Noisy neural coding and decisions under uncertainty

I derive a noisy neural coding model (NCM ) and pit its performance against prospect theory plus additive noise (PT) using some prominent recent datasets collected to measure PT parameters. The NCM is based on the premise that choice patterns observed under uncertainty may originate from noisy perceptions of choice stimuli, which are optimally combined with mental priors to obtain actionable decision parameters. This contrast with PT, which models preferences as deterministic, but adds a noise term for empirical implementations. I show how the parameters emerging from the NCM naturally map into PT parameters. The NCM can thus be seen as a generative model for PT. At the same time, the NCM departs from PT in that it is inherently stochastic. This results in novel predictions about systematic correlations between PT parameters, as well as pointing to instances under which PT will be violated. Using Bayesian hierarchical models to fit the data, I find substantial support for these predictions. The NCM further consistently outperforms PT in terms of predictive ability. These results contribute to the nascent literature documenting the role played by imprecise cognition in economic decisions.