In our uncertain and dynamic world, we must select appropriate adaptation strategies to reach desired outcomes. The literature describes two types of such strategies: model-based and model-free. Model-based strategies are flexible but cognitively demanding, as they rely on an internal representation of dependencies within the environment. Model-free strategies, in turn, depend on direct associations between stimuli and rewards learned through experience, which makes it effortless but more rigid. Therefore, a different strategy may be reasonable depending on the circumstances. This study investigates how knowledge of task structure influences strategy selection in a probabilistic reversal learning task, where reward contingencies can change over time, requiring adaptive flexibility. We hypothesized that informed individuals would adopt a model-based strategy reflected by a counterfactual reinforcement learning model, whereas uninformed individuals would rely on a model-free approach, aligning with basic reinforcement learning. Contrary to our hypothesis, the behavior of the informed and uninformed participants did not differ, and both groups showed evidence of the model-based strategy. Additionally, we examined neural activity in the ventromedial prefrontal cortex, considered a key region in the brain’s valuation system, to find a signal associated with model-based valuation. Surprisingly and in contrast to previous literature, we found no such activity in this region. These unexpected and mixed results should, however, be interpreted with caution, as they may reflect specific features of the study design rather than effects generalizable to broader contexts.