Comparisons of keystroke-level model predictions to observed data

Abstract

Comparison of model prediction against observed data is an investigative step used in cognitive modeling research for human-computer interaction. In this paper we describe comparisons between Keystroke-Level Model (KLM) predictions and user behavior by total duration, aggregated events and Cohen's Kappa. Our preliminary investigations support the validity of KLM mental preparation duration and placement rules in modeling interaction with handheld devices but suggest changing a previously-published parameter.