Act or Clarify? Modeling Sensitivity to Uncertainty and Cost in Communication

When deciding how to act under uncertainty, agents may choose to act to reduce uncertainty or they may act despite that uncertainty. In communicative settings, an important way of reducing uncertainty is by asking clarification questions (CQs). We predict that the decision to ask a CQ depends on both contextual uncertainty and the cost of alternative actions, and that these factors interact: uncertainty should matter most when acting incorrectly is costly. We formalize this interaction in a computational model based on expected regret: how much an agent stands to lose by acting now rather than with full information. We test these predictions in two experiments, one examining purely linguistic responses to questions and another extending to choices between clarification and non-linguistic action. Taken together, our results suggest a rational tradeoff: humans tend to seek clarification proportional to the risk of substantial loss when acting under uncertainty.

💡 Research Summary

The paper investigates how people decide whether to act immediately under uncertainty or to seek clarification through a question before acting. The authors formalize this decision as a trade‑off between the expected regret of acting with incomplete information and the cost of asking a clarification question (CQ). Expected regret is defined as the difference between the payoff that would be obtained with perfect information and the expected payoff given the current level of knowledge; it captures the potential loss from a suboptimal action. The central hypothesis is that both the level of uncertainty about the interlocutor’s goal and the cost of acting incorrectly jointly determine the likelihood of asking a CQ, with an interaction such that uncertainty matters most when the cost of a mistake is high.

Two experiments were conducted.

Experiment 1 examined a pure linguistic setting. Participants played the role of a bartender responding to a customer’s “What drinks do you have?” query. Four response options were available: (1) give an exhaustive list of all drinks, (2) give a partial list (mention‑some), (3) ask a clarification question (“Do you prefer cocktails or soft drinks?”), or (4) other. Uncertainty was manipulated by a cover story indicating whether the customer was equally likely to prefer cocktails or soft drinks (high uncertainty) or strongly biased toward one category (low uncertainty). Cost was operationalized by the size of the option space: a small space (two drinks per category) versus a large space (four drinks per category). The design was a 2 × 2 within‑subjects factorial. Bayesian logistic regression revealed a strong main effect of uncertainty (β = 0.34, 95 % CrI