Thank you for attending the workshop!

We hope to see you next year and on our next initiatives
If you would like to provide feedback on the workshop or to opt in for future RaD-AI events, please fill out this following poll: https://forms.gle/rqN45pq4sP3UGjhdA

2025 AAMAS RaD-AI Workshop: Program

 Detroit, Michigan (USA) | 20 May 2025
All times are in local time in Detroit, MI (USA) (i.e., EDT, or GMT-4)

09:00-09:15 Introduction (Co-Chair David Aha)

09:15-10:00 Session 1 (Theme: GenAI)

09:15-09:35 GPT-4 as a Moral Reasoner for Robot Command Rejection (Invited HAI-2024 Paper)

Ruchen Wen, Francis Ferraro, & Cynthia Matuszek (University of Maryland, Baltimore County; USA)

09:40-10:00 Disobedience for a Cause: Leveraging Implicit Objectives in User Plans by LLM Surrogates 

Dale Peasley, Zachary Gray, & Sandip Sen (University of Tulsa, USA) paper

10:00-10:30 Coffee Break

10:30-12:30 Session 2 (Theme: Mental Modeling & HAI)

10:30-10:50  Mental Model-based Generation of Lies for Insider Threat Modeling 

Brittany Cates & Sarath Sreedharan (Colorado State University, USA) paper

10:55-11:10 Rebellion for the Greater Good: When AI Agents Disobey to Optimize Team Performance

Eric Peterson & Sandip Sen (University of Tulsa, USA) paper

11:15-11:35 Obey, Refrain, or Contravene?

Divya Sundaresan (NCST, USA), Gordon Briggs (NRL, USA), & Munindar P. Singh (NCST, USA) paper

11:40-11:55 Frictive Policy Optimization for LLM Agent Interactions 

James Pustejovsky (Brandeis University, USA) & Nikhil Krishnaswamy (Colorado State University, USA) paper

12:00-12:30 Discussion: Morning Presentations

12:30-14:00 Lunch

14:00-15:45 Session 3 (Theme: