Decision makers frequently reject mixed gambles offering a larger possible gain and a smaller possible loss, the classic phenomenon of loss aversion. Zhao, Walasek and Bhatia (2020) showed with a drift-diffusion model (DDM) of accept/reject decisions that this rejection tendency decomposes into three distinct mechanisms: the prospect-theory utility-weighting bias, a fixed utility bias, and a pre-valuation bias — a starting-point bias of the evidence-accumulation process that favors rejection before the gamble is even valued. The pre-valuation bias proved to be the largest single contributor, adapts to the desirability of the experienced gamble environment, and appears to be carried into the laboratory from pre-experimental tendencies. The open problem, posed by the source paper itself, is to explain where this bias comes from and why it resists shifting toward acceptance even in highly desirable environments; to build formal models that connect risky choice to long-term decision experience in everyday environments and give a detailed account of the contextual determinants of loss aversion (including the status quo and the sure-zero framing); and to determine which psychological, clinical, and neurobiological correlates of mixed-gamble rejection track the pre-valuation bias rather than valuation mechanisms.

Background

Loss aversion — the observation that losses loom larger than equivalent gains — is a cornerstone of behavioral decision research. In prospect theory (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992), it is modeled as a utility-weighting asymmetry: the subjective value function for losses is steeper than for gains, captured by a coefficient λ>1\lambda>1 that overweights losses relative to gains. A standard laboratory probe is the mixed-gamble accept/reject task, in which participants decide whether to accept gambles offering equal chances of a gain and a loss; higher rejection rates of gambles with larger gains than losses are conventionally interpreted as evidence of a large λ\lambda. This interpretation underlies a large correlational literature, including the influential finding that striatal and prefrontal activity tracks mixed-gamble rejection rates (Tom, Fox, Trepel & Poldrack, 2007), where neural activity was attributed to utility weighting.

That inference has been questioned on several grounds. Ert and Erev (2008) showed that rejection of attractive mixed gambles diminishes when the task is reframed as a choice between the gamble and a sure outcome of zero rather than an accept/reject decision with a status quo of doing nothing, suggesting that rejection partly reflects the response format rather than valuation. Walasek and Stewart (2015) showed that loss aversion can be made to disappear or reverse by manipulating the distribution of gains and losses participants are exposed to, supporting environment-based (decision-by-sampling) origins of the phenomenon. Process-level modeling sharpened these concerns: Clay, Clithero, Harris and Reed (2017) fitted drift-diffusion models to mixed-gamble decisions and argued that loss aversion reflects the rate of evidence accumulation rather than a response bias.

The drift-diffusion model (Ratcliff & McKoon, 2008) describes two-choice decisions as noisy accumulation of evidence toward one of two decision boundaries; its central parameters are the drift rate (the average speed and direction of evidence accumulation, governed here by the utilities of the gain and the loss) and the starting point (how close the accumulation process begins to one boundary before any evidence is seen, which biases response times and choices toward the nearer boundary). Zhao, Walasek and Bhatia (2020) used this framework to decompose mixed-gamble rejection into three simultaneously estimated mechanisms: a utility-weighting bias λ=β^L/β^G\lambda=\hat{\beta}_L/\hat{\beta}_G operating through the drift rate, a fixed utility bias α\alpha (an additive preference for rejection inside valuation), and a pre-valuation bias γ\gamma implemented as a starting-point shift toward the rejection boundary, with the standard logistic choice model recovered as the special case γ=0\gamma=0.

