A note on “When are decisions improvable? An evaluation of diagnostic methods”

In their study of various improvability diagnostics, Bernheim, Lucia, Nielsen, and Sprenger (2026) argue that experimental subjects often interpret the elicitation of cognitive uncertainty (CU) - which asks them to report the likelihood that they made the best decision for themselves - as instead asking about the presence of ex-post outcome (payoff) uncertainty. First, we amend Bernheim et al.’s survey to disambiguate the interpretation of their results. Subjects may either misconstrue the CU elicitation as asking about outcome uncertainty, or they may report uncertainty about their best decision—in line with the intended meaning of the question - because outcome uncertainty makes the choice subjectively difficult. Our results support the latter interpretation. This conclusion is corroborated by both a literature review and an analysis of large-scale data on choice inconsistencies. Second, we study whether responses to the CU elicitation are predictive of choices in line with Bernheim et al.’s interpretation. In a standard multiple price list experiment, we find that, in the sub-sample that - according to Bernheim et al.’s methodology - understands the elicitation as asking about outcome uncertainty, CU predicts probability weighting (behavioral attenuation) just as strongly as in the rest of the sample. This is a pattern implied by the intended meaning of the CU question but not by Bernheim et al.’s interpretation.