Research

Improving Decision-Making Under Risk: The Effect of Information Processing Guidance

with Abigail Barr and Martin Sefton

Abstract: We introduce Information Processing Guidance (IPG) — a novel Bayesian boost that directs attention to decision-relevant information and scaffolds the reasoning process — to test whether it reduces Bayesian decision errors, whether any improvement survives the withdrawal of guidance, and how it interacts with task difficulty and cognitive traits. In a laboratory experiment, participants complete multiple rounds of a Bayesian investment task across eight problems of varying difficulty, measured by a newly constructed Cognitive Complexity Score (CCS). Continuous guidance significantly lowers decision errors; the reduction does not survive withdrawal on average, but it does persist for participants with the highest Cognitive Reflection (CR). Task difficulty itself reliably raises errors but does not moderate the guidance effect, unlike CR, which significantly amplifies it. Guidance therefore functions as a short-term attentional scaffold for most participants, but as a durable, long-term boost for the most reflective, whose gains survive its removal.

Work in progress

From Metric to Mind: Cognitive Complexity Scores and Subjective Perception

with Abigail Barr and Martin Sefton

Abstract: We ask whether an objective measure of task complexity, the Cognitive Complexity Score (CCS), captures how difficult people perceive a decision to be. CCS is an a priori measure of the information, operations, nesting depth, and working-memory load a task requires. We test it in an online experiment in which participants solve incentivised investment tasks that call for Bayesian updating. Subjective difficulty is elicited in two ways: a Likert rating and pairwise difficulty comparisons scaled by a ridge-penalised Pairwise Empirical Bayes model. We find that the objective measure is strongly associated with subjective difficulty, but far more sharply when difficulty is elicited by pairwise comparison than by Likert rating, consistent with direct ratings carrying more scale-use noise. The paper contributes an externally validated complexity measure and a benchmarked comparison of two elicitation methods for measuring perceived task difficulty.

The Role of Observation and Advice in Overcoming Reinforcement Heuristics Bias

with Abigail Barr and Martin Sefton

Abstract: We investigate whether naive advice and observational learning reduce susceptibility to the reinforcement heuristic in a binary-choice task under risk. Pairs of participants make repeated choices either independently, after observing their partner's prior choice, or after receiving their partner's advice. We find that neither advice nor observation reduces the frequency of non-optimal choices on average, but a small learning effect emerges in both the advisor and advisee, with advisors learning faster than participants who decide independently, suggesting that giving advice sharpens attention to decision quality. In addition, participants are more likely to follow advice than to imitate observed choices, and, counter-intuitively, comply more readily with low-quality (heuristic-consistent) advice than with high-quality (Bayesian-consistent) advice.

Interrogating Reinforcement Heuristics

with Abigail Barr and Martin Sefton

Abstract: We examine the role of noisy outcome feedback in the binary-choice task, comparing participants who receive feedback on their choices with those who do not. We find that feedback does not improve overall decision quality: Bayesian-optimal choice rates and reinforcement heuristic error rates are similar across the feedback and no-feedback conditions. Clustering analysis of the No-feedback group shows that only about 20% of participants behave consistently as Bayesian decision-makers, very few adhere consistently to the reinforcement heuristic, and the majority fluctuate between the two, in a pattern closer to random choice. The results suggest that the reinforcement heuristic does not stem from a simple information-processing deficit that feedback can correct, but instead reflects a more fundamental feature of how people navigate repeated choices under uncertainty.

The Right-to-Buy Scheme on the Development of Children

with John Gathergood, Matteo Carlo Maria Sandi, Richard Disney, and S. J. Machin

Abstract: Right to Buy (RTB), a large scale housing policy by which incumbent tenants in public housing could buy their properties at heavily subsidised prices, increased the homeownership rate in the UK by over 10 per cent between 1980 and the late 1990s. Using the ONS Longitudinal Study, this project will study the impact of childhood exposure to RTB on their educational and professional trajectories. The research will provide evidence of a potential novel means by which gentrification, and housing provision, may contribute to increase human capital accumulation and improve the educational and professional trajectories of children born in low income settings.