| Our goal is to develop a paradigm to train novices on complex visual tasks more efficiently. We explore the value of attentional highlighting---which involves making saliency modulations to guide novices---by studying the domain of fingerprint matching. Using data collected by Busey et al. (in press) we construct probabilistic models to predict fixation sequences of novices and experts. We then conduct experiments with novices and record eye gaze as they examine prints for similarities and differences. During a training phase, attentional highlighting cues subjects along an expert fixation path. We evaluate the effect of training by comparing pre- and post-training performance on a set of images including training and transfer images. Using our models, we compute a log odds score that reflects the degree of expertise exhibited by a subject's fixation sequences. Our results show that the gaze behavior of subjects becomes more expert-like on trained and untrained images. |
