A Bayesian Experimental Design Approach Maximizing Information Gain for Human-Computer Interaction
Abstract
A new information-theoretic approach based on Bayesian Experimental Design (BED) is applied to human-computer interaction, and in particular to multi-scale navigation. Instead of simply executing user commands, our BIG (Bayesian Information Gain) technique is modeling user behavior and tries to gain information by maximizing the expected mutual information provided by the users' subsequent input.
Origin : Files produced by the author(s)