This paper focuses on ways to improve computer models of decision-making to predict human behaviour in the “Iowa Gambling Task (IGT)”. This work concentrated on faulty pattern unlearning from the human decision making (HDM) model through the concept Redundancy Count (RC), which controls learning of game play when human losses interest in the game. In this work, we tested several values of RC to find out the optimum one. HDM model based on decay reinforcement learning, prospect theory and trial dependent choice rule has been used. For validation of proposed concept, mean square deviation (MSD) is used. Result show improvement in MSD after 100 simulations, while the MSD value degrades in lesser simulations. This work is inspired from the concept of machine unlearning, where the learned model will unlearn the faulty patterns.