Stereotypes, conceived as heuristics facilitating navigation of the social world, can be learned based on statistical contingencies between social group concepts and a stereotype-compatible characterization. However, a negative stereotype applied to an individual’s ingroup can conflict with the general tendency to see the self in a positive light, the self-positivity bias. Here, we used a modified minimal-group paradigm to create self-associations with members of an artificial social group (i.e., Fribble group) and test the impact of (counter-) stereotypical incidental learning on self- and group associations. Based on one pilot study (N = 43), a related simulation study, and two main experiments comprising human (N = 89 and N = 119) and simulated participant data, we tested whether minimally defined self- and group associations can be modified through incidental stereotype learning at the task temporal scale (i.e., duration of an experimental session rather than long-term socialization history). Results of the pilot and related simulation study established (i) that the pilot procedure for ingroup associations had to be strengthened; (ii) proof of concept of theoretical assumptions; (iii) feasibility of the experimental implementation of hypotheses; and (iv) optimal design for experimental studies to maximize statistical power. Findings of Experiment 1 subsequently showed that a stronger procedure was efficient to create self-ingroup associations and, by extension, ingroup stereotypes at the pre-learning phase. Furthermore, after incidental exposure to counter-stereotypical information, we observed changes in ingroup associations and self-related associations, even though the latter were never targeted by the counter-stereotypical information. Findings of Experiment 2 showed that the changes in self- and group associations observed in Experiment 1 were moderated by the proportion of positive/negative ingroup information to which participants were exposed during the learning phase. Experiment 2 also allowed discarding an alternative explanation of these changes in terms of memory loss. Together, findings provide an in-depth mechanistic account of the social-cognitive processes involved in stereotype learning and illustrate the benefits of mixed-methods (formal modelling, human and simulated participant data) to understand them.