Managing an information environment like Twitter must be very difficult. The people who run the system have great control — and responsibility — over what the place allows and encourages. In a conversation platform (which is what Twitter is at its core), the primary question is: How do you allow for freedom of expression while also steering people away from harmful speech? This isn’t an easy question to answer. What is “harmful”? For whom? How and where does the environment intervene?
Episode 148 of Sam Harris’s Making Sense podcast features a conversation with Jack Dorsey, Twitter’s CEO, that addresses some of these questions head-on. I was very impressed by how much thought Mr. Dorsey has given to these issues. It’s clear that he understands the systemic nature of the challenge, and the need for systemic responses. He expressed Twitter’s approach with a medical analogy:
Your body has an indicator of health, which is your temperature. And your temperature indicates whether your system more or less is in balance; if it’s above 90.6 then something is wrong… As we develop solutions, we can see what effect they have on it.
So we’ve been thinking about this problem in terms of what we’re calling “conversational health.” And we’re at the phase right now where we’re trying to figure out the right indicators of conversational health. And we have four placeholders:
- Shared attention: What percentage of the conversation is attentive to the same thing, versus disparate.
- Shared reality: This is not determining what facts are facts, but what percentage of the conversation are sharing the same facts.
- Receptivity: Where we measure toxicity and people’s desire to walk away from something.
- Variety of perspective.
What we want to do is get readings on all of these things, and understand that we’re not going to optimize for one. We want to try to keep everything in balance.
I’d expect the idea to be to incentivize “healthy” conversations over “unhealthy” ones. This would be implemented in the design of the environment itself, rather than at the policy level:
Ultimately our success in solving these problems is not going to be a policy success. We’re not going to solve our issues by changing our policy. We’re going to solve our issues by looking at the product itself, and the incentives that the product ensures. And looking at our role not necessarily as a publisher, as a post of content, but how we’re recommending things, where we’re amplifying, where we’re downranking content.
Twitter has a great responsibility to get this right, because in some ways the system is becoming key public infrastructure. As Mr. Dorsey acknowledged,
Ultimately, I don’t think we can be this neutral, passive platform anymore because of the threats of violence, because of doxxing, because of troll armies intending to silence someone, especially more marginalized members of society. We have to take on an approach of impartiality. Meaning that we need very crisp and clear rules, we need case studies and case law for how we take action on those rules, and any evolutions of that we’re transparent and upfront about. We’re not in a great state right now, but that is our focus. I do believe that a lot of people come to Twitter with the expectation of a public square. And freedom of expression is certainly one of those expectations. But what we’re seeing is people weaponize that to shut others’ right to that down. And that is what we’re trying to protect, ultimately.
As a Twitter user, I was pleased to see the depth of the thinking and care that is going into these issues. I learned a lot from this podcast about the reasons for some of Twitter’s controversial design decisions. (E.g. I now know why Twitter doesn’t have an “edit” button.)
Unfortunately, the conversation didn’t address the elephant in the room: Twitter’s business model. Ultimately, Twitter makes money by showing ads to its users. A good public square shouldn’t attempt to sway our opinions; it should provide the venue for us to form them through engagement with others. How might “conversational health” might be used as a means for persuasion?