As interactive systems increasingly engage multiple users with varying attention spans and cognitive states, research on optimising the timing and content of explanations gains even more importance. Existing research focuses on multi-step explanations, where part of the explanation is available at the start of a scenario, and the remaining details are obtained at the halfway point. The work at hand explores whether it is more effective to deliver both, the partial and full explanations, at these time points or only the full explanation, depending on the users’ attention levels. We employ a Markov Decision Process (MDP) framework, utilizing backward Bellman induction to determine the optimal timing and content of the explanation delivery in an autonomous vehicle setting. A game-theoretic approach, guided by the SEEV (Salience, Effort, Expectancy, Value) model, is used to adapt the explanation strategy to the users’ varying needs and cognitive states. By balancing attention and minimizing cognitive strain, the model enhances user comprehension and trust. The model’s performance is evaluated in simulated multi-user scenarios, demonstrating its capability to dynamically adjust explanation delivery based on attention patterns. Our results suggest that this approach can significantly improve user experience by reducing cognitive load and optimizing information delivery.