I should state up front that I do not use Twitter. I have occasionally followed a link to Twitter, but I don’t linger there. It is a confusing mess that seems to bring out the very worst in people. It seems that Twitter is starting to realize this, and to understand that the solution may not be in controlling who has access to Twitter, but in how the system rewards its users. We all respond to incentives. We all, to some degree, are rewards junkies – when certain behaviors are rewarded, we repeat and amplify those behaviors to receive more of those rewards. Twitter’s problem, as its CEO Jack Dorsey has begun to understand, is that it rewards awful behavior, rage, groupthink, bullying, and dehumanizing its users.  A Buzzfeed article from May 15th details how Twitter is experimenting with a new interface – one that reduces the incentives for the worst of behavior, and perhaps restores some humanity.

In its early years Twitter optimized for engagement, which engagement features (replies, and the like and retweet buttons) and metrics (number of followers, likes, retweets, and replies) help to deliver. So now it’s trying to shift what it encourages people to do.

Advertisement revenue largely funds Twitter, and advertisers pay based on how often their ads are seen by Twitter users. For Twitter to maximize revenue, therefore, it needs to keep people active and engaged on the platform, and to keep them engaged it needs a constant churn of tweets. These tweets can either form conversations, or they can be endless re-tweets of what others have said, either to support or tear them down. Their character limit precludes long-form replies, so what gets attention tends to be either the outrageous or the pithy. The incentives at play are the accumulation of followers, as well as replies, retweets, and likes — the more people you get to follow you, the more your tweets are recirculated, the more people respond to them either in outrage or support (or outrage at the support, etc.), and the more they are liked, the more you as a user are rewarded, and the more attention you get. With hundreds of millions of users worldwide, a large count of followers helps you to rise above the din – especially as those followers will retweet your own work to people who would otherwise not know who you are. This system is especially good at rewarding outrage and anger – there are ample incentives to be horrible, and to band together with other horrible people either for prestige or protection. And Twitter is losing members.

One user, @matthewreid, replying to Stone, summed up the issues facing Twitter nicely: “A quick scroll through many of these replies illustrates what made this place I love so toxic. Bullying. Mob mentality. Insufferable knowitalls.” Twitter CEO Dorsey has admitted the same himself: “I also don’t feel good about how Twitter tends to incentivize outrage, fast takes, short term thinking, echo chambers, and fragmented conversation and consideration.”

In 1998, Adam Hochschild published King Leopold’s Ghost, about how Leopold II of Belgium, through fraud and barbarism, carved out the Congo River basin as his personal slave plantation and turned it into one of the worst examples of colonialism ever to darken human history. The podcast series Our Fake History, hosted by Sebastian Major, recently ran a three-part series exploring both this book and other source material on the Congo. One of the especially notable horrors of that time was how human hands became a sort of nightmarish currency of trade. You see, colonial troops would be issued ammunition for their rifles, but were not allowed to expend that ammunition except for “legitimate” cases of need. If you used all of your allotment of ammo, you would not be issued more without proof. What was that required proof? The severed hands of the people you shot. Of course, in a small skirmish, one might fire many rounds that would miss, or fail to kill. Well, better make up those numbers…

In just a couple of years, this led to soldiers massacring or maiming entire villages, tribes raiding and slaughtering other tribes to have hands to barter for goods or weapons, and all manner of other brutalities one can envision. Terrible incentives lead to terrible behavior.

Twitter’s incentives (particularly likes and retweets) reward vanity, rage, and humiliation — we’re not seeing the bartering of severed hands, but we are seeing the rewarding of other forms of dehumanizing behavior. Whatever behavior you reward, you will see more of that behavior. If you reward barbarism, you will get more barbarism. Other social media platforms are suffering similar problems:

Twitter is not alone in this. Tech leaders across the industry are rethinking the role of their platforms’ incentives, in response to mounting criticism that technology platforms do more harm than good. Instagram is running a test where like counts are hidden to followers, but are viewable by the post’s account holder. Head of Instagram Adam Mosseri told BuzzFeed News that the test wasn’t about incentivizing specific behavior but “about creating a less pressurized environment where people feel comfortable expressing themselves” and focus less on like counts.

If this pans out, it may tone down the rewarding of vanity. But there are other pressures at play. In the current structure, conversations are often impossible just due to the scale and the nature of the feed — it’s nearly impossible to tell who is replying to whom, or to which comment.. Moreover, aside from a Twitter user name and profile picture, you may know nothing about to whom you are replying. Are they really a Russian troll (as you’ve just accused them of being), or are they a married mother of three in suburban Atlanta? Unless you take the time to look into that user’s profile, you do not know, and unless you look at their other tweets, you do not know if their comment (which so enraged you into shouting back) is at all reflective of who they are. Maybe you don’t know the context, or the person at the other end?

Haider explained, “If you can see who you’re replying to, then you may change the tone in which you reply. I know this anecdotally, when people tweet [redacted] to me, and I respond, right? Imagine if in this reply state, you could actually see ‘oh, this person’s a father of two.’”

Beykpour acknowledged that the feature feels “tiny” but said Twitter is starting to “tug on this string” related to empathy: “How do you surface context around who this person is, but ideally also give you a little bit of empathy around who you’re talking to? It’s harder to have an uncivil dialogue with someone that you know a little bit more about than if it was an anonymous face.”

He added, “Civility is one really important aspect of [healthy conversations]. It’s very difficult for people to feel like they can speak freely, if they don’t feel safe enough to speak in the first place.”

There have been a number of arguments on Ricochet and elsewhere about “deplatforming” and the rights of Twitter and others to kick off especially awful users like Alex Jones or Louis Farrakhan. Ultimately, though, I would argue that booting people off, or blocking them, is a last resort. If you get to the point on a platform where you have to ban a number of bad actors, then it is time to look at the platform itself and try to adjust it such that the bad actors themselves have less incentive to join the platform and abuse it in the first place — you cannot reward them, and you must find ways to reward those who try to keep the forum civil. It’s actually similar to how a lot of corporate employee handbooks evolve — there are rules in those books for certain behaviors because someone, at some time in the past, exploited how things used to work (I joke that my own company handbook has several “Person X Memorial Rules” because I can point directly to the series of incidents that led to that specific rule being in there). You have to have a system that rewards the behavior you want, or what your best customers want, or what your employees need and want.

It will be interesting to see if Twitter or the other social media giants can rebuild themselves, and I do wonder at what new incentives they are, perhaps unwittingly, creating. Will Twitter become less of howling mob than it now is? Hard to say. But all online forums, large and small, have to grapple with what behaviors they are rewarding, and what they are punishing, and whether they are even aware of it. Leopold II claimed he wasn’t rewarding the maiming of others (though his defense was itself horrible — he claimed that cutting off the hands meant his subjects wouldn’t be able to work for him, so he would have suggested cutting other things off). I’m sure Dorsey at Twitter never thought he would be psychologically rewarding the very behavior he now laments, though the evidence is now clear that this is exactly what his platform does. Twitter as it now stands rewards people for being terrible to each other. If they are able to remake their system to end that, I wish them well.