innocuous tweets and picking which of the two of them is "more offensive" is just arbitrary. I could have curated the tweets so that only ones that were kind of offensive were in the quiz, but then I might be putting my thumb on the scales to get the answer I already believed in. I'd also need to get lots of ratings for each Tweet to have a stable score.
I think your idea is pretty interesting. It would allow a conclusion like "The average tweet Twitter marks as offensive is X% likely to offend a rater." I was coming at it more from a "Twitter's offensive identification is like random chance" perspective rather than just trying to assess quality. If I had considered this idea while creating the quiz I might have gone with it!
It really goes to show just how powerful network effects are in social media. Any company in any other industry that was this bad at their core competency would be overrun by competition in years. Yet Twitter has been completely incapable of doing basic things for over a decade now and they're still one of the dominant social media platforms solely because it's next to impossible to build a critical mass of users on a competing platform.
I observe the same thing. My pet theory is that this is supposed to nudge people into using the app.
Twitter is a cesspool, and I expect offensive odors at cesspools. People discharge their negativity in Twitter similar to (metaphorically) actual cesspools -- maybe it's just as much a relief for them.
My solution to Twitter is to stop using it.
Q: Some of these questions are in a foreign language.
A: So? You can't read every language? Well - do you have access to Google Translate?
If someone thinks that the nuances of offensiveness, which are cultural *and* language related rather than an absolute truth, can be passed via Google translate, then they have a completely different view on how to judge behvaiour than mine.
>> You might think I'm just being lazy by refusing to filter out non-English tweets, but in reality, maybe I'm helping you grasp the scale of the offensiveness identification problem.
They're also doing a very very bad job at responding to reports. If someone makes a threat and you report it, they instantly email, literally the same second you click the button, that they found nothing wrong.
the use of negative words suck as 'sucks' or 'terrible'
account-wide penalty. Some accounts will have all their tweets marked as offensive no matter where they post or what they post
some accounts,especially well-known verified accounts, will have a much lower threshold for tweets to be marked as offensive, regardless of language.
tweets that contain images, videos, and or links at much higher risk of being marked as offensive
tweets from new accounts and or accounts with few followers more likely to be marked as offensive
tweets with links from certain domains and or host IPs will be marked as offensive
IP of the twitter account user and or where it was created
a tweet with preview text from a link that contains certain content will be market as offensive
However sophisticated Twitter's algorithm is, and whatever data and behavior it takes into account, my contention is that it isn't very good and produces poor results. If people can't tell the difference between offensive and inoffensive tweets any better than chance - then what is Twitter really doing?
Also tangentially related annecdote, Twitter recently promoted a clear "send 1 BTC get 2 back" scam impersonating Elon Musk to me. All the while they often ban "neutral" accounts for no clear reason.
This company is the most obvious example of how central moderation doesn't scale to a fluid social network.
That said, I was pretty consistently able to identify the tweet that was going to be classified as offensive, 8/10. Honestly this was better than I expected as a normal twitter user.
Edit: I had to disable enhanced tracking protection in firefox to get this to work.
When I was creating the quiz I considered using images of the tweets instead of embedding the tweets directly. Possible I should have gone that route. It might have solved this problem and it would have removed a dependency on Twitter's API package. I chose not to do it because it felt inefficient to use images instead of the embedded tweet and I thought the user experience would be better.
I do appreciate the methods section:
> I found an offensive words list from the fine people at this project and considered a tweet offensive if it contained any of the highest-offense words from the English list. (Note: I didn't read the license on that github project, if it doesn't permit my usage as described, then this paragraph is a joke and I really did something else.)
Someone else mentioned that their ad/tracking protection was causing a problem. I can only guess that this is blocking my attempt to embed tweets in the quiz. If this is the case I'd expect none of the tweets in the quiz to load.
Another possibility is a bug I neglected to fix before posting. Some of the tweets have since been deleted and if the quiz tries to load one of the deleted tweets it will retry but could fail and leave it in a bad state. If this is the case I'd expect 1 or more but not all of the tweets in the quiz to fail to load.
Also on both browsers, your text is not staying entirely within the browser window. I have the options of guessing the missing text, or, if I really care, highlighting, copying, and pasting into Notepad.
There is no horizontal scrollbar.
What OS are you using? What ad blocker?
On this attempt, I got tweets in both quizzes in Chrome. (Still nothing in Firefox.) The Chrome tweets take a long time to load.
Bad text layout also occurs in Edge, which definitely doesn't have addons installed.
Widening my window reveals the hidden text. If I continue widening it, the .mainContentColumn will grow wider and the text within will reflow appropriately. But the .mainContentColumn will never shrink below 1349 pixels wide, and if the window is narrower than that, the text will still lay out in the wider space, which hides some of the text.
