Decision making and the intrinsic cost of cognitive control
Humans are constantly faced with dilemmas between choices options with rational, more optimal, control-demanding, outcomes (maintaining a diet) and options with outcomes that are tempting in an automatic or habitual fashion (a slice of cake). Although these decisions carry great importance in our everyday lives, attempts to exert appropriate control often fail. I will present behavioral, neural, and computational findings that suggest that the mobilization of cognitive control in these choices is a form of cost-benefit decision making. First, I will present behavioral evidence that demands for cognitive control register as costly. Next, I will describe an integrative model of cognitive effort-based decision making which draws from ideas from economic labor supply. In this model, the benefits from control are weighed against its cost in a nonlinear fashion. This nonlinearity motivates the worker to strike an optimal balance, or trade-off, between labor and leisure. Even though this model provides a normative account of effort-based decision making, it does not provide a deep understanding of its cognitive mechanisms. Recently, I have used reinforcement learning (RL) to investigate the mechanistic principles behind the cost-benefit analysis. Modern RL theories formalize the trade-off as a competition between a cheap but inaccurate “model-free” system that gives rise to habits, and an expensive but accurate “model-based” system that implements planning. In recent behavioral and modeling work, I show how humans integrate these systems’ accuracy and cognitive demands to adaptively arbitrate between them. This work points to a new perspective on effort-based decision making that centers on motivation or value, rather than limited resources.