The Host's Privilege
Advancing our thinking about online moderation
It’s been funny watching Elon’s takeover of twitter. Not like the “haha” kind of funny, more like the “that’s funny, who let 4 drunk golden retrievers try to drive a car?”
Anyhow, this whole debacle started over the presumption that speech should be ‘more free’ in some ephemeral way, largely due to a bunch of people feeling like they were being oppressed. And yet, Twitter is a private business (then and now), so 1) it’s not clear it was ever a free speech venue, and 2) the whole free speech thing is more about government reprisal anyhow.
But what should speech moderation look like in online communities? It’s an interesting question. Surely the host of any party has some rights in determining what’s acceptable discourse and what’s not, and to date neither the former ‘shadow council’ approach nor the ‘rule by billionaire demagogue’ has been received as overly palatable, with both approaches causing a fair bit of backlash.
As a primer on moderation, I’d recommend reading this awesome blog post by Mike Masnick that details why moderation norms tend to converge over time: at the end of the day, it’s about running a functional business. If you want to run an online community at a scale sufficient to pay for servers and ongoing development and whatnot, you need a revenue model. That correct revenue model ends up being advertising, because advertisers are willing to pay more for access to an audience than that audience will (in aggregate) tend to want to pay to participate.
Can this model be improved? Yes, probably. One imagines you probably still need all of the things in Masnick’s post above, but after that there are some other things we could try. Two ideas seem interesting: audience participation and increasing the cost of fraudulent behavior.
Audience participation would involve a feedback loop for the community in determining what sorts of things they want to allow. Not overly long ago, the Rudy Havenstein account was banned for something or other, inspiring a small user revolt which led to it getting reinstated. This is a form of audience participation — the mods made a decision, and the community overrode that decision by objecting to it rather loudly. What if we institutionalized that?
Imagine a framework for deciding what sorts of things we want to see in a community, a sort of collective trial framework. Audience members are selected to offer input on some sort of issue like “is this lab leak discussion something we want to allow on here or do we consider it dangerous misinformation?” (Twitter shut this down and was accused of bias for doing so). This seems like the sort of thing that could be incorporated into a notion of moderation, and would allow for community-driven input on what the boundaries of the discussion are.
Similarly, raising the cost of entry for systemic ‘fraudulent comments’ seems interesting. Imagine a world in which signing up for twitter has a nominal one-time cost ($5 or something). This is fairly easy for a real user to pay, but makes running a bot farm to shill NFTs or whatever a much less economic proposition: each bot will cost $5 to set up, and if it’s banned that cost is gone forever. What this does is allow legitimate advertisers to outcompete (via raw economics) bot farms that can, in absence of the fee, write scripts to set up thousands of new accounts for free and simply not care if they’re eventually banned. Existing accounts could be grandfathered in, and bad actors slowly pruned while the cost of reentry is increased.
This is different from the current verification scheme which, aside from being a hilarious UX failure and causing open rebellion, charges a persistent fee and also doesn’t really do anything to clean the place up either.
In short, I think there are some concrete steps that could be taken to make a place like twitter great. Those steps are:
Return to the former moderation framework as a baseline.
Allow for that framework to be evolved incrementally via audience participation and moratoriums on things that aren’t working well.
Increase the economic cost for bad actors in a way that increases the relative value to advertisers (bad actors can no longer compete against them for free), and doesn’t have a persistent pay-to-play component for users.
I’m not sure what’s next for Twitter, we seem to be in a strange sort of crisis moment where the current owners really want to express their vision for the place, and… well I suppose that is their right in some sense.
It’s just a silly expression of the host’s privilege. Ultimately, just as we’ve found with representative governance, the highest value outcomes are ones that reflect outcomes desired by the whole. When we can find ways to express the will of the people and create economic disincentives for bad actors, that’s generally good for commerce.
And Twitter is, after all, a business.
Or at least trying to be.
I was a moderator on one of the StackExchange communities for a year or two. The StackOverflow/StackExchange management team had put a lot of thought into moderation, and set up a framework that seemed reasonably successful.
They'd realise that moderation effort had to scale with community size, so volunteer moderators were drawn in from each community's most responsible and dedicated users, as it grew. They shared a single online chat-room ("The Teachers' Lounge") for moderators across all Stack communities, where they could support each other and seek advice from their peers, and from staff members. And the staff members acted as supervisors, and Appeal Judges, when a moderator's judgement was questioned.
And crucially, there was quantitative support too: any user could raise a flag for content to receive a moderator's attention: the handling moderator would mark the flag as helpful or unhelpful; and a user's running proportion of helpful flags, weighted to more recent results, was used to prioritise that user's future flags.
One other thought here, Travis. I liken comments to shopping. If you go to a discount retailer looking for clothes, you find a lot of gems in a mass of stuff. But it takes hours and hours to sift through. If you're time constrained, sifting through stuff is something you just can't do so you shop somewhere where you're likely to find good stuff in a very short period of time even if it is more expensive to shop there. If you take the same thought to comments, then Twitter is like the discount retailer but one with tons of stuff you don't want. The key for Musk should be to make the 'shopping' for replies to power users beneficial enough to keep them on the platform because I think a small subset of active non-company users drive engagement. And it's those users who are least likely to get quality mentions.