Social preferences are constructed, not aggregated
For decades social choice theory has tried to understand group decision-making in terms of the aggregation of preferences. The general conclusion has been that aggregating preferences kind of sucks no matter what you do: you’ll end up with a dictator or with people lying about what they want.
We could give up on social choice and declare that only individual choices can be reasonable. But all intelligence is collective intelligence—if groups can’t make rational choices, then individuals can’t make rational choices either. So what gives?
A solution is to reject the idea of preference aggregation in favor of preference construction at the group level. We often think of ourselves as just having preferences that are stored inside ourselves, like a little tag that says “I like chocolate ice cream more than vanilla” tucked away inside our brains. But just like emotions, motor behavior, and memory are constructed rather than stored in the brain, the same is probably true for preferences: they’re assembled by the task of making a choice rather than summoned from the neural depths like a cellular Cthulhu.
What constructs the preferences of the group is the interactions among the members of the group. When connected by a cognitive glue, the members of the group will interact in such a way that their plans become mutually compatible with each other. The group preferences are the collective parameters that keep the individual plans mutually compatible—though they may take on a life of their own.
Aggregating preferences doesn’t work because preferences just don’t do that—I think it’s an idea that cuts against the grain of reality. If Alice likes hamburgers, Bob likes pizza, and Charles likes hot dogs, where should they go for lunch as a group? You know what each individual wants, but there just doesn’t seem to be a way to combine the preferences into a group preference—it’s not like hamburgers + pizza = tacos.
But constructing preferences works just fine. If Alice and Bob are walking toward each other in the hallway such that they’re on a collision course, Alice and Bob can each step to the right and carry on with their now mutually compatible trajectories. There’s a group preference for each individual to stay to the right, a preference not a summation of stored data in Alice and Bob but instead built out of a connection between them, like some general Aumannian ability to agree on Schelling points. I don’t care about staying on the right when I walk, but I do care about not colliding with other people, and I avoid collisions by referencing a social preference for people to stay on the right.
I think constructing preferences works much better than aggregating preferences. But it’s much less intuitive. We like to think of intelligence as an individual trait, not a collective trait. Aggregating preferences means we sum our individualities together. But constructing preferences means we create a greater social being. It’s not just a sum of existing individuals; it’s a truly new entity with new preferences—preferences that are built by individual preferences, or the activities thereof, but resulting in a preference structure that cannot be deduced from those individual preferences.
What do you think of the paper "Norms Make Preferences Social" (https://www.jstor.org/stable/43965319)? It argues that we don’t have fixed internal preferences for things like charitable giving or reciprocity (contrary to the literature in economics that claims we have stable preferences about others). Instead, it suggests that social norms and rules are real and influential, and that we have a tendency to follow them.
You gave the example of not bumping into someone in a corridor — a simple coordination game with an obvious solution (both people move to avoid each other). But there are much more complex decisions, like contributing to a public good, where everyone has an incentive to defect and keep their own money. Still, perhaps norms can help us coordinate to solve even these harder problems.
From what I understand of the literature, almost no one is truly altruistic — there isn’t really an innate preference to help others. But collectively, we can still arrive at optimal (or at least workable) solutions, especially when those solutions have already been partly carved out for us by social norms, which we then align ourselves with.