Integrating an exit choice model into a microscopic crowd dynamics model is an essential approach for obtaining more
efficient evacuation model. We describe various aspects of decision-making capability of an existing rule-based exit
choice model for evacuation processes. In simulations, however, the simulated evacuees clogging at exits have behaved
non-intelligently, namely they do not give up their exits for better ones for safer egress. We refine the model to endow
the individuals with the ability to leave their exits due to dynamic changes by modifying the model of their excitement
resulted from the source of panic. This facilitates the approximately equal crowd size at exits for being until the end
of the evacuation process, and thereby the model accomplishes more optimal evacuation. For further intelligence, we
introduce the prediction factor that enables higher probability of equally distributing evacuees at exits. A simulation to
validate the contribution is performed, and the results are analyzed and compared with the original model.