Can 5‐month‐old infants consider the perspective of a novel eyeless agent? New evidence for early mentalistic reasoning

Abstract

To make sense of an agent's actions in a scene, we generally consider what information is available to the agent. When this information is less complete than our own, so that the agent is ignorant about critical aspects of the scene, we adopt the agent's perspective to predict, interpret, and respond to the agent's actions. What are the infant roots of this perspective-taking ability? Are infants in the first year of life (henceforth young infants) already able to attribute to an ignorant agent a representation of a scene that differs from their own? When researchers began studying epistemic reasoning in infancy, two types of accounts were proposed that offered different answers to this question (for a recent review, see Scott et al., in press).According to mentalistic accounts, young infants are capable of tracking what information is available to ignorant agents. Mentalistic accounts differ in how they describe these epistemic attributions, from simple registrations dependent on a minimal psychologicalreasoning system (two-system accounts; Butterfill & Apperly, 2013;Low et al., 2016;Low & Watts, 2013) to more sophisticated mental states dependent on a psychological-reasoning system akin to that of older children and adults (one-system accounts; Baillargeon et al., 2016;Carruthers, 2013;Leslie et al., 2004;Luo & Baillargeon, 2010). Despite these differences, however, mentalistic accounts agree that young infants possess rudimentary perspective-taking skills: Under some conditions at least, infants are able to recognize that an agent's representation of a scene is less complete than their own, and they then use this incomplete representation, nonegocentrically, to reason about the agent's actions.In contrast, teleological accounts argue that young infants lack any perspective-taking skills. In this view, the cognitive system responsible for reasoning about agents' actions is initially limited to processing physical variables, as opposed to mental states (Csibra et al., 2003;. When watching an agent act in a scene, infants generate a teleological action explanation that specifies the layout of the scene, the agent's actions, and the end-state the agent achieves. This explanation, together with a core principle of rationality (and

Understanding preferences in infancy

“…This comparison helps to highlight the distinction between preferences and goals or intentions: whereas goals or intentions denote outcomes agents seek to achieve (e.g., to grasp an object), preferences are dispositional states that explain why agents choose one option over another. Critically, if object‐B is initially hidden from the agent, but not the infant, infants still appear to construe the situation as the agent having no choice because object‐A is the agent's only option and hence do not attribute a preference to the agent (Choi et al, 2022; Luo & Baillargeon, 2007; Luo & Johnson, 2009). This set of results underscores the non‐egocentric feature of infant preference understanding.…”

“…Further, when both choice and effort information is given in an experimental situation (e.g., as an agent chooses between two options, A and B, with A easier to reach than B), infants seem to prioritize the agent's choice over effort to make decisions on the agent's preferences. This is because around 5–6 months of age, infants were still found to attribute to an agent a preference for A if the agent consistently chose A over B when both objects were of equal distance to the agent, but B was separated by a short barrier from the agent, and hence, hypothetically less accessible and a more‐effortful option than the directly reachable A (Choi et al, 2022; Luo & Johnson, 2009).…”

“…While both age groups acted prosocially because they gave the experimenter something, the older age group's actions were more precise and tailored to the experimenter's specific preferences. This developmental improvement to take notice of others' individual preferences in infants' overt actions, above and beyond their implicit understanding of others' representations from an early age as shown by their looking times (e.g., Choi et al, 2022; Luo & Baillargeon, 2007), can undoubtedly facilitate infant participation in social interactions, which requires considerations of others' different perspectives.…”

The Violation-of-Expectation Paradigm: A Conceptual Overview

“…When we introduced Woodward's (1998) preference task earlier in this article, we noted that its findings were open to a novelty-based interpretation: In the habituation trials, infants attended mainly to the object grasped by the agent (object-A); in the test trials, they dishabituated when this object changed from familiar object-A (expected event) to novel object-B (unexpected event). However, recall that additional data refuted this novelty-based interpretation: Infants looked about equally at the test events if in the habituation event object-B was either absent or hidden from the agent by a screen (Bíró et al, 2011;Choi et al, 2022;Luo & Baillargeon, 2005a;Luo & Johnson, 2009).…”

From infancy, we model people as minds, acting in a physical world

“…compared to when a hand comes to rest passively on an object; Woodward, 1999). Infants also use what agents can see to constrain their interpretation of goal-directed action-they posit an object as a potential goal to an agent only if the agent was in a position to know about it (Choi et al, 2022;Luo & Johnson, 2009).…”