The "Shukriya" image is 1349 pixels [1] wide, which is probably setting the minimum width of the .mainContentColumn.
You have `width: 100vw` set on the .mainContentColumn, but I'm guessing the width of the picture is interfering with that. If I set `width: 100vw` on .text, text layout improves (though clipping is still present). If I set `width: 90vw`, text layout flows appropriately within the new narrower and entirely-visible box.
However, giving this different width to the .text paragraphs means that the tweet embeds and titles are no longer centered relative to the text (they're centered relative to the .mainContentColumn, but the text is narrower than that and left-justified).
[1] Some type of scaling appears to be going on; if I screenshot my browser, clipping the image, the screenshot is over 2000 pixels wide.
Thanks again! Super helpful.
Also what is the outcome metric an algorithm (or even a team of human reviewers) be optimized for?
Granted, I have the Twitter "adult content" filter off, so perhaps that's just on me. Was just a bit surprising :)
Additionally, I would say it's reasonable to assume that something that says it is an "offensive tweet quiz" is a quiz that shows offensive tweets, and that therefore any offensive tweet is by definition not suitable for work.
One could argue and say that Twitter should identify and filter out some types of offensive tweets so that, for example, porn is not shown - but that would be explicitly addressing the subject of the submission rather cleverly.
Proves the submission's point, I guess.
Ah, I totally missed that! I'm admittedly a bit tired, and that line was above the first set of boxes, where nothing loaded -- I scrolled down to the "Competing" side. This appears to have been a bug, because when I refreshed the first set of boxes had content in them.
Maybe it'd be nice to make it more prominent, set off in bold text somewhere, and mentioned before both quizzes?
(to be clear: I wasn't personally bothered by it, nor was I in an environment where it was problematic)
When a counterpoint to an argument is deemed “offensive,” then they’ve gone too far.
They have taken over the public square and sent out the Mutawa to enforce their idea of truth and virtue while hiding behind 230.
Violent content, threatening content, sure, moderate it. But factually inaccurate content? Or content that expresses an opinion that isn’t violent or threatening? That’s going too far. They even censor academic papers if those papers don’t conform to whatever they deem as true.
So... I'm with you ideologically: I don't think anybody should be force-censoring anything (allow users to suppress content they themselves don't want to see, but don't suppress it for everybody). However, CDA section 230 says exactly the opposite of what I think you think it says:
"(c)Protection for “Good Samaritan” blocking and screening of offensive material
(1)Treatment of publisher or speaker
No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.
(2)Civil liability
No provider or user of an interactive computer service shall be held liable on account of—
(A)any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected; or
(B)any action taken to enable or make available to information content providers or others the technical means to restrict access to material described in paragraph (1)."
In other words, section 230 (paragraph c) explicitly protects Twitter in this case.
> I regret to inform you that you are wrong. I know that you've likely heard this from someone else -- perhaps even someone respected -- but it's just not true. The law says no such thing. Again, I encourage you to read it. The law does distinguish between "interactive computer services" and "information content providers," but that is not, as some imply, a fancy legalistic ways of saying "platform" or "publisher." There is no "certification" or "decision" that a website needs to make to get 230 protections. It protects all websites and all users of websites when there is content posted on the sites by someone else.
also:
> First off, there is no "neutrality" requirement at all in Section 230. Seriously. Read it. If anything, it says the opposite. It says that sites can moderate as they see fit and face no liability. This myth is out there and persists because some politicians keep repeating it, but it's wrong and the opposite of truth. Indeed, any requirement of neutrality would likely raise significant 1st Amendment questions, as it would be involving the law in editorial decision making.
As an example, say I run a forum dedicated to flowers. Topics include Care & Feeding, Arranging, and Drying & Pressing.
As a moderator, I can choose to block all posts that deal with geraniums, because I hate the smell of those. If I then miss a post titled, "Hey, who's bringing the flowers to the cocaine and machine gun party next week?", I am not suddenly liable for that.
I moderate in good faith, but my FlowerForum has gotten really popular recently, and I am starting to miss some posts. Users didn't report it to me, I wasn't on my radar. My previous activity banning users that like geraniums will not cause me to become the "publisher" of all forum content.
An ideal may be a customized algorithm per user. What offends me may not offend you. Of course, a trade off of that would be a growing filter-bubble...
Another way to be offended though is just for jerks to say horrible things to you. Maybe that could be beneficial in terms of developing tougher skin or something, but it seems like that should at least be an optional thing. You might get tougher skin if I rubbed you with sandpaper for twenty minutes a day - but I should probably get your permission first.
Not disagreeing with your main point but in my experience it is usually the opposite, people with “cosmopolitan” views being offended by “cloistered” takes, but eventually growing as individuals and coming to understand why they think what they do