The Conversation Game

I've been thinking about filter bubbles aka. echo chambers. This is the phenomenon where, on social media, you find yourself connected to a group of people who mostly share the same opinions and values as you do.

It makes sense. It's nice when you say something and people agree with you. It also gives us a little dopamine hit when someone says something that conforms to our own opinions. There's a corresponding negative feeling when someone disagrees with us, or holds an opinion we find incorrect, or even offensive (especially when it's your mum, but that's a different problem).

The trouble with these bubbles is that there's an insidious tendency to make the world seem a lot simpler than it is. If I look at my twitter or facebook feed (and presumably this is the same on other social networks), the world is filled with people who all believe the same things I do. It's easy, right? [Insert politician here] is obviously a lying moron, while [insert other politician here] is clearly the better choice. Everyone knows this and everyone agrees, because I keep reading tweets and posts that say so - therefore it must be true.

Except, it's not. The world is a lot more complicated than that, and for any position or opinion I hold, there are a ton of people who believe the opposite - and they're probably not all idiots.

This oversimplification might be the underlying reason why filter bubbles tend to push people towards extremes. Put a mildly right-wing person in an echo chamber full of right-wing people for a few months, and if you test them before and after, you'll probably find that their positions have hardened and polarised.

This year (2017), we had actual nazis marching in the streets in the US, and far right political parties winning seats in Germany. I think there's a lot more to this than just social media echo chambers, but I also think they contribute to the problem.

So, what can we do about it? We can't just tell people to go and get out of their filter bubbles and engage with people they disagree with. Well, we can tell them to do that, but they won't.

Stepping out of your nice, comfortable, self-affirming echo chamber is hard. Being disagreed with makes us uneasy, and being forced to defend our opinions is difficult. We don't like being uneasy, or doing difficult things, so we don't.

I'm just as guilty of this as anyone else. Right now, I'm reading 'The Machinery of Freedom' by David Friedman, which is a book I largely disagree with. I'm finding it quite hard going. It's difficult even to read a book whose arguments I dislike. How much harder would it be to engage with a real person about this stuff in a constructive way?

And yet, that's exactly what we need to do. One of the first rules of the internet these days is, "never read the comments" because the quality of the conversations is so awful that reading them is the mental equivalent of wading through a sewer.

So, I've been thinking about ways to nudge things back the other way - to encourage meaningful debate across the boundaries of our filter bubbles, and this is one idea I had.

The Conversation Game

It's a crappy working title - do you have a better one? Here's how it works;

• Alice signs up to play the game, and picks a few topics she cares about, and indicates which side of those particular fences she is on, and how strongly she feels.
• The game server picks an opponent for her, by looking at twitter and (probably using machine learning) choosing someone who holds the opposite view on one of Alice's topics - abortion, maybe. Depending on Alice's cumulative score, the server picks someone with a higher or lower number of followers. Let's say the server picks Bob (who doesn't even know about the game, at this point).

Now it's up to Alice to engage in a dialogue with Bob about abortion, by replying to one of his tweets. Alice wins points in the following way;

• Points for keeping the conversation going - so some score every time Bob replies to Alice
• A few more points if Bob loses his cool, and just insults Alice with an ad hominem attack. There's no point in trying to continue the discussion at this point, so Alice just wins, her score increases, and the server picks someone else for her to play against. Bonus points for instances of Godwin's Law.
• A lot more points if Bob's position shifts in any meaningful way - e.g. if he acknowledges one of Alice's points as valid.

The rules for players would include;

• All tweets must be polite and courteous. The whole point of the game is to foster discourse, not shouting matches
• No lying. If you state a fact, you need to be able to back it up (preferably quoting a source at the time)

Scoring could initially be done by human moderators, but ultimately it ought to be possible to use machine learning to categorise the tweets and score the round accordingly.

As Alice's score increases (and she could lose points too, if she breaks the rules), the server will award her stronger and stronger opponents, where 'strength' is measured by either the number of twitter followers they have or, if they're also players of the game, by their score. This should provide some degree of graduated difficulty.

That's as far as I've got. I think there's something here, but I'm sure it needs more thought. If you have any ideas about this, I'd love to hear from you. That goes double if you're interested in building it. I like the idea, but I know I'll never have the grit to actually make it a reality by myself.