Blink : observing thin slices of behavior to determine users' expectation towards task difficulty

Abstract

This work aims to address the following question: is it possible to infer the users' expectations regarding task difficulty by watching them just before the actual start? We present a study where people acting as evaluators determined users’ expectations based on non-linguistic social signals in a 20 seconds video clip. The evaluations were performed using a five-point scale and the average error of the evaluations was of one point. Preliminary results suggest what type of signals was used by the evaluators to determine the users’ expected difficulty with the task.