The goals of this study were to investigate the timing and the mechanism by which two types of model-centered instruction (MCI)--expert modeling (EM) and self-guided modeling (SGM)--might be made increasingly effective, efficient, and engaging for learners with different levels of expertise. The 62 pre-service and in-service evaluators who participated in the study were randomly assigned to one of these two types of MCI. The participants in the EM group were provided with the conceptual models used by experts to solve ethical conflicts within program evaluation. The participants in SGM group received no guidance in developing their own mental models. Regarding instructional effectiveness measures, there were no significant differences between the two types of MCI. However, inexperienced learners in the EM group invested less instructional effort and time than did those in the SGM group. In addition, inexperienced learners in the EM group also exhibited more engagement than did those in the SGM group. Therefore, EM is likely to be the more appropriate instructional design for inexperienced learners. Expert modeling required experienced learners to invest more mental effort, because if the conceptual model of the expert was redundant, it required them to integrate the previous schema with resulting overload of their working memory. Regardless of the types of MCI employed, the inexperienced participants showed significantly higher levels of attention and satisfaction than did the experienced participants.