Research

Working papers

  1. Confounds Protocols: Collecting & Analyzing Participant Explanations to Identify Confounds (and Sometimes Mechanisms)Previous title: “Don’t Ask Don’t Learn”Daniel BankiUri SimonsohnConditionally accepted at Psychological SciencePDFShow abstractHide abstract

    Abstract

    To protect against unexpected confounds, we propose collecting participant explanations in psychology experiments by default. We introduce the "Confounds Protocols", a five-step procedure for collecting and analyzing such data. We demonstrate their value by replicating three well-known studies that had already sparked published debate. In each example, participant explanations allowed us to discover something new and fundamental, be it confounds or mechanisms, about these mature paradigms.

  2. The Gambler's Fallacy Fallacy FallacyDaniel BankiUri SimonsohnAccepted at Psychological SciencePDFShow abstractHide abstract

    Abstract

    In contrast to the results and conclusions by Xiang et al. (2025), reanalyzing their data we find significant and substantive evidence of gambler's fallacy beliefs. Our results differ from theirs because our analytical approach differs from theirs. While they average probability estimates across judgments, we count probability estimates exhibiting the gambler's fallacy. For example, after a streak of length four, 57% of responses deem a streak as overly likely to end, compared to 31% that do the opposite. Moreover, Xiang et al. report that the median participant doesn't exhibit the gambler's fallacy, while we find that a substantial minority of participants do exhibit it.

  3. Decisions under Risk Are Decisions under Complexity: CommentDaniel BankiUri SimonsohnRobert WalatkaGeorge WuR&R at American Economic ReviewSSRNShow abstractHide abstract

    Abstract

    Oprea argued that prospect theory's risk attitudes reflect "complexity" rather than underlying risk preferences. A re-analysis reveals that this claim is driven by measurement error caused by a confusing experimental design. Participants valued risky lotteries and riskless "mirrors" similarly, but this was largely due to the 75% who failed comprehension checks. These subjects had noisy valuations and frequent dominance violations, which artifactually produce a fourfold pattern, loss aversion, and positive correlation between mirrors and lotteries. In contrast, participants who likely understood the task valued mirrors at expected value and lotteries in line with prospect theory.

Publications

  1. Justification aversion: The Road to More Effective DefaultsDaniel BankiDaniel Navarro-MartinezOrganizational Behavior and Human Decision Processes, 2026PDFDOIShow abstractHide abstract

    Abstract

    Default options are one of the most frequently used tools of choice architecture, yet the factors that influence their effectiveness remain imperfectly understood. As a way to improve the effectiveness of defaults, this paper explores the idea of adding a justification requirement if people want to switch away from the default. Across six pre-registered studies, we find a consistent justification aversion effect: Depending on the context, the requirement to justify one's choice results in a 22% to 58% increase in default adherence relative to comparable non-justification requirements. This effect is robust to different decision contexts and alternative switching requirements. We test various potential mechanisms, and the evidence is most consistent with image concerns associated with having to justify a socially undesirable deviation from the default. Importantly, image concerns do not appear to undermine socially desirable deviations from defaults.