I have a secret technique that will cause any group of engaged smart people to start squabbling. I haven’t seen it fail to work yet, with the caveat that I’ve gotten bored with seeing the same patterns of squabbling recur, so I haven’t deployed it much recently.
Stop me if you’ve heard this one before: is a hot dog a sandwich?
No wait, come back. It’s a meme question because it works. Well. It doesn’t really work any more because now everyone knows everyone knows it’s a meme question. But asking questions of that nature does work. Consider a fictional example:
- Mallory: Does a tuna salad count as a salad?
- Alice: Hmm. I don’t think so, since the tuna’s cooked and salads only contain raw things.
- Mallory: Interesting. So is a beet salad a salad?
- Alice: Only if the beets are raw? That suddenly doesn’t feel right to me.
- Bob: Wait, no. Those are both obviously salads. They have salad in the name.
- Mallory: Huh. Then would you say a fruit salad is a salad?
- Carol: That can’t be a salad! It’s sweet and salads can’t be sweet.
- Alice: What? You’re crazy. That’s definitely a salad.
You can see how it would work. If everyone leans toward accepting, come up with less and less salad-like examples until you uncover an argument. And do the opposite if everyone instead leans toward rejecting. Is a pile of croutons a salad? Is a chicken fajita a salad? Is a salad wrap a salad?
Why does this work so well? I think it has to do with how people tend to deal with the fuzzy boundaries around categories. My current mental model of how most people think about categories is: ask system 1 what feels right, then get system 2 to come up with a post facto justification. This results in inconsistencies and contradictions if you drill down into a specific category with just one person, and results in discord in a group setting.
If you see something like this happening and want to stop it, the most common defense is to categorically reject this kind of categorization question. While this does successfully protect you from my fellow agents of chaos, it does not protect you from FOMO. And, I think we can do better. Categories are useful tools for making sense of the world, and we don’t want to dismiss interrogating their inner workings out of hand.
My current mental model of how to think about categories in a healthy way is informed by semantics, a subfield of linguistics. The theory goes, each category I have easy mental access to is represented by a prototypical element in my brain. So when I think of “sandwich”, I think of a prototypical ham and swiss on sliced white bread with lettuce, tomato, and mustard. When I want to see if something else fits the category of sandwich, I mentally compare it to the prototype and produce a binary yes or no based on its similarity. The individual variation we observe results from how everyone’s internal learned encoding of “similarity” can be very different from each other and still yield the same result for nearly all relevant questions, nearly all of the time.
The way I cut this particular Gordian knot is that I reject your sandwich binary. I instead think of that ham and swiss as 1.0 a sandwich, a BEC bagel as .9 a sandwich, a taco as .6 a sandwich, a salad as .2 a sandwich, and so on. It’s our old friend cosine similarity, modeled in our own organic brains. Then you can set your cutoff for whether something is a sandwich based on what the question is asking about. You can even re-weight each property of sandwichness depending on what the question is. Sometimes you care more about whether the maybe-sandwich is edible and sometimes you care more about whether the maybe-sandwich will come apart if you throw it. I will not be taking further questions at this time.
And now you know the secret. Please don’t actually cause discord on purpose, unless you really want to, or it would be really funny.