The Social Self

Five posts into this series, I need to acknowledge something I’ve been getting wrong: I’ve been treating decision-making as if it happens inside individual minds.

A person weighs evidence, calibrates confidence, manages uncertainty, chooses actions. Even when I discussed irrationality, I framed it as internal struggle within a single self.

But this misses something fundamental: we make decisions our “rational selves” disagree with constantly, not because we’re confused, but because we understand we live in a society, and social context influences what we do more than what we think.

The Dinner Invitation

#

You’re exhausted. Your friend invites you to dinner. You don’t want to go. Your rational self says: decline, rest, optimize for wellbeing.

But you go anyway. Not because you suddenly want to. Because the relationship matters more than your preference. Because this is what friends do. Because saying no creates awkwardness you’ll regret more than the fatigue.

This isn’t weakness, it’s sophisticated social reasoning overriding individual optimization.

Aristotle knew this: humans are political animals, constituted by our relationships. Hegel developed it: the self emerges through recognition by others. Heidegger made it central: we are always already Being-with. Margaret Gilbert formalized it: much of social life involves “plural subjects” where the “we” acts, not just aggregated “I"s.

The philosophical insight: The isolated individual who makes autonomous rational choices is a fiction. We’re fundamentally social beings whose decisions emerge from relationships.

Three Levels of Social Decision-Making

#

Level 1: Social preferences (still individual). You care about others’ happiness. You factor their wellbeing into your utility function. But you’re still the decision-maker, just with expanded preferences. This is the level most AI systems can model.

Level 2: Role-dependent (social norms). Your professional obligations override personal preference. You don’t want to work late, but your role requires it. The role decides, not you. Social norms determine behavior independently of individual preferences.

Level 3: Relationally constituted (deepest). The “we” decides, not the “I.” You’re literally different selves in different relationships. Margaret-with-daughter isn’t Margaret plus context. It’s a different decision-making entity. The relationship itself has preferences, not just the individuals in it.

Seven Ways Social Reality Overrides Individual Rationality

#

Face-saving and honor. Erving Goffman showed how we perform selves in social situations, maintaining “face” even at significant personal cost. Reputation management trumps preference satisfaction.

Gift-giving and reciprocity. Marcel Mauss demonstrated that gifts create obligations that bind societies together. You give and receive not for utility but for relationship.

Conformity and belonging. Solomon Asch’s experiments showed people denying their own perceptions to match group consensus. Belonging overrides accuracy.

Moral disgust and taboo. Jonathan Haidt’s work shows moral reasoning is often post-hoc rationalization of disgust reactions. We feel first, reason second.

Loyalty overriding judgment. You defend your group’s bad decisions because loyalty matters. The relationship trumps the rational assessment.

Politeness overriding truth. Paul Grice’s maxims show how conversation requires cooperative violation of pure information transfer. We sacrifice accuracy for relationship.

Self-sacrifice for group. The ultimate override: dying for others. No individual utility function captures this.

The Fundamental Shift

#

The isolated mind model: Individual → Processes info → Forms preferences → Decides → Acts in social world.

The embedded mind model: Social context → Constitutes self → Self-in-relation → Decisions emerge from relational space.

Consider a jazz musician. You can’t isolate one musician’s “contribution”, the music emerges from interaction. Similarly, human decisions emerge from social contexts, not just individual minds.

What This Means for AI

#

AI can model multiple preference contexts, social norms and roles, and relationship patterns. But AI cannot model the relational constitution of self, Margaret-with-daughter isn’t Margaret plus context, it’s a different entity. AI cannot access social meaning and face, the felt pull of obligation, or emergent collective decision-making where the “we” is the decision-maker.

The challenge to rationality: If we constantly act “against our interests” for social reasons, are we irrational?

My answer: Rationality is context-dependent and socially embedded. We navigate among multiple rationalities (individual, social, collective, moral). Social rationality usually wins because we’re fundamentally social.

Implications for AI Design

#

If human understanding is socially constituted, then AI that models only individuals will systematically fail. We need:

Relationship-aware modeling. Track not just Margaret’s preferences, but Margaret-with-daughter, Margaret-with-physician, Margaret-alone preferences.

Social context detection. Recognize which relational mode is active and adjust predictions accordingly.

Respect for distributed agency. Sometimes the right “decision-maker” to consult isn’t the individual but the relationship or family unit.

Humility about individual models. Accept that individual preference models will always be incomplete because individuals are always embedded.

The individual is a mode we can enter, not the human baseline. AI that treats humans as isolated decision-makers will fail to approximate human understanding in its most fundamental dimensions.

This is the sixth in a series exploring how AI approaches understanding. Previous articles examined individual cognition. This one reframes everything by showing that even the “individual” making calibrated uncertain decisions is itself a social construct.