The objective of this study was to gain insights into how victims use their visual attention to determine the sincerity of an offender’s apology during simulated victim–offender mediation. We hypothesized that the victims’ visual attention (gaze fixation duration) would be focused more on the offender’s upper (than lower) face area, especially the eyes and the eyebrows, to infer the degree to which the offender suffers, takes responsibility, and has empathy for the victim. In turn, we expected these inferences to positively predict the perceived sincerity of the apology. Additionally, we took into account the victims’ a priori expectations regarding the sincerity of the apology and (positive) attitudes toward resocialization programs (ARPs). We expected both variables to enhance the above proposed process through which victims determine the sincerity of the apology. Fifty-eight students took the victim’s role in a fictitious crime scenario and watched a video in which the offender offered a remorseful apology. We obtained eye tracking data to determine the participants’ fixation and attention distribution. As expected, the participants’ gaze fixated significantly longer on the upper face. The results also showed that their prior expectations, positive ARPs, and inferences of suffering and responsibility taking after the apology all positively predicted the perceived sincerity. However, unexpectedly, gaze duration was not directly associated with these inferences. The fixation duration on the upper face instead appeared to moderate how ARPs predicted inferences of responsibility taking. More concretely, the exploratory path model analyses revealed that when the participants had more positive a priori ARPs, the longer they focused on the offender’s eyes and eyebrows and the more they concluded that he took responsibility for his actions (which in turn predicted more sincerity). However, for those with relatively negative ARPs, it was the other way around: the more they focused on the eyes and the eyebrows, the stronger they inferred that the offender did not take responsibility (which predicted less sincerity). Our findings demonstrate the vital role of the victims’ a priori attitudes, expectations, and eye gaze behavior in understanding the reception and the evaluation of offenders’ apologies. This study also suggests how novel technology can be used to investigate gaze behavior in the field of victim–offender mediation.