Going at the Bar

That's the heart of the debate. For instance, John Searle (1995) famously described a similar scene. Searle notes how cumbersome life would be if we were constantly aware of every norm and cultural convention. Paraphrasing a little, he compares us to fish in water, immersed in a shared 'We' or COLLECTIVE INTENTIONALITY (CI) that transcends individual intentions. This is why he argued that French cafés—and social institutions more broadly—are IRREDUCIBLE to just individuals; they possess an institutional reality. Intentionality has long fascinated philosophers. Consider the following. When 'I raise my glass,' my arm goes up. Ludwig Wittgenstein, half a century ago, asked what remains if you subtract the fact that my arm goes up from the fact that "I" raise my glass. The answer lies in individual intentionality. In a similar vein, we can ask: 'When the three of us raise our glasses together,' we raise our arms together. What remains if you subtract the fact that our arms go up from the fact that we raised our glasses together? Wouldn’t you agree that this 'We' who raise our glasses feels different from saying that each of us decided on our own to raise a glass? The answer lies in SHARED INTENTIONALITY. You get the idea. There's more. We can ask: 'What's left of the café if you subtract the fact that "You" and "I" are engaged in cultural conventions and norms, such as cheering together and perhaps philosophizing at the bar?'

I love what you’re saying, Philosopher. I agree that the Café could be thought of as more than the sum of its parts. In our field, we call this "emergence"—but I’m not convinced that intentionality is necessary to explain it. We’ve recently started modeling groups as irreducible units, much like in philosophy. But "groups" has been an inconsistent term for us. Sometimes it refers to latent social structures or COMMUNITIES, while at other times it means tight-knit subgroups, like the three of us here—what we might call a CLIQUE. We realized that in such cases, it might be better to model interactions for what they are: interactions involving more than just dyads. For example, when we cheer together, we think of it as a HYPEREDGE, or clique, as the phenomenology differs from each of us cheering individually with one another. The same applies to households, research collaborations, or group chats. Bartender, if I understand your question: 'When can we say that interactions in French cafés are HIGHER-ORDER, meaning they’re irreducible to just pairwise interactions?' One answer is that GROUP INTERACTIONS relate to the dynamics of the groups themselves. Take our example of cheering. When my friends start to cheer, am I going to join in only with a probability proportional to the number of people cheering? No, if we decide to cheer, that probability is highly NONLINEAR; as soon as one raises their glass, we all follow suit, as the Philosopher would say. We see irreducibility in group interactions with the interplay of structure and dynamics of networks; the probability that I raise my arm as others start cheering is more than the sum of INDEPENDENT INFLUENCES. With that in mind, we have developed very neat models that we can analyze with our mathematical toolsets. I guess one would wonder, then, what we would gain by adding this extra layer of intentionality?

This is an interesting question! Don’t you want to model what makes humans special? Collective intentionality has tangible influence that I bet can mislead your simpler models. This is not just philosophy. The science of group minds is expanding every day; there is social cognition talking about the "We-mode" of collaborating minds, or the emergence of collective intentionality in developmental psychology. Consider the work in developmental psychology by Michael Tomasello. As early as four years of age, children are highly motivated to enforce fairness in collaborative activities. In many studies, groups of children were asked to distribute rewards among themselves, varying the relative fairness of the distribution scheme. Researchers found that even at four, children react strongly to unfair distributions, rejecting rewards that disadvantage them. Consider that—they prefer to receive nothing rather than be subject to perceived unfairness! As they grow, children start making reference to the "right" ways of doing things, which extend beyond rules established in MOMENTARY INTERACTIONS. They effectively become agents motivated to uphold SUPRA-INDIVIDUAL CONSTRUCTIONS established by their group. This sense of fairness is a simple example of how groups govern our social lives. As children become moral agents, they learn that morality doesn’t depend on "You" and "I" doing the right things. They feel increasingly empowered by this more abstract "We", which we think of as the root of institutional facts. What happens when you ignore that fuzzy layer in your representation of human networks? Let’s consider what was happening before you embraced higher-order structures. You felt motivated to explain the conditions under which higher-order interactions better represent social systems than pairwise interactions. Your argument was that hyperedges are more accurate for representing us cheering because 'that’s what it is' (your words). Make a similar leap in thinking once again: why is being an intentional group more significant than a mindless cluster of interacting individuals? Because that’s what it is! By accepting what we call this NON-REDUCTIVE interpretation of groups, you are confronted with a whole new world of modeling choices. If you keep all your representations as mindless entities, you miss the key idea of individuals acting in response to what they think the 'group is thinking.' By distinguishing group minds from the sum of individual intentions, we emphasize the role of intentional groups in shaping our individual lives. Yes, the structure and dynamics of groups matter, but humans are unique in how we vary in our responses to group beliefs and norms, perhaps leading to qualitatively different group dynamics.

If I may—Tomasello, the guy we’re talking about, is a fellow cultural evolutionist! His idea of 'We-' intentionality is central to how humans build on cultural knowledge over generations—a process we call the RATCHET EFFECT. It’s the idea that our evolutionary success stems from constantly building on previous achievements. For human populations to exhibit cumulative cultural evolution, it needs to be transmitted accurately. Cultural evolutionists have shown how humans are unique in their ability to learn from each other, imitate, and teach, thereby increasing transmission accuracy.

Let me flesh out how cultural evolution connects with philosophy. We rarely discuss irreducibility but think a lot about GROUP-LEVEL TRAITS. As evolutionists, we see groups as carriers of cultural variations that are irreducible to INDIVIDUAL TRAITS. That is, a collection of individuals interacts together to produce a trait, rather than collective intentionality. For example, being a guitarist is an individual trait, while playing punk music as part of a band is a group-level trait. In both cases, we’re distinguishing groups by what they produce.

We also believe the emergence of group-level attributes has something to do with institutions, but with a little extra. We think that institutions often provide a framework for DIFFERENTIATED individuals to COORDINATE and COLLABORATE toward a shared purpose—I guess similar to the concept of CI. Consider again punk music, because why not. The shared purpose is producing music, but this arises from each individual playing their part. The institution includes how to produce punk music, but also political beliefs, ethics, and norms of addressing the audience that structure the ways in which group members interact. As such, non-reducibility arises as individuals take on distinct roles within a context of COLLABORATIVE INTERDEPENDENCE. Group-level traits are derived from norms and organization that make social groups more effective as a unit in ways that mere aggregates of individuals cannot.

This idea of group effectiveness hints at another key concept in cultural evolution missing from the Philosopher’s perspective: groups often compete in various ways, influencing their growth, production of daughter groups, and management of the commons. They compete directly but also indirectly, with differential group survival and reproduction. Since we assume that group-level variation has an impact on all that, some of us suggest that group-level traits may be subject to SELECTION. Although controversial, there are strong reasons to believe this process could help explain how humans scaled up collaboration among large groups of unrelated individuals. Without going into the details, it led us to study in depth the spread of GROUP-BENEFICIAL TRAITS, where individuals are willing to pay some personal cost for the group to succeed.

So, you see, we agree with the Philosopher that group interactions are not reducible to nonlinear dynamics in momentary higher-order interactions. For us, group interactions are also about that additional institutional layer that is the product of individuals interacting toward a shared goal. But we reach this conclusion from a very different place. The ratchet effect suggests that, rather than evolving big brains simply to be smart, we evolved them to learn strategically from each other. Cumulative cultural knowledge led to the idea that more structured organizations, with rules and norms, ought to stabilize cooperation in larger and more heterogeneous social networks.

You two are blowing my mind. I see that you're both saying related things, but not exactly. Let me get this straight. Philosopher, you're arguing that we need intentionality because it shapes group interactions, beyond momentary interactions. Intentionality might influence things like fairness, such as how groups distribute rewards according to social norms. These norms are non-reductive, partly because they rely on individuals' perceptions of what the 'group is thinking.'

Cultural Evolutionist, you also see norms as constitutive of social groups, but in the form of group-level traits. Group-level traits are non-reductive because they depend on individuals being differentiated and organized. Here, it’s not really about intentionality; rather, it’s about how groups can be seen as more or less successful based on traits that cannot be expressed at the individual level. These group-level traits cannot be expressed by individuals because they emerge from organized interactions between group members.

Am I getting this right? So let’s assume that groups are not mindless but purposeful, either because they can be more successful or they exhibit intentionality. How do I represent purposeful groups?

I agree; this is challenging. But first, I just wanted to mention something to cultural evolutionist. It seems we are on the same page about groups! A colleague of mine has reached a similar conclusion about the role of norms in characterizing the ontology of social groups. What she calls internal and external norms set expectations about how to interact with in- and out-group members, while structuring norms set expectations about roles and positions in the hierarchy. But again, for us, norms are interesting because of what they mean for group members; social groups exist because they provide normative structure in how WE live our lives together in a variety of scenarios. Going back to your punk music example, she would be interested in the function it might serve; how people use punk music to self-identify, or perhaps as a derogatory term against the youth. This reminds me of how John Robb' starts his Punk Rock: Oral History Anyway, in both cases, Physicist, we want to be able to distinguish situations where individuals respond to the norms of groups they identify with, which influence their behaviors and potentially feed back into the group’s actions. The Cultural Evolutionist was saying something similar with individuals having a psychology geared towards learning from each other. In my mind, there’s also an element of meta-cognition at work here. In your jargon, the states of both nodes and groups are influenced by an additional layer—whether it’s intentionality, beliefs, or perhaps norms of rationality and truth. Whatever it is, it’s certainly beyond the realm of mindless bouncing balls.

Alright, this is tricky, but I’m up for the challenge. What do we want to keep track of here? What can we observe? What are the units of interaction? The perception of a punk band by an audience could be seen as band-to-audience interactions; each audience member is influenced to talk to friends about punk music after seeing a show. In this case, the group—the band playing together—is interacting as a whole with individuals. But what you’re describing is more abstract; it involves the interaction between group-level traits or intentionality and what I’ve been calling group interactions—like those within a household. For simplicity, let’s refer to my concept of group interactions as MANY-BODY INTERACTIONS, as I sense something physical or tangible in momentary group dynamics that seems absent from group stuff.

You both mentioned the coevolution of group-level phenomena, like institutions, with individual behaviors; they are mutually reinforcing. Punk music shapes how individuals act and adopt certain traits, but, of course, people also influence punk music in return.

I think two mechanisms are essential to represent purposeful groups: (i) a system where groups display emergent, purposeful states, and (ii) a multiscale model where the ‘group mind’ both shapes and is shaped by individual states, capturing the meaning the group holds for its members. There’s also the interaction between node heterogeneity and the group mind, or social organization—whatever forms the ‘framework for cooperation and coordination.’ Let’s start by modeling the coevolution of individual and group minds, and then we can add the rest…