About Me

We present a new behavioral foundation for regret theory. The central axiom — trade-off consistency — renders regret theory observable at the individual level and makes our foundation consistent with the existing measurement method. For the first time, our behavioral foundation allows deriving a continuous regret theory representation and separating utility from regret. The axioms clarify that regret theory minimally deviates from expected utility by relaxing transitivity only.

We examine risk attitudes under regret theory and derive analytical expressions for two components — the resolution and regret premiums — of the risk premium. We posit that regret-averse decision makers are risk seeking for low probabilities of gains and risk averse for high probabilities, and that feedback reinforces risk attitudes. We test these hypotheses experimentally and estimate both premiums empirically.

We experimentally characterize and measure the interaction between risk and time preferences. Decision makers are insensitive to time delay for small probabilities of gains but become progressively more sensitive as the probability of gain increases. Models allowing probability-time interaction and capturing magnitude effect fit the data better. Accounting for this interaction also leads to lower estimated discount rates.

ACs account for a significant fraction of energy consumption in tropical countries. We conduct a randomized control trial in which people are financially rewarded for setting higher AC temperatures. The gradual treatment — raising 1°C in each of two periods — outperformed the abrupt 2°C increase in maintaining higher temperatures during and after the intervention. Energy data confirmed lower consumption as a result.

Through four experiments (N=478) we show consumers are unduly averse to new technology vaccines (e.g., mRNA) due to higher perceived uncertainty of side effects. A herd behavior nudge — communicating increasing population vaccination rates — effectively increases uptake of new technology vaccines by alleviating perceived uncertainty rather than through social conformity or free-riding concerns.

We study how managers allocate resources in response to algorithmic recommendations programmed with specific risk aversion levels. Highly risk-averse algorithmic recommendations have a strong and persistent influence on order decisions even after removal — effects that hold regardless of whether advice comes from a human or algorithm, and regardless of decision autonomy.

Based on 6,731 incentivized daily forecasts over 377 days, individuals who engaged in counterfactual thinking prior to forecasting exhibited greater judgmental bias — underestimating the death toll, anchoring on more favorable scenarios, and showing insensitivity to trend changes. This was also observed among individuals who had recovered from COVID-19.

A randomized field experiment (N=1,121) showed an educational show on 5-minute showers triggered a boomerang effect — children below the target increased their shower time. Private and public pledges effectively countered this, reducing overall shower time and increasing adherence by raising willingness to meet the target and ensuring greater reductions for those most willing to reduce.

In two pre-registered RCTs on mobile usage, subjects pre-announced future targets and incentives behaved in a forward-looking manner consistent with habit formation — reducing usage before incentives begin, sustaining reductions during the incentivized period, and maintaining lower usage after incentives are removed. This provides empirical support for rational addiction theory in the domain of digital consumption.

Across two longitudinal field studies (5,686 observations, 153 individuals), individuals incentivized to both adopt a beneficial activity (language learning, daily walks) and curb smartphone use showed greater long-term reductions than those targeting smartphone use alone — especially when the positive alternative occupied a higher proportion of idle time.

We model consumers who anticipate elation and disappointment from information. Our key finding: a consumer facing a large potential loss of low probability may seek non-instrumental information purely for reassurance. This paradoxically causes a consumer less likely to face a loss to value information more than one more likely to face it — an effect that disappears at higher levels of information instrumentality.

We examine whether inequity persists on platforms like ride-share due to momentum in repeated contests. Momentum enables inequity to persist even long-run. Two interventions — modifying winner-take-all incentives and enabling participant communication — both reduce inequity in controlled lab experiments. The former reduces earnings asymmetry directly; the latter diminishes propensity to play momentum strategies.

Behavior change interventions often involve experimenter-assigned goals. We test whether granting individuals autonomy in goal-setting enhances motivation and behavioral change when average goal difficulty and incentives are held constant. In an eleven-week randomized field experiment (N=149), participants chose their own goal, were assigned a matched goal, or received no goal. Despite identical incentives and comparable average goal difficulty, participants who chose their own goals reduced phone use by 73% more and met their goals 11% more often than those with assigned goals, and reported lower depression, anxiety, and perceived addiction — providing causal field evidence on the motivational benefits of autonomy in goal-setting.

Diecidue and Somasundaram (2017) provide an axiomatic foundation for regret theory. This note shows that their original axioms are insufficient and provides the appropriate axioms that characterize regret theory. We also provide axiomatic foundations for the regret theory generalizations proposed by Loomes and Sugden (1987) and Bikhchandani and Segal (2011), defining clear boundaries between increasingly general versions of the theory.

Managers routinely face high-stakes, irreversible go/no-go decisions when evaluating innovation projects, and often seek external advice to reduce risk. We develop a model of information acquisition that jointly incorporates project uncertainty and signal fidelity, showing that optimal information demand is non-monotonic — managers should seek the most information only when uncertainty and signal fidelity are at intermediate levels. In controlled lab experiments, however, individuals instead show a strictly monotonic "uncertainty-reduction heuristic," demanding more information as uncertainty rises rather than adapting to the optimal baseline. We design a cognitive nudge — refocusing decision-makers on the project's probability of success or failure rather than its baseline uncertainty — that realigns behavior with the optimal policy, and show via a structural model that individuals systematically underweight external advice, especially when signal fidelity is high.

Long-run behavior change is usually evaluated using average outcomes, but behavior also has structure in how it unfolds over time. In a 51-week randomized field experiment with 523 adults using wearable activity trackers (170,000+ daily observations), participants received one of four incentive designs — fixed gain, fixed loss, gain-based streaks, or loss-based streaks — tied to personalized daily walking goals, followed by a 12-week period without incentives. All designs increased walking during the intervention, with gains persisting afterward, but different structures shaped distinct dynamics: gain-based streaks strengthened day-to-day momentum, while loss-based streaks accelerated recovery after lapses — patterns that persisted even after incentives ended.