When I Need a Stand-in
M.S. Thesis — AI agents that negotiate on your behalf
The High-Stakes Context
Negotiations require high emotional intelligence and strategic information disclosure. This M.S. Thesis targets the high-stakes domain of agentic negotiations, creating AI agents that represent users in sensitive situations.
The Architectural Orchestration
Architected a multi-agent reflective pipeline that orchestrates user-guided inference and dynamic disclosure mechanisms. This ensures AI agents can negotiate on behalf of users while strictly respecting their predefined communication boundaries and identity.
The Deployment & Scale
Deployed as a fully functioning, highly usable system prototype capable of handling complex multi-turn negotiations.
The Empirical Evaluation
Evaluated through rigorous mixed-methods research, combining system logs and user feedback to prove the system's ability to maintain trust and accurately represent the user's intent.
The Outcome
- M.S. Thesis completed and defended (KAIST, Feb 2026)
- Being repackaged for conference submission (1st author)
- Live demo available