Comparing the Input Validity of Model-based Visual Attention Predictions based on presenting Exemplary Situations either as Videos or Static Images

Functional cognitive models are used to explain observed human behavior. Applying such models to predict behavior requires generalization of the model to be applied in different application domains but also a careful consideration of model input data validity. Visual attention models have already been validated in various domains. But elicitation techniques to collect valid input data that is reproducible by others are still missing. For visual attention prediction model input data is determined mainly based on discussion between experts and individual experience, which is difficult to reproduce. We use a software tool to support input validity. The tool helps users to create attention models. It uses images of the situations that are investigated for stimulating the users to virtually put themselves into these situations. An experiment (n=40) showed that using looping videos instead of static images stimulates imagination in a different way. It has an effect on the models generated by the users and needs careful consideration