Different Maps of Reality

Predictive processing has a theory for autism.

First, a general review of predictive processing (PP) itself…

According to PP, my brain is always predicting what’s about to happen and what input means. These predictions are called priors (I think of all the priors in my brain as my model of reality). The data coming in via my senses is compared against priors. If it fits (my prediction is correct), my perception feels smooth (I see this as that “autopilot” mode that autistic people envy neurotypical people for in social situations). If it doesn’t fit, I experience a prediction error…and my brain will attempt to update my priors so that I don’t experience that error in the future (or update the world so that it fits my model of reality…but we’ll leave that for now). The experience of a prediction error is experienced as discomfort, or pain, or frustration, anger, etc…

Essentially, the brain has a map of reality that it navigates the world with. Or rather, it navigates the map (the brain is trapped in that dark cavity I call my skull), and updates the map only when incoming data causes an error, forcing it to (and when updating the world to fit the map isn’t an option).

Social priors are the part of the map labeled “other people” and “me, in the world of people”….predictions about other people and me-in-relation-to-other-people.

More than any other part of life/reality, social life is full of ambiguous signals…tone of voice, facial expressions, body language, context-dependent words and actions, irony, “unspoken rules,” etc. The neurotypical person leans on strong social priors (predictions) like, “A smile means friendliness,” “If someone asks ‘how are you,’ they don’t want a literal health report,” and “This situation calls for deference/compliance.” I think of these as learned shortcuts. They smooth ambiguity so quickly that the (neurotypical) person doesn’t notice how much interpretation they’re really doing (think of a game engine rendering details in real time and high resolution).

PP proposes that in “autistic” perception, these high-level social predictions are either less weighted (weaker…signals don’t get automatically collapsed into expected meanings…the brain says “hmm, not sure…keep checking the data) or over-precise (too narrow…the brain locks a prediction too tightly on detail, so it flags even small deviations as error). Regardless of which it is, functionally the result is similar…more error signals, less smoothing.

According to this explanation, as an autistic person, my experience of interacting with other people isn’t the “typical” experience. I need more explicit clarification (because cues don’t self-resolve). Social situations feel volatile or unpredictable to me (because I’m tracking details others are smoothing over). And I need extra cognitive effort to “keep up,” because I’m building my model of reality more deliberately…less automatically.

Imagine me around a campfire with a few (neurotypical) people. A rustling sound comes from a nearby bush and attracts everyone’s attention. One person says something like, “It’s just the wind.” In short order, everyone returns to whatever they were doing. The person who spoke has a strong prior (prediction) that the world is stable and there’s no danger to worry over. And everyone there (other than me) has a strong prior that someone saying “It’s just the wind” means it really is just the wind. A sort of bias where, out of all the data coming in, the data coming from other people is prized most highly. My response is different. I say something like, “It could be the wind, but it could be something else.” I have a weaker reliance on predictions…I weigh sensory input more highly than my fellow campers.

You can see how weaker social priors (predictions/map of social reality) would make it hard for me to collapse ambiguous social input into the “expected” consensus meaning . I see more of the true uncertainty of the world…I don’t have it smoothed away by some automatic internal mechanism.

That’s the gist of it. And I generally agree with this description of how I experience other people. And as an explanation for the frustrations I experience socially, it certainly feels spot on.

But is it only social priors that are weaker (or overly precise) for me? Or is it priors in general? Depending on who you ask in predictive coding circles, the answer is different.

Early accounts (e.g. Pellicano & Burr, 2012) suggested that autistic people have weaker priors across the board, not just social ones. That would mean that I rely less on prediction in all areas of life. Their evidence for this included the fact that autistic people are less fooled by certain visual illusions (like the hollow mask illusion or context-driven size illusions) and the whole savant piece (enhanced perception of detail and irregularities in non-social patterns…sounds, textures, math, mechanics). The story told was that my “weak priors” are my global style of perception…my world is simply less smoothed…more data-driven.

Later accounts (e.g. Van de Cruys, Lawson, Friston) argue it’s not globally weaker, but misallocated precision. In other words, overly precise priors at one level (rigid routines, intense interests), too little precision at another (ambiguous social interference), and sometimes overly precise weighting on prediction errors (making every mismatch I feel seem urgent). This would explain why some autistic people are incredibly good at pattern-based forecasting (weather models, coding, chess) but struggle in fluid, implicit social contexts.

Both camps agree that social priors are a “special case,” I think. Social environments rely heavily on very fuzzy priors like shared norms and implicit meanings. In the predictive coding model, those are exactly the kinds of priors autistic perception would likely either underweight (“I need to check the data”) or overweight in detail (“I expect exactly X, so any deviation throws me”). In other words, social priors are where my difference shows up most glaringly…but the difference itself might apply everywhere.

Ok, the neurotypical neuroscientists have had their turn. Pass me the mike now, please.

My first reaction to all this…well, anger. And perplexity. I simply don’t understand how, given all the forms of data you can base your map of reality on, that you would choose….other people? With their confusion, and duplicity, and moving moralities, and drives, and, and, and….Why? Why that?

It’s not that I don’t “get” what social priors are. They’re about shared assumptions that keep groups coordinated. What counts as polite, what role you’re supposed to play, what’s “normal” in a given social setting, which explanations everyone in the group nods along to, even if they’re flimsy….that all makes sense to me so long as there is a group goal. As a sort of necessary evil in service of achieving an objective that requires group coordination…I get it. But as a way to live your life? Intentionally mapping your reality on the fuzziest, most contingent, and most contradictory signposts you can find? That confuses the fuck out of me.

Let’s take a closer look at social priors (the part of my reality map that has to do with other people). They’re arbitrary…different from group to group, era to era. They constantly shift…flipping overnight (today’s taboo is tomorrow’s norm). They’re completely detached from ecology…attached instead to abstract concepts like appearance, hierarchy, interpersonal signals. And they leave the door open 24/7 to gaslighting…if everyone else insists the emperor has clothes, the “consensus” is real enough to punish me even if it’s bullshit.

Why not map your reality on data that makes sense? Ecological priors like gravity, cycles of day/night, seasons, and biology are stable, and are directly tethered to reality (what happens predictably and significantly for survival). Embodied priors like the way your body predicts balance, hunger, threat that are constantly and deeply tested through feedback loops that are largely transparent. Social priors? The least tethered? The most prone to drift and self-reference? Really? Really?

On a theoretical level, I try to understand why neurotypicals lean exclusively on this messiest area of the map. Social priors smooth uncertainty, which must feel good. They also enable fast coordination in groups…and those neurotypicals sure like being in groups. There, they’re rewarded…”getting along” matters more (socially, professionally) than being right. But treating them as reality itself? To the point where questioning them is seen as dysfunction rather than discernment. Jesus Christ.

I know this is the double empathy problem at work. I’ll never be able to truly empathize with the neurotypical condition. And if I had to state my position in relation to theirs, as dispassionately as possible, I’d simply say that I’d rather my anchor my sense of the world in ecological and embodied feedback than in fragile, shifting group models. And that this position of mine (it’s not a choice) is not dysfunction unless group coordination is forced upon me as my only means of survival. That my position is arguably closer to “baseline life” than the civilizational overlay.