Many organisms, including humans, face the dilemma between choosing an option with a well-known output (exploitation) and sampling from a lesser known alternative (exploration). There are two common strategies used in solving this dilemma. One strategy is to guide decisions based on an “information bonus”, by choosing the option that provides more information (directed exploration). A second strategy is to maintain a certain decision noise, thus with an increased sampling of other options “by chance” (random exploration). Previous work by Wilson et al. (2014) showed that humans use both of these strategies when faced with this dilemma. The present study asks: How do humans tackle the exploration-exploitation (EE) dilemma within a social context? We addressed this question by empirically using the “Horizon Task” (Wilson et al., 2014), where participants choose between two options to maximize rewards, with information about these options either equal or unequal. The experiment consisted of 320 trials under two conditions. In the individual condition, participants completed the task alone. In the social condition, participants worked in dyads and completed the task as a team. Each participant completed 160 trials individually and 160 trials in dyads (counterbalanced order). During the collective part, dyads discussed their choices, and their conversations were recorded. Pilot data (N=18 individuals, 9 dyads) indicated that dyads and individuals were equally likely to make exploratory choices aimed to increase information (directed exploration). However, dyads were much less likely to make random, impulsive errors (random exploration) compared to individuals. When equal information was available about the two bandits, individuals made random errors (i.e. choosing the bandit with lower expected value) significantly more frequently compared to dyads. Based on these findings, we are currently collecting data from 17 dyads of participants. Preliminary results showed that our findings in the pilot study are being replicated. Understanding the dynamics of collective decision-making in exploration tasks can illuminate how social interactions influence cognitive processes. Our study seeks to provide insights into collective intelligence and its potential applications in collaborative environments.