Individual risk attitudes arise from noise in neurocognitive magnitude representations

Abstract

Humans are generally risk averse, preferring smaller certain over larger uncertain outcomes.Economic theories usually explain this by assuming concave utility functions. Here, we provide evidence that risk aversion may also arise from relative underestimation of larger monetary payoffs, a perceptual bias rooted in the noisy logarithmic coding of numerical magnitudes. We confirmed this with psychophysics and fMRI, by measuring behavioural and neural acuity of magnitude representations during a magnitude perception task and relating these measures to risk attitudes during separate risky financial decisions. Computational modelling suggested that participants employ similar mental magnitude representations in both tasks, with correlated precision across perceptual and risky choices. Participants with more precise magnitude representations in parietal cortex showed less variable behaviour and less risk-aversion. Our results highlight that at least some individual characteristics of economic behaviour may reflect capacity limitations in perceptual processing rather than processes that assign subjective values to monetary outcomes.

A nonlinear code for event probability in the human brain

“…This method consists of applying basis functions to a quantity of interest to obtain features and then modelling the fMRI signal in each voxel as a linear combination of these features, following the general linear model approach that is massively used in fMRI (Friston et al, 2007). This approach is related to other methods that can accommodate nonlinear tuning curves, such as population receptive field (pRF) mapping (Barretto-García et al, 2023; Dumoulin & Wandell, 2008; Harvey et al, 2013). The key difference is that pRF methods assume a specific form of nonlinearity, typically bell-shaped tuning curves, corresponding to the idea that a voxel is selective for a range of values.…”

Compressed Representations and Attentional Competition in Numeric Integration for Average Estimations

“…This demonstrates the importance of a cognitive model to explicitly take cognitive noise into account. Recently, models have been developed that combine noisy logarithmic representations with optimal Bayesian decoding of this noisy information (Barretto-García et al, 2023;Khaw et al, 2021;Petzschner et al, 2015). These models predict that higher levels of representational noise should lead to stronger compression of the numeric representations.…”

“…In general, our everyday valuebased decisions depend on the processing of the numerical magnitudes represented in symbolic terms (Brezis et al, 2018;Olschewski et al, 2024). Therefore, a topical research agenda examines communalities in the cognitive processes between numerical judgments and value-based tasks, and how cognitive processes in the integration of numerical information shape value-based behavior (Barretto-García et al, 2023;Dutilh & Rieskamp, 2016;Frydman & Nave, 2017;Khaw et al, 2021;Olschewski et al, 2021;Schley & Peters, 2014;Trueblood et al, 2013;Vanunu et al, 2019; see also Smith & Krajbich, 2021). Similarly, representing and processing numerical magnitudes is a cornerstone in the development of mathematical skills (De Smedt et al, 2013;Merkley & Ansari, 2016) and statistical reasoning (Schulze & Hertwig, 2021).…”