Biases about Sunk Costs: Children Do Not Anticipate Irrationality

• In Event: Poster Session 04
  In Poster Session: PS 04 Topic - Cognitive Processes

Abstract

Children expect agents to behave rationally, considering constraints, costs, and rewards (e.g. Gergely & Csibra, 2003; Jara-Ettinger et al., 2015). For example, infants expect agents to move towards goals via the least costly path (Liu & Spelke, 2017). However, biases lead people to make irrational choices (Arkes & Ayton, 1999; Thaler, 1974). One instance is the sunk cost bias, when people overvalue things that they spent more time, money, or effort pursuing. For example, if someone booked two non-refundable trips for the same day, they would likely go on the more expensive trip over a cheaper one, regardless of which one will be more enjoyable (Arkes & Blummer, 1985). Accurately predicting others’ behaviour, then, requires not only anticipating rational choices, but also anticipating irrational ones. In three experiments, we explored if children and adults predict others’ irrational choices about sunk costs.
In Experiment 1, 60 5-6-year-olds and 153 adults heard stories about an agent who collected two identical objects, one easy to obtain and one difficult (e.g. a girl collecting flowers atop a small and a large hill). The agent then learned that they could only keep one object, and participants indicated which object the agent would keep. Choosing the high-effort object would indicate that participants expected the agent to value it more than the low-effort object. Adults were more likely than children to choose the high-effort object (p<.001), and they chose it significantly above chance (p<.001). However, children chose objects at similar rates (p=.315).
Since children did not expect an agent to irrationally overvalue sunk costs, Experiment 2 explored if children expect others to rationally minimize future costs. Six-year-olds (N=24) were either shown stories about sunk costs (similar to Experiment 1) or future costs. In the future cost condition the agent retrieved the two identical objects but instead of asking which they would keep, two new objects appeared in the same locations and children predicted which the agent would retrieve. There was no significant difference between conditions (p=.127), and children in both conditions chose the low- over high-effort objects (sunk cost p=.029; future cost p=.006). Thus, children expect agents to rationally minimize future costs, but not irrationally overvalue sunk costs.
Experiment 3 explored children’s sensitivity to others’ sunk costs by asking them to make a first-person judgment, as we anticipated that this might be easier. We presented 34 6-year-olds with two agents who each collected identical objects to give to the participants themselves. Children were asked which gift they would choose; the object from the agent who expended more effort or the object from the agent who expended less effort. Children chose objects at similar rates (p=.069), suggesting that they do not overvalue sunk costs (and if anything, children leaned towards choosing the low-cost object).
Together these findings reveal that children do not anticipate others’ irrational choices about sunk costs, nor do they consider others’ sunk costs in their own choices. When considering costs leads to rational behaviour (e.g., minimizing future costs), children can accurately predict behaviour.

Authors

• Claudia Sehl, University of Waterloo
  Presenting Author

• Stephanie Denison, University of Waterloo
  Non-Presenting Author

• Ori Friedman, University of Waterloo
  Non-Presenting Author