This study investigates the potential for a computer agent that considers group norms to mediate interest coordination regarding game scores in ultimatum games. Studies are being conducted to explore interactions between humans and computational agents. Implementing sociality into agents is important in order to be accepted by humans. Our previous work has shown that, in an interest-based group ultimatum game, group norm-aware agents promote fairer behavior in human players compared to agents that do not adapt to the behaviors of other players. However, because no comparison has been made between games involving only humans and those involving both humans and agents, further research is necessary. In this study, we explore the effect of agent participation by comparing games with and without agents. Specifically, we focus on how the agent adapts to the implicit group norms specific to each human group. By doing so, instead of addressing social dilemmas through a top-down approach by establishing rules for the group, we examine an agent-based method that aims to solve social dilemmas in a bottom-up manner by using group norm-aware agents both as intermediaries and group members. The results show that human players in the human–agent group exhibited fairer behavior than in the human-only group, although the benefits to the players were relatively small. This suggests that agent mediation promoted fair behavior, but at the cost of lower profits for the players in the group.