In two preregistered experiments, Zhao et al. (2020) found that removing the pre-valuation bias hurt model fit more than removing either valuation mechanism in Experiment 1 and in the low-payoff condition of Experiment 2 (the paper notes that the relative importance of the pre-valuation bias and utility weighting bias varies across the high- and low-payoff conditions), that the bias carries a unique behavioral signature — faster rejections than acceptances once choice difficulty is controlled — and that it correlates with individual acceptance rates more strongly than utility weighting does. Experiment 2 showed that γ\gamma is adaptive: participants exposed to a low-payoff gamble environment developed a more negative pre-valuation bias than those exposed to a high-payoff environment, while for gambles shared across the two environments the two mechanisms dissociated (relative comparisons shifted the drift rate, prior experience shifted the starting point). The paper's conclusions were shown to generalize to a model incorporating risk aversion, with adequate parameter recovery in supplementary analyses. Critically, the bias is present from the very first trials of the experiment, indicating that participants bring a rejection-favoring starting point into the laboratory; and even in the highly attractive high-payoff environment participants retained a mild tendency to reject — an asymmetry in adaptation that the paper documents but does not explain. Related dynamic approaches, such as mouse-tracking measures of decision conflict that predict loss-aversion parameters (Stillman, Krajbich & Ferguson, 2020), likewise measure the tendency without identifying its source.

What remains unresolved is the mechanistic account of the pre-valuation bias itself: what establishes it (long-term experience in everyday decision environments, pre-experimental response tendencies, or other learned expectations), why it adapts asymmetrically, how contextual determinants such as the status quo and the sure-zero framing modulate it, and which of the many psychological, clinical, and neurobiological variables previously correlated with mixed-gamble rejection actually track this pre-valuation channel rather than valuation.

Problem Statement

The problem, as posed by Zhao, Walasek and Bhatia (2020) at the close of their drift-diffusion decomposition of loss aversion, is to give a mechanistic account of the pre-valuation (starting-point) bias that their experiments identify as the largest single mechanism behind mixed-gamble rejection. Its unified objective is to explain where this bias comes from and how context shapes it, jointly addressing three linked questions the source paper explicitly leaves open:

  1. Origin and asymmetric adaptation. What establishes the rejection-favoring pre-valuation bias — long-term experience in everyday decision environments, pre-experimental response tendencies, or other learned expectation mechanisms — and why are decision makers reluctant to shift toward positive (acceptance-favoring) pre-valuation biases even when immersed in highly desirable gamble environments, retaining a mild rejection tendency despite large positive expected values?

  2. Formal account of contextual determinants. Develop formal models that connect risky choice to long-term decision experience in everyday environments, specifying how that experience maps onto the starting-point and valuation parameters of the decision process, and thereby give a detailed account of the contextual determinants of loss aversion — including the role of the status quo in accept/reject tasks and of external frames such as the choice-between-gamble-and-sure-zero setting in which loss aversion diminishes.

  3. Correlate mapping. Determine which of the psychological, clinical, and neurobiological variables previously correlated with mixed-gamble rejection rates track the pre-valuation bias rather than the valuation mechanisms (for example, whether striatal and prefrontal correlates of the kind reported by Tom et al., 2007, reflect utility weighting or pre-valuation tendencies).

All three numbered questions are mandatory and none is illustrative, since these components jointly support the single objective of explaining the pre-valuation bias. The boundary of the problem is a source-faithful mechanistic account validated against the multi-mechanism decomposition that (1) explains what establishes the pre-valuation bias and why it adapts asymmetrically, (2) develops the formal models connecting risky choice to long-term decision experience and thereby the contextual determinants of loss aversion, and (3) assigns the psychological, clinical, and neurobiological correlates of mixed-gamble rejection to the pre-valuation or the valuation mechanisms.

