The Social Cost of an AI Teammate: How an Artificial Teammate Reshapes Human-Human Communication in Small-Team Decision-Making

Source: arXiv Computers & Society (Academic)

The Challenge

As conversational artificial intelligence evolves from a passive assistance tool into an active collaborator, organizations are increasingly positioning AI agents as functional team members. However, little socio-technical research examines how the inclusion of an artificial peer alters interpersonal dynamics among human co-workers. This paper investigates the theoretical and empirical problem of how an AI teammate reshapes human-to-human communication patterns, individual perception of status, and team cohesion during small-group decision-making. The central issue is whether conversational AI creates hidden socio-cognitive costs that erode mutual engagement, psychological safety, and felt value among human team members during collaborative problem-solving.

Core Findings

The authors conducted a randomized controlled experiment comparing 16 hybrid teams (two humans plus one conversational AI teammate) against 17 all-human control teams (three humans) performing a high-stakes moral-dilemma decision task. Utilizing Group Communication Analysis (GCA), lexical analyses, and post-task surveys, the study evaluated discourse across six communication dimensions. Results indicate that while the AI teammate was consistently the most talkative and self-cohesive entity in every hybrid group, its utterances yielded the lowest information density and least novel information. Critically, the AI's presence disrupted human-to-human relational dynamics: human teammates exhibited reduced mutual responsivity and diminished social impact toward one another. Human participants in hybrid teams reported significantly lower levels of belonging and status, with conversational dominance by the AI directly correlating with humans feeling less valued. Notably, this social cost manifested immediately at baseline rather than developing over time.

Strategic Takeaway

For organizational leaders and digital transformation strategists, this study underscores a critical friction in human-AI teaming: granting AI conversational agency can inadvertently degrade human social capital and team dynamics. Digital leaders must move beyond viewing AI solely in terms of efficiency and evaluate its socio-technical impacts on collaboration. To mitigate baseline disruptions in belonging and status, managers should carefully calibrate AI participation levels, establish explicit norms for AI interaction, and design collaborative workflows that prioritize human-to-human dialogue, ensuring AI remains a supportive augment rather than a dominant, cohesion-eroding team member.

Deep Dive Q&A

How does adding an AI teammate alter communication between human team members?

The inclusion of an AI teammate significantly decreases human-to-human responsivity and social impact, causing human team members to interact less dynamically with one another and report lower feelings of personal status and team belonging.

Did the AI teammate provide high-quality conversational input?

No. Although the AI agent was the most talkative and self-cohesive entity in every hybrid team, its contributions contained the lowest information density and generated the least new information compared to human participants.

Does the social friction caused by AI teammates worsen over time?

The study found that the social costs—such as reduced human responsivity and diminished feeling of value—were present immediately at baseline rather than gradually emerging over the course of the team's conversation.