To elucidate the mechanisms of human error, this study aims to construct a simulation model using the cognitive architecture ACT-R that precisely reproduces the variability in reaction times during the Color Cued Operation Switching (CCOS) task. First, we conducted an experiment with six participants performing the CCOS task to measure reaction time data. We then constructed an ACT-R model that assumes interruptions in the cognitive process to account for this variability. Finally, the model’s parameters were optimized using a multi-objective genetic algorithm to fit the cumulative distribution of the experimental reaction times. The results show that the optimized model accurately reproduced the experimental reaction time distributions in simulation. The structure of this successful model suggests the existence of adaptive cognitive processes under time pressure, such as the parallelization of knowledge retrieval and processing