Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Amplifying Minority Voices: AI-Mediated Devil's Advocate System for Inclusive Group Decision-Making

Soohwan Lee, Mingyu Kim, Seoyeong Hwang, Dajung Kim, Kyungho Lee

TL;DR

This paper tackles the problem of minority voices being overwhelmed by majority influence during group decisions. It proposes an LLM-powered devil's advocate that anonymizes dissent and presents it as its own to reduce social pressure and increase psychological safety. A multi-agent architecture—Summary, Paraphrase, Conversation, and AI Duplicate Checker—supports anonymous dissent representation, contextual counterarguments, and novelty checks while balancing participation with an approximately eight-turn cadence. The work identifies limitations of text-based interaction and fixed timing, outlining directions toward adaptive, multimodal, and ethically aware designs to improve inclusivity and decision quality.

Abstract

Group decision-making often benefits from diverse perspectives, yet power imbalances and social influence can stifle minority opinions and compromise outcomes. This prequel introduces an AI-mediated communication system that leverages the Large Language Model to serve as a devil's advocate, representing underrepresented viewpoints without exposing minority members' identities. Rooted in persuasive communication strategies and anonymity, the system aims to improve psychological safety and foster more inclusive decision-making. Our multi-agent architecture, which consists of a summary agent, conversation agent, AI duplicate checker, and paraphrase agent, encourages the group's critical thinking while reducing repetitive outputs. We acknowledge that reliance on text-based communication and fixed intervention timings may limit adaptability, indicating pathways for refinement. By focusing on the representation of minority viewpoints anonymously in power-imbalanced settings, this approach highlights how AI-driven methods can evolve to support more divergent and inclusive group decision-making.

Amplifying Minority Voices: AI-Mediated Devil's Advocate System for Inclusive Group Decision-Making

TL;DR

This paper tackles the problem of minority voices being overwhelmed by majority influence during group decisions. It proposes an LLM-powered devil's advocate that anonymizes dissent and presents it as its own to reduce social pressure and increase psychological safety. A multi-agent architecture—Summary, Paraphrase, Conversation, and AI Duplicate Checker—supports anonymous dissent representation, contextual counterarguments, and novelty checks while balancing participation with an approximately eight-turn cadence. The work identifies limitations of text-based interaction and fixed timing, outlining directions toward adaptive, multimodal, and ethically aware designs to improve inclusivity and decision quality.

Abstract

Group decision-making often benefits from diverse perspectives, yet power imbalances and social influence can stifle minority opinions and compromise outcomes. This prequel introduces an AI-mediated communication system that leverages the Large Language Model to serve as a devil's advocate, representing underrepresented viewpoints without exposing minority members' identities. Rooted in persuasive communication strategies and anonymity, the system aims to improve psychological safety and foster more inclusive decision-making. Our multi-agent architecture, which consists of a summary agent, conversation agent, AI duplicate checker, and paraphrase agent, encourages the group's critical thinking while reducing repetitive outputs. We acknowledge that reliance on text-based communication and fixed intervention timings may limit adaptability, indicating pathways for refinement. By focusing on the representation of minority viewpoints anonymously in power-imbalanced settings, this approach highlights how AI-driven methods can evolve to support more divergent and inclusive group decision-making.

Paper Structure

This paper contains 6 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. AI-assisted Group Decision-making
  4. AI-mediated Communication
  5. System Overview
  6. Limitations & Future Work

Figures (2)

  • Figure 1: Four patterns of AI-mediated communication in group settings: (A) human requests and relays AI-generated content, (B) human selectively shares AI's output, (C) AI reformulates and presents human's message, and (D) AI directly facilitates communication between participants. Arrows indicate information flow, with numbered sequences showing order of interactions.
  • Figure 2: System Overview and Example Task Scenario. The figure illustrates a team leader promotion decision task, where participants discuss candidate qualifications in a chat interface. Minority members can privately share dissenting views via direct messages(DM) to the system, which reformulates and presents them as AI-mediated messages. If there is no DM with an opposing opinion, the system will send out a counterargument that it has generated on its own. The system architecture consists of a chat interface, database, and server, processing both public discussions and private DMs through four key agents: (A) Summary Agent for analyzing public opinion, (A') Paraphrase Agent for rephrasing minority views, (B) Conversation Agent for generating contextual counterarguments, and (C) AI Duplicate Checker for ensuring message novelty via cosine-similarity comparison.