Modeling is an activity in the software development life cycle in which different experts and stakeholders collaborate as a team. In collaborative modeling, adhering to the optimistic versioning paradigm allows users to apply concurrent changes to the same model. In such a situation, conflicts may arise. To have an integrated yet consistent merged model, conflicts have to be resolved. To this end, automation is currently at its limit or is not supported at all, and user interaction is often required. To alleviate this flaw, there is an opportunity to apply Artificial Intelligence techniques in a collaborative modeling environment to empower the provisioning of automated and intelligent decision-making. In this paper, we propose the use of reinforcement learning algorithms to achieve merging conflict resolution with a high degree of automation. This enables the personalized and quality-based integration of model versions. To evaluate our idea, we demonstrate the resolution of UML class diagram conflicts using a learning process in an illustrative modeling scenario. We also show the applicability of our approach through a proof of concept implementation and assess its accuracy compared to the greedy and search-based algorithms. Moreover, we conducted an experience with five experts to evaluate the satisfaction of actual users with the selection of resolution actions for different conflicts. The result of the assessment validates our proposal with various syntactic and semantic conflicts.