Empowerment as a metric for optimization in HCI

Abstract

We propose a novel metric for optimizing human-computer interfaces, based on the information-theoretic capacity of empowerment, a task-independent universal utility measure. Empowerment measures, for agent-environment systems with stochastic transitions, how much influence, which can be sensed by the agent sensors, an agent has on its environment. It captures the uncertainty in human-machine systems arising from different sources (i.e. noise, delays, errors, etc.) as a single quantity. We suggest the potential empowerment has as an objective optimality criterion in user interface design optimization, contributing to the more solid theoretical foundations of HCI.