Towards simpler models of choice behaviour

Abstract

This paper argues that modelling specific psychological processes, rather than modelling observed behaviour, enables the design of simpler models that provide a psychologically grounded explanation of choice under risk. As a particular case, it studies the process of weighing the more favourable and less favourable parts of each alternative and shows how this process yields a model that, through a single function, captures behavioural patterns typically described by multiple components. Formalising this process leads to two structural assumptions: within each choice set, a common separation point divides the outcomes of the alternatives into favourable and unfavourable outcomes, and this point acts as a reference point for relative valuation. Under these assumptions, a minimal valuation structure based on sensitivity to less favourable outcomes is sufficient to account for different attitudes towards risk, the reflection effect, and biases in probability perception. The approach thus provides a parsimonious framework in which loss aversion is reinterpreted as sensitivity to relatively less favourable outcomes, without the need to introduce separately a probability-weighting function and a value function over monetary outcomes.