A Neural Network Approach Of The Dynamical Inconsistency Problem Of Decisions.

Abstract

The intertemporal inconsistency problem in decision making refers to those cases in which the
individual has chosen in favor of those alternatives that ensure a delayed retribution, but in the interim
he changes his choice to an alternative that is realizable early, even when the alternative available in
long time is optimal for him.
This kind of inconsistencies over choices has been modeled as the interaction between two systems: a
cool system which has an automatic and slow operation, and another hot system which operates in a
fast and automatic fashion. The hot system makes its choices biased towards those alternatives with
early realization while the cool system has a remarked preference for the delayed gratification
alternatives. The individual decision results as the solution to the competition in operation between the
two systems.
In this article, we model the dynamical mechanisms that govern both systems functioning and their
interaction when they are engaged in decision making. These mechanisms are revealed by an analysis
of a neurocaomputational approximation to the network operations. Then we consider the dynamical
system that measures the activation of the minimal network when a stimulus input excites both
systems.
As each system's activation is represented by nonlinear equations, the nullclines structure generates
multiple equilibrium nodes which are found and classified given its local stability. As a final exercise,
this work studies the relevance of each parameter in the dynamics of the system, and in the
determination of the final decision.