To reduce uncertainty, individuals in groups can use personal and social information (i.e., information provided by others).Individuals are both emitters and receivers of social information and have to integrate personal and social information,giving rise to complex, poorly understood, collective dynamics. Here we applied evidence accumulation models (the drift-diffusion model) to group decision making to describe and understand these dynamics. We modelled the choice behavioras a process where evidence, in the form of sequentially arriving social information from other participants choices, isaccumulated until a threshold is reached. Our results show that highly confident individuals start close to the threshold andthus respond fast. Such early responders affects the subsequent dynamics, whereby humans weighted social informationas a linear function of the size of the majority for a particular option. Our results provide new insights into how socialinformation impacts the dynamics of decision making within groups.