References

  1. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291. DOI: 10.2307/1914185. https://doi.org/10.2307/1914185
  2. Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323. DOI: 10.1007/BF00122574. https://doi.org/10.1007/BF00122574
  3. Tom, S. M., Fox, C. R., Trepel, C., & Poldrack, R. A. (2007). The neural basis of loss aversion in decision-making under risk. Science, 315(5811), 515–518. DOI: 10.1126/science.1134239. https://doi.org/10.1126/science.1134239
  4. Ert, E., & Erev, I. (2008). The rejection of attractive gambles, loss aversion, and the lemon avoidance heuristic. Journal of Economic Psychology, 29(5), 715–723. DOI: 10.1016/j.joep.2007.06.003. https://doi.org/10.1016/j.joep.2007.06.003
  5. Walasek, L., & Stewart, N. (2015). How to make loss aversion disappear and reverse: Tests of the decision by sampling origin of loss aversion. Journal of Experimental Psychology: General, 144(1), 7–11. DOI: 10.1037/xge0000039. https://doi.org/10.1037/xge0000039
  6. Clay, S. N., Clithero, J. A., Harris, A. M., & Reed, C. L. (2017). Loss aversion reflects information accumulation, not bias: A drift-diffusion model study. Frontiers in Psychology, 8, 1708. DOI: 10.3389/fpsyg.2017.01708. https://doi.org/10.3389/fpsyg.2017.01708
  7. Ratcliff, R., & McKoon, G. (2008). The diffusion decision model: Theory and data for two-choice decision tasks. Neural Computation, 20(4), 873–922. DOI: 10.1162/neco.2008.12-06-420. https://doi.org/10.1162/neco.2008.12-06-420
  8. Zhao, W. J., Walasek, L., & Bhatia, S. (2020). Psychological mechanisms of loss aversion: A drift-diffusion decomposition. Cognitive Psychology, 123, 101331. DOI: 10.1016/j.cogpsych.2020.101331. https://doi.org/10.1016/j.cogpsych.2020.101331
  9. Sheng, F., Ramakrishnan, A., Seok, D., Zhao, W. J., Thelaus, S., Cen, P., & Platt, M. L. (2020). Decomposing loss aversion from gaze allocation and pupil dilation. Proceedings of the National Academy of Sciences, 117(21), 11356–11363. DOI: 10.1073/pnas.1919670117. https://doi.org/10.1073/pnas.1919670117
  10. Stillman, P. E., Krajbich, I., & Ferguson, M. J. (2020). Using dynamic monitoring of choices to predict and understand risk preferences. Proceedings of the National Academy of Sciences, 117(50), 31738–31747. DOI: 10.1073/pnas.2010056117. https://doi.org/10.1073/pnas.2010056117
  11. Walasek, L., & Stewart, N. (2021). You cannot accurately estimate an individual's loss aversion using an accept–reject task. Decision, 8(1), 2–15. DOI: 10.1037/dec0000141. https://doi.org/10.1037/dec0000141
  12. Mulder, M. J., Prummer, F., Terburg, D., & Kenemans, J. L. (2023). Drift–diffusion modeling reveals that masked faces are preconceived as unfriendly. Scientific Reports, 13, 16982. DOI: 10.1038/s41598-023-44162-y. https://doi.org/10.1038/s41598-023-44162-y
  13. He, X., Qiu, B., Deng, Y., Wang, Z., Cao, X., Zheng, X., Zhu, J., & Zhang, W. (2024). Material hardship predicts response bias in loss-averse decisions: The roles of anxiety and cognitive control. The Journal of Psychology, 158(4), 309–324. DOI: 10.1080/00223980.2023.2296946. https://doi.org/10.1080/00223980.2023.2296946
  14. Chen, Z., Eben, C., & Verbruggen, F. (2024). Differential effects of prior outcomes and pauses on the speed and quality of risky choices. Journal of Behavioral Decision Making, 37(2), e2370. DOI: 10.1002/bdm.2370. https://doi.org/10.1002/bdm.2370
  15. Walasek, L., Mullett, T. L., & Stewart, N. (2024). A meta-analysis of loss aversion in risky contexts. Journal of Economic Psychology, 103, 102740. DOI: 10.1016/j.joep.2024.102740. https://doi.org/10.1016/j.joep.2024.102740
  16. Cotet, M., Zhao, W. J., & Krajbich, I. (2025). Deliberation during online bargaining reveals strategic information. Proceedings of the National Academy of Sciences, 122(7), e2410956122. DOI: 10.1073/pnas.2410956122. https://doi.org/10.1073/pnas.2410956122
  17. Chen, C., Lin, X., Yang, Y., & Huang, J. (2026). Disentangling computational and neural mechanisms of evaluation direction in value-based decision-making. NeuroImage, 333, 121945. DOI: 10.1016/j.neuroimage.2026.121945. https://doi.org/10.1016/j.neuroimage.2026.121945