Before talking more about self-identifying, it’s worthwhile to consider why we even think that there is anything to identify  — a “self” that we can attach labels to. This is not a given; many people “…report no individualized sense of self, no self-related thoughts” This is from a paper on Persistent Non-Symbolic Experiences, a sciencey-sounding term for what is normally called “enlightenment”.

Who’s “enlightened”? If you ask them, any and all of:

• An old monk wireheading in a cave, oblivious to suffering and desire.
• A rationalist breaking through his default ontology to play with frameworks at will.
• Me, after eating way too many mushrooms.

Whatever do these all have in common? Here’s my Predictable Identities conjecture: the sense of self comes from habitual and repeated thoughts; novel thoughts make the self dissolve.

“Habitual thoughts” means everything from high-level cogitation to low-level processing of stimuli. Kaj Sotala points out that something as simple as the perception of hearing a stick hitting a woodblock is the result of complicated processing that combines several auditory frequencies, your previous experience with wood objects, and mental imagery. This processing usually happens automatically: [hear sound] -> [imagine block]. 

But, with enough concentration, one can break this habit, leaving just the raw sounds to do with as one wishes. The same is true of “habits” like [feel pain] -> [experience suffering] or [see object of desire] -> [experience craving] that are the common focus of the monk’s meditation, or the frameworks of social interaction Valentine writes about. 

When the part of your brain that monitors itself notices repeated patterns of thought it creates a high-level model called “self” that it can use for prediction. “I” am interpreting the sound as a stick hitting a woodblock. “I” suffer when in pain. “I” think of everything in terms of predictive processing. And “I” will likely continue to do so in the future.

But when a thought or interpretation arises that can’t be predicted from the habits of my mind, there is no reason to assign it to a consistent thinking “self”. The thoughts I’m used to thinking are mine, but the novel ones could be anyone’s or no one’s.

The last post set up the motivations and trade-offs of applying identity labels to yourself and to others. One can choose to oppose this ubiquitous labeling or indulge in it. With these two axes, we can sketch out the blogchain’s first (though inevitable) 2×2:

The anti-label corner is exemplified by Paul Graham’s famous exhortation to keep your identity small and Scott Alexander’s reminder that categories are just (occasionally) useful abstractions. This corner sees other-labeling as a distraction from more pertinent questions of what someone actually does or believes; they see self-labeling as a sacrifice of clear thinking for tribal conformity. “Nerd” is a bad label for this corner, but being resistant to labels is its very nature.

Identity politickers sit at the top right corner. They reinforce the centrality of identity built by stacking labels (intersectionality) and demand that others be loyal to their identity markers, as in the quote about black faces and voices. Identitarians are not fans of rich white guys like Paul and Scott, but these two at least stick to their expected role of adversary. They have much more contempt for identity traitors like Candace Owens.

On the top left are people who gleefully toss labels at their outgroup to reinforce its homogeneity and overall wickedness. This usually serves to distance oneself from a label. On the bottom right, a person can attach a self-label to some conventional opinion to demonstrate how nicely predictable they are and to associate with a group. 

These two corners are less of a committed stance and more of a tactical application of labels, but the same is true across the board – people change their willingness to label themselves and others depending on the situation. Labeling is a political tool; we should expect hypocrisy to be the norm rather than the exception.

After consistency, the most important feature of predictable identities is a good label, a pithy description that tells others (and yourself) what to expect of you. 

Let’s say that you are consistent in the sentiment that individuals should be free from government intervention, and thus oppose business regulation and gun control. What should your position be on abortion, immigration, or campaign finance? It’s difficult to extrapolate a coherent position from your basic values, and often not worth the effort  — your take on campaign finance likely has zero impact either on government legislation or on your own life. It’s uncomfortable for your political stances to be so unpredictable.

On the other hand, you may simply adopt the label of “Republican” and acquire a set of stances on all political issues. The Republican position on anything is common knowledge, and anyone who knows that this label is part of your identity should not expect any surprises. Sticking to the label is often valued much more than consistency of actual opinion:

People strongly dislike labels that don’t actually help prediction, as illustrated quite hilariously by the recent backlash to “sapiosexual”. From what I can tell, here’s what people who call themselves that mean by it:

• 10% are sexually aroused purely by intelligence, not appearance (yes, they exist).
• 20% like hot people but only if they’re smart.
• 30% weigh personality more than physical attributes in romantic partners, relative to others.
• 40% use the term merely to signal their own intelligence.
• 1 person (me) insists that it should mean “attracted exclusively to Homo sapiens”.

The end result is that people who haven’t met a single self-identified sapiosexual write articles titled You’re Not Sapiosexual, You’re Just Annoying in their frustration at the unpredictability of the label.

Self-labeling is a powerful tool for shaping your behavior and the reactions of others, to be used with care.

We like and reward people who are consistent, categorizable, and cooperative because it’s easier to predict them. And since prediction shapes action shapes reality, we make other people be so. This was the first half of this blogchain; the second will cover how we do this to ourselves.

Let’s start with consistency. Students who did well on an exam said that they weren’t anxious before it, those who did poorly said that they were. Pre-exam questionnaires, however, show that these memories are false. We tell ourselves (and our friends) how we immediately fell in love with our partner, even if in reality the first couple of dates were awkward and skeptical. We tell anecdotes about things we did as kids that predict our current vocation, even though at age 6 we were mostly obsessed with spiders and the Power Rangers. We also predict that our attitudes will persist into the future to a much greater extent than they actually do.

Self-consistency bias affects not only our memories, but also the decisions we make. In fact, we couldn’t make decisions at all without assuming some consistency in our preferences and abilities. And by convincing ourselves of our consistency, we reassure others as well. 

I’m writing this post because I predict that, at least for several more months, I will be as obsessed with identity and predictive processing as I am right now. You’re reading this blogchain partly for the same reason  — you predict that new posts will keep coming every other(ish) Wednesday and will stay on topic. If I actually analyzed the half-life of my fascinations with esoteric topics, I may discover some discouraging patterns. It’s better for all of us that I don’t. 

I have always been writing Predictable Identities. 

I will always be writing Predictable Identities.

Isn’t that reassuring?

Predictable Identities is 16 posts in, which is going much better than I would’ve predicted. It’s time to review what we covered so far: the principles of predictive processing and how we apply them to other people.

Our brains constantly predict sensory inputs using a hierarchy of models. Learning new and better models improves our predictive ability in the long term but can be so painful in the short term that we will fight against updating, and often fight the people who force us to update. It’s important to take all models with a grain of salt and resist the lure of all-explaining ideologies.

We predict the world to exploit and act on it, and the same applies to other people. We need to know how to get people to do nice things for us, using stereotypes for strangers we don’t know and more detailed mind-simulations for people we do. We don’t need detailed models for people unlikely to be nice no matter what, and we’re creeped out by people who don’t fit our models at all like those who blow their non-conformity budget. We encourage those around us to conform to our narratives and predictions of them, which means that changing one’s opinions and behavior takes great effort in the face of the expectations of your social surroundings.

Finally, our predictions of ourselves interplay in complex ways with how others see us, since our minds are neither transparent to us nor opaque to those around us. For example, to be treated nicely we must honestly believe that we are nice, even if that is self-deceiving. Our models, predictions, and beliefs about our selves form our identities. This will be the topic for the second half of this blogchain.

Thanks for joining me on this journey, I predict exciting things ahead.

A super-predictor millionaire child may seem like a fanciful example, but the general principle of Newcomb’s problem applies to almost every social interaction. People do nice things for you if they predict you are trustworthy, and the only way to be predicted as trustworthy is to be so: leave the cash, work hard, maintain the apartment.

Why not just deceive people? Lying is hard. People put significant cognitive resources into reading your mind, from processing your uncontrollable microexpressions to explicitly simulating your behavior. They also do so as a group, sharing information about your reputation. Conscious deception requires keeping track of two separate stories about your intentions. This imposes an extra cognitive burden when you are already facing a lot of combined brainpower that’s trying to predict you.

One way to lessen the effort of deception is to keep track of only one version — the wrong one. Self-deception is indispensable in social situations, believing yourself to be smarter, better informed, more talented and moral than you really are so that other people treat you as a smart, informed, talented and moral. But for self-deception to successfully affect people’s predictions about you, it must not leak into your conscious awareness. You must sincerely believe you will leave the cash on the ground at least until the child disappears.

If you make it too hard for people to predict you well, you will not be presented with Newcomblike opportunities in the first place. The child wouldn’t offer you money if they couldn’t hack your phone to study you. You won’t get the job offer if you refuse to drink at the networking event and “get to know each other”. The reason I write personal things on the internet every week is to make it easy for faraway strangers to predict my thoughts and values, and to occasionally offer me nice things.

“Do you like money?”

You spin around to see a strange child. You could’ve sworn you were alone on this street just a second ago. “Everyone likes money, kid,” you reply, “but I’m not getting sucked into another MLM scheme.” 

“This isn’t MLM, it’s LDT!” The child presses an envelope into your hand. “Open it.”

You do, it contains several $100 bills. “What’s the trick?” you ask. “And what’s your name?”

“The name’s Newcomb, and the trick was hacking into your phone last month.” The child smiles. “I’ve studied your behavior, and made some predictions. Specifically, I predicted whether you will take the envelope or leave the bills on the ground and go straight home after you hear me speak. If I predicted you would do the latter, I deposited $1,000,000 into your bank account 5 minutes ago. If I predicted you’ll take the cash, I retweeted Trump from your Twitter handle. By the way, I’ve been playing this game for a while and never got a prediction wrong. Goodbye!”

You look down at the cash. “Wait, what if…” but as you look up, the child is gone like an expired snap.

Do you take the cash or leave it and go home? Or do you hate philosophical thought experiments? Alright, let’s talk about something completely different.

Your employer decides on bonuses in December, but the bonuses are only paid out (and found out) in March. If the company predicted that you would slack off in Q1, your bonus is small, and if they thought you’d hustle it’s generous. You planned to quit at the end of March anyway, how hard do you work until then?

A leasing agent says she has another application for your dream apartment, but she’ll let you rent it because she feels you’ll work extra hard to maintain the furnishings. Do you?

Every rich-enough belief system wants to be the single source of answers to all questions. To fight against that it’s important to remember that every belief system is fake.

The first step is to notice that every single thing can be described using different frameworks, the way a single plot of land is represented using different maps.

Which of these two maps of Princeton is more “true”? Neither, since neither one is actually the town of Princeton itself. They each contain different elements: one has trees and roofs, the other has streets and parks. Certainly, Princeton contains all the above, and many things besides.

If you visit Princeton, you would notice other things it’s made of that don’t show up in the maps, things like people and pizza and noise. But those are also mere “maps” that your brain projects on the town, maps that miss a lot of detail. For example, you miss the fact that both the trees and the pizza are made mostly of carbon atoms. And if you think that “carbon atoms” are a real thing that exists you are at least a century behind on physics.

The fact that frameworks made of cities or trees or carbon atoms are fake doesn’t mean that they’re not useful. Your brain can only predict the world using concepts like “pizza” and “person”, fake though they are. But remembering that all frameworks are fake allows you to nimbly shift among them instead of believing that any single framework contains the only thing reality is really made of. Your brain won’t like this juggling effort at first, but it will appreciate not slamming into massive prediction errors caused by the limitations of a single ontology.

And if “pizza” is fake, how much flimsier are maps made of concepts like “conservative”, “chaos”, “female”, “sin”, “petit-bourgeoisie”, “freedom”, or “privilege”?

More than inspired by In Praise of Fake Frameworks