The Naïve Utility Calculus: Computational Principles Underlying Commonsense Psychology

Abstract

We propose that human social cognition is structured around a basic understanding of ourselves and others as intuitive utility maximizers: From a young age, humans implicitly assume that agents choose goals and actions to maximize the rewards they expect to obtain relative to the costs they expect to incur. This "naïve utility calculus" lets both children and adults observe others' behavior and infer their beliefs and desires, their longer-term knowledge and preferences, and even their character: who is knowledgeable or competent, who is praiseworthy or blameworthy, who is friendly, indifferent or an enemy. We review studies providing support for the naïve utility calculus, and we show how it captures much of the rich social reasoning humans engage in from infancy. Commonsense PsychologyTheories of decision-making have been at the heart of psychology since the field's inception, but only recently has the field turned to the study of how humans -especially the youngest humansthink humans make decisions. When we watch someone make a choice, we explain it in terms of their goals, preferences, personalities, and moral beliefs. This capacity -our commonsense psychology -is the cognitive foundation of human society. It lets us share what we have and know, with those from whom we expect the same in return, and it guides how we evaluate those who deviate from our expectations.The representations and inferential power underlying commonsense psychology trace back to early childhood -before children begin kindergarten, and often even in infancy. Work on how children reason about other agents' goals [1][2][3][4][5][6][7][8], desires [9][10][11], beliefs [12][13][14][15][16][17][18], and pro-social behavior [19][20][21][22][23][24][25][26][27][28][29] has advanced our understanding of what in our commonsense psychology is at work in early infancy [30][31][32] and what develops [16][17][33][34][35]. Nonetheless, major theoretical questions remain unresolved. What computations underlie our commonsense psychology, and to what extent are they specific to the social domain? Are there a small number of general principles by which humans reason about and evaluate other agents, or do we instead learn a large number of special case rules and heuristics? To what extent is there continuity between the computations supporting commonsense psychology in infancy and later ages? Is children's social-cognitive development a progressive refinement of a computational system in place from birth, or are there fundamentally new computational principles coming into play?In this article we advance a hypothesis that offers answers to each of these questions, and provides a unifying framework in which to understand the diverse social-cognitive capacities we see even in young children. We propose that human beings, from early infancy, interpret others' intentional actions through the lens of a naïve utility calculus: that is, people assume that others choose actions to maximize utilities -the rewards they expect to obtain relative to the costs...

Learning the Structure of Social Influence

“…The flexibility of children's inferences suggests that they may be supported by an abstract understanding of preferences (i.e., utility functions that underlie choices) rather than mere associations (Jara-Ettinger, Gweon, Schulz, & Tenenbaum, 2016). Around 18 months of age, children already understand that another agent's food preference may differ from their own (Repacholi & Gopnik, 1997) and provide food items that she likes even when it conflicts with their own preferences.…”

Children understand that agents maximize expected utilities.

“…For us to know what someone likes, she has to know it herself first. other people make decisions, and it may differ in important ways from how people actually make decisions (see Jara-Ettinger, Gweon, Schulz, & Tenenbaum, 2016). Under this account, agents' knowledge of the costs and rewards should influence their behavior in two ways.…”

Listener’s vmPFC simulates speaker choices when reading between the lines

“…This generative account, also known as the Rational Speech Act (RSA) model (10), provides precise and falsifiable behavioral predictions in a parsimonious framework for social signal 55 interpretation that can be extended from communication (10,12,13) to social perception (16,17) and interpersonal decision-making (18). However, no direct evidence is available suggesting that pragmatic interpretation indeed involves rational, context-specific simulation of speakers.…”

“…Speakers are expected to compare candidate expressions to make a choice for best helping the audience recognize the intended meaning in a given context. In addition, listeners need to monitor 50 knowledge and beliefs shared with the speaker (common ground) for simulating speaker behavior based on mutual, rather than the listener's own private knowledge (4, 6).This generative account, also known as the Rational Speech Act (RSA) model (10), provides precise and falsifiable behavioral predictions in a parsimonious framework for social signal 55 interpretation that can be extended from communication (10,12,13) to social perception (16,17) and interpersonal decision-making (18). However, no direct evidence is available suggesting that pragmatic interpretation indeed involves rational, context-specific simulation of speakers.…”

“…In addition, it raises an intriguing possibility that additional neurocognitive processes are involved in selecting mental models appropriate to the information structure within communicators, perhaps through hierarchical generative mechanisms (11)(12)(13). 275 The notion of probabilistic generative model has deep roots in artificial intelligence (14) and cognitive neuroscience (11) such as visual perception (15) and motor control (31), and has been recently proposed to account for various social inferences (10,12,13,16,17). Our results shed light for the neural instantiation of this generative account in high-level cognitive systems involved in communication and social reasoning.…”