An economic perspective on the binding problem

The binding problem is the question of how “items that are encoded by distinct brain circuits can be combined for perception, decision, and action”. In economics, the answer to this kind of question is basically always “prices”—the production process that produce sausages need not overlap with the process that produces loaves of bread, yet the coordinative powers of prices make it easy to combine sausages and bread loaves to create hot dogs. There is no particular point at which or process by which these production activities are united into a single one.

More generally, the economy is the process of the coordination of many different plans. Those plans can be wildly different from each other in terms of how they are constructed, what the specific goals are, and what kinds of worldviews and sensory modalities those plans reflect. The economic binding problem should be the problem of how and where these plans are brought together into a single plan, but of course, nothing like that happens. There is no economic binding problem. Instead, these plans act as agents that coordinate with each other to construct a group-level plan, a process that happens everywhere and always throughout the economy, not at a particular point and time.

The actual economic equivalent of the binding problem is the problem of figuring out how to use a voting procedure to aggregate social preferences into an economically rational social choice. The key results of social choice theory show that this is basically impossible: you cannot aggregate preferences via a particular process to produce some kind of combined outcome. But economic coordination happens all the time and is totally non-mysterious.

It’s the difference between aggregation and construction. You can’t glue people’s pereferences together. If you know that Alice wants pizza for lunch, Bob wants hamburgers, and Charles wants hot dogs, there’s no way to stick these preferences together into a combined Alice-Bob-Charles lunch preference. Instead, group level preferences are constructed out of the coordination of the members of the collective. The old stuff doesn’t get combined; rather, something new gets built.

In economics, what makes this possible is the bow tie structure of prices. Prices are constructed out of a huge amount of information—all of the information that goes into supply and demand—but they don’t exactly sum up that information. Instead, a price is just a simple number. There’s no way to disaggregate a price back into the information that created it because it’s not an aggregation at all! What prices remember about the past is entirely a function of how they’re used by the economic agent making plans based on prices. The result is a structure that looks like this: prices can take in pretty much anything and be used to produce pretty much anything.

Taken from here

You don’t need to ask, “where and how does all the stuff on the input side get combined?” It’s not what’s happening. Nothing is combined, everything is coordinated.

The same kind of economic coordination is going on in the brain and body. If we imagine—however incorrectly—that one kind of neuron “codes for” color, and another codes for shape, etc., then we do not need to ask how color and shape get combined into a single perception. They can be coordinated into a single perception. This coordination doesn’t happen in a particular place in the brain at a particular point in time. It’s happening all throughout the brain and body, an economy of cells.

This idea becomes more challenging when we realize that it implies we probably need to reject the idea that familiar psychological categories are being combined in any sense. Historically, many psychologists and neuroscientists assumed a more or less one-to-one relationship between our psychological categories and our psychological processes. That is, for every category we have to describe our psychological experiences, like color, shape, emotion, memory, action, etc., there must be some mechanism in the brain that is responsible for producing phenomena that fall into that category, like color neurons for color and shape neurons for shape. (This perspective makes a lot more sense when you assume that cells are dumb parts with simple jobs assigned to them by the genome rather than agents that figure out what to do in the context of a developing organism.) The binding problem is a direct consequence of this background theoretical assumption: if there is a color mechanism and a shape mechanism, then there has to be a third mechanism, a combining mechanism, for putting color and shape together so that we don’t mix up red squares with blue triangles.

But in reality, there are no mechanisms in the brain for doing anything. Phenomena that we categorize as if they’re innately distinct are in fact all produced by the same general process of interoception-and-allostasis. By taking economics as the background assumption, we can see that it makes no more sense to think there are color neurons and shape neurons than to think that there are sandwich people and tax preparation people and online influencer people. People can only form these kinds of plans by coordinating with other people: sandwiches and tax preparation and online influencers are produced, not received from the environment. The same is probably true for color and shape and every other part of vision as well. (Otherwise, you run into the degrees of freedom problem, needing a grandmother cell to recognize a grandmother.)

Group preferences, in general, seem to be wildly different from the preferences of the parts that make them. Humans are made of cells, for example, but we have no innate knowledge of cells and no innate consideration for them—even after learning that you’re made of cells, you probably won’t shed any tears for them if a few million cells die when you scrape your hand. If our senses are also group-level constructions, then there’s no reason to believe that if we see color and shape, then neurons must in some way perceive color and shape as well. It just doesn’t have to be that way—or if it does, we don’t know why.

The perception-cognition-action fallacy is another example of a binding problem created by a theoretical supposition that our categories as scientific observers have to each be produced by a distinct dedicated process in the brain and body: one for perception, one for cognition, and one for action. Perception, cognition, and action don’t have to be bound together into a coherent, goal-oriented stream because they are themselves produced by the general process of interoception and allostasis as the economic coordination of the cells, organs, and subsystems in the brain and body.

And indeed, recent evidence suggests that the binding problem may not be a problem at all once you relax the assumption that vision has to be an aggregation of what we think of as the parts of vision, and with each part of vision produced by a distinct mechanism in the brain.

In the end, there’s no binding to be done because nothing exists to be bound. Information isn’t stored in the environment waiting to be discovered and put together by an intrepid mind. It’s created by the economic coordination of the parts that make up a self.