A user may be exposed to multiple token instances representing stimuli that may influence the affective state of the user. Described herein are embodiments of systems, method, and computer programs for estimating affective response to a token instance of interest, selected from among the token instances. The token instance of interest is selected based on attention levels derived from eye-tracking data. In one example embodiment, the token instance of interest is a token instance in which the attention of the user is higher than the attention level of the user in at least one other token instance to which the user is exposed at the same time.