The eyes are considered a ‘window’ into the mind, shedding light on the cognitive processes leading to explicit decisions. Yet, eye movements are also a distinct oculomotor decision that is the outcome of an independent cognitive system. When exploring novel environments, the convergence and divergence of gaze and explicit decisions highlights the eyes’ dual role. However, the computational mechanisms underlying this interplay and its contribution to adaptative behavior remains unclear. Combining virtual reality based associative learning paradigms with computational modeling across one exploratory and two pre-registered experiments (Total N = 115), we examine learning as multi-system process encompassing explicit decisions and ocular expectations. The two exhibited a partial overlap in their learning. On the one hand, participants’ explicit predictions about target location and their subsequent anticipatory gaze direction robustly converged, and gaze reveals the latent probability of the explicit choice. Moreover, gaze exhibited computational hallmarks of confidence, reflecting an oculomotor assessment of the explicit prediction. On the other hand, ocular and explicit decisions consistently diverged indicating an independent computation reflected in gaze. Although the two systems learn and ascribe value similarly, they utilize different decision mechanisms. Oculomotor decisions were less driven by choice consistency across trials enabling increased exploration. Under uncertainty explicit and oculomotor decisions tended to diverge, enabling a hedging of one’s decision. These findings highlight how together, explicit and oculomotor systems’ learning enables adaptive embodied behavior.Significance StatementThe eyes often reveal additional information beyond one’s explicit decision, offering a ‘window into the mind’ (e.g., leading poker players to wear sunglasses). Yet, eye movements are the product of a distinct system, occasionally diverging from explicit decisions. However, what additional information is conveyed by gaze, and what leads to its occasional divergence remain unclear. Using virtual reality and computational modeling, this study demonstrates a partial overlap between the two. Generally, they robustly converge, and gaze reflects a confidence-like assessment of explicit decisions. Yet they also diverge due to distinct decision mechanisms. Their divergence enables a ‘hedging’ of one’s decision under uncertainty. This demonstrates a multimodal decision process, where different action systems make overlapping but independent choices.