The Price of Knowledge: Children Infer Epistemic Desires and States from Exploratory Behavior

• In Event: Diverse Knowledge, Diverse Cues: Implications for Learners’ Inferences and Actions

Abstract

To decide whom to believe, how to inform, and even whether to punish or forgive, it’s critical to understand what others know. Because we cannot see into others’ minds, we must often infer their epistemic states from their behavior. But while much prior research has investigated how children infer others’ goals, intentions and desires from their actions, far less work has tested whether and how children infer others’ epistemic states.
In the current work, we test whether four- and five-year-olds make rich epistemic inferences from others’ actions, and begin to investigate the mechanisms underlying their abilities. Specifically, we test whether the expectation that others maximize rewards and minimize costs — shown to underlie children’s goal, intention, and desire inferences — also explains their epistemic inferences.

In Experiment 1, four- and five-year-olds (n=48) were introduced to a strong puppet, a weak puppet, and a box. The experimenter told the puppets that there was a special toy hidden under the box. She gave each puppet a chance to lift the box, but both puppets refused. The experimenter then told children that one of the puppets had actually seen her hide the toy under the box, and asked them to judge which puppet had already known what was there. If children look to agents’ costs to infer their epistemic states, they should judge that the strong agent already knew: it would have been easy for her to lift the box, so if she chose not to, she probably didn’t need the information. Indeed, 75% of children judged that the strong puppet had already known what was under this box (36 of 48, bootstrapped 95% CI: 62.5 – 87.5; see Figure 1).

Experiment 2 was nearly identical, except that both puppets agreed to lift the box. Four- and five-year-olds (n=48) were now asked which puppet had really wanted to know what was there. If children attend to agents’ costs, they should say that the weak agent had the stronger epistemic desire, because it was a lot harder for her to lift the box. Indeed, four- and five-year-olds thought the weak puppet was the one who had really wanted to know (32 of 48, bootstrapped 95% CI: 54.2 – 79.2; see Figure 1).

Finally, in two control experiments (n=48) we show that these inferences cannot be explained by a simple association between strength and competence (see Figure 1).

Taken together, these results demonstrate that from others’ information-seeking actions and their costs, preschoolers can infer both what they knew and wanted to know. This suggests that by the preschool years, children’s mental-state inferences are already unified under an assumption that others maximize utilities, choosing actions that produce the greatest rewards while incurring the fewest costs — thus highlighting the role of mental state reasoning in children’s knowledge attributions.

Authors

• Rosie Aboody, Harvard University
  Presenting Author

• Caiqin Zhou, University of Michigan
  Non-Presenting Author

• Julian Jara-Ettinger, Yale University
  Non-Presenting Author