Animals receive many signals from the environment, and use these signals to decide their behavior. Many studies discuss how animals decide their behavior in response to signals, with many theoretical studies focusing on this problem. Many animals decide their actions based on both innate traits and their experience. However, the question on when animals adopt innate decision-making and when animals behave based on their experience remain. It is difficult to resolve this question using existing mathematical models. Here, we constructed a mathematical model that included decision-making using both innate and learning behaviors by extending the Bayesian learning model. In particular, we assumed an initial prior distribution in the learning model as an evolutionary trait. This approach allowed us to construct a simple framework to consider decision-making by both innate and learnt behavior. Furthermore, we applied this model to avoidance behavior, and analyzed when animals avoid signals innately and when animals avoid signals by using their experiences. We showed that innate avoidance evolves when the cost of attacking signals and the frequency of the signal is high. In contrast, animals avoid signals based on experiences under the opposite conditions.