Visualizing Toxicity in Twitter Conversations | Hacker News Reader

Visualizing Toxicity in Twitter Conversations

medium.com

41 pointsbetolink24 comments

The visualization doesn't interest me nearly as much as how Twitter classifies something as "toxic". It seems subjective enough that one person's "toxic conversation" is another person's "sick burn". Unfortunately, the post doesn't contain any explanation or links regarding their classification algorithm(s).

I was curious about their definition as well. Our world is so hyper political and polarized that the definition of “toxic” and “hate speech” varies widely depending on your political ideology. There’s zero space left in the middle.

holding a middle ground is frowned upon on the internet.

I've confirmed this myself by writing torridly on occasion and for sites which allow voting (such as HN/Reddit) I am rewarded.

When I make a nuanced, balanced and reasoned reply I am punished.

I bring this up because on twitter there is no 'disagree' button, there is only shouting at your enemy in 280 characters, that used to be 140 characters, enough that you can never actually make a reasoned assertion about anything.

I still have my twitter account, it's a shame to remove it because I have a good handle- but I can't escape the blatant fact that it is only inhabited by bullies. Both sides of a weirdly distorted political spectrum drumming/repeating their own rhetoric to each other -- and anyone who might try to strike a balance is seen as an enemy by both sides of the extreme.

The extremes of twitter almost caused me to hate feminism, A cause I actually have always agreed with. In turn it also caused me to hate ideologues who espouse fears of restricted freedoms/end of privacy. Which I also agree with.

Twitter is _not_ a place to go to have a conversation, it's a place where a conversation is thrust upon you if you speak, whether you want it or not.

I'm writing this because I agree with you. I've had a growing resentment for twitter due to the blatant, crushing pile-on that it allows.

Eh, I think you’re experiencing something else. Moderation and hedging phrases seem more like blue-chip stocks in the internet points risk/reward game. You’re almost guaranteed not to lose that many points with a stance fewer folks will brigade, and if you hit the right tone, odds are you’ll get a modest reward for pitching in. Picking a side and going all-out on some cultish activism is more like a venture capital penny-stock situation. With a perfect storm of the right audience, day, etc, you could get a huge internet karma ROI! And the other times, you lose everything.

YMMV of course, a lot depends on how “rational” the “market” happens to be.

I usually see the opposite effect, at least on HN: the more extreme a view, the more likely to be downvoted/ignored.

Directly center/nuetral is usually ignored, and the main upvoting comes through slightly, but not strongly, shifted from the center. Even on being on the wrong side of things, dissent is usually limited but mostly nuetral

But then again, my stronger opinions tend to be on the wrong side of things, though rarely commenting on political hot-topics(these are more like claiming the best way to shop for things is to first filter on whether their marketing is up to snuff: if they tried to advertise and managed a shitshow, probably their whole operation is just as bad.. got a lot of shit for defending that one)

There's even a subreddit dedicated to criticizing people that don't hold extreme political views, /r/enlightenedcentrism. It's pretty much a super left-wing echo chamber though, they view their "side" as 100% correct, and anybody who falls in the middle of the two extremes is considered too lazy and apathetic to see that being super liberal is the only correct political ideology.

yeah, judging from a few of those samples, i'm assuming toxicity means being on the "incorrect" side of the gun and immigration debates.

The author of this article suspiciously leaves out any formal definition of toxicity, but there is related work that does go in-depth into how to define it from an objective standpoint: https://perspectiveapi.com/

Agreed, the visualisation itself is pretty pointless also as it doesn't show the data in a meaningful way.

Definition of "toxicity" is nowhere to be found in this post. Plenty of cool pictures, though.

Someone needs to write a joke medium article from the perspective of Winston Smith. "Unpersoning thought criminals with AWS"

So toxic, but what defines toxic, how were they categorised etc? Just can't take the term seriously anymore after all the buzzfeed videos educating me on American politics that's been shoved into my feeds.

Yup, they've muddied the waters, possibly on purpose.

For a more formal definition of "toxic", see these research efforts by Google and Kaggle, respectively:

https://perspectiveapi.com/ https://www.kaggle.com/c/jigsaw-toxic-comment-classification...

It seems unfortunate that the goal was to build a good internal marketing video, rather than a useful user interface that could be used for navigating Twitter?

The final viz looks awful. In the attempt to make it look like a tree, you did get a tree but lost the informational value. I thought the earlier iterations had less noise and more signal.

There are plenty of books on visualization theory. Exposes pitfalls.

This is an awesome related API by Google: https://perspectiveapi.com/

Also check out this Kaggle competition: https://www.kaggle.com/c/jigsaw-toxic-comment-classification...

Red and green are not a great choice for color blind people.

It’s amazing to me that people look at an online community such as twitter and then determine if it’s toxic or not based on observed behaviors. They’re just people. People are toxic on twitter. Twitter is not toxic. Twitter is exactly the way that people who use it want it to be.

How do you expect to change twitter without changing people? It’s an exercise in futility. You can regulate it for eternity and it will still reflect the people who use and engage in the site.

The same people can behave totally differently depending on circumstances. For example: many people experience road rage, but they'd never behave that way outside of their car. Something about being in a car separated by a couple panes of glass causes a lot of people to act insane. It's similar for social media:

- Twitter and Tumblr both have retweet/reblog with commentary, which makes it extremely convenient for people to tell their followers about something they dislike. Just click retweet, type, "Look at this dumbass.", and watch the likes come in.

- Twitter notifies you if someone replied, quote-tweeted, mentioned you, or linked to a tweet of yours. This means that people definitely know when somebody disagrees with or dislikes them. If those same people were mocking them on some other forum, they'd never know and a cycle of hostility wouldn't form.

- Twitter's character limit makes it hard for people to keep nuance in their ideas. Replies to tweets are sorted by time, so whoever replies first (i.e. gives their response the least thought) gets the most eyeballs.

- Lastly, Twitter defaults to only accept direct messages from people you follow, meaning that most people can't criticize in private. They have to broadcast the criticism to the whole world, making it much harder for both parties to keep their egos in check.

These "features" combine to make civil conversation all but impossible.

All good points. One I would add: there is far less assumption of good faith from strangers on Twitter, and most parts of the internet in general.

For whatever reason, in-person interactions have a level of charity in them. You'll clarify misconceptions, give people the chance to make and walk back mistakes, and generally assume that your conversational partner is reasonable until proven otherwise.

Twitter is not like this. Many people seem to find it very easy to start a conversation with criticism, looking for evidence of stupidity or malice rather than trying to clarify it away. And there is almost no ground-level assumption that the other person is actually "trying" the way you are.

Seems easily traced back to Twitter not being a very human medium. As in the car example, it's easier to strongly dislike something when it's not obviously another human. But of course, the existence of trolls on the internet and mostly nonexistence of tolls in everyday life is a factor too.

For this reason, I've mostly given up on having any kind of remotely grey-area discussion on Twitter. I just use it to follow a few dozen people whose thoughts I'm interested in. And outside of HN and email/messaging, I don't talk through the internet at all.

I mean, moderation policies, tools to change signal to noise ratio, reward incentives for certain types of behaviours, and community mores all affect the observed level of toxicity. There are clearly places and periods on the internet that are more or less abusive and it's not because massive social change is happening in the outside world, right?

Awesome project. Always interesting to see original ideas in data visualization.

Reply on news.ycombinator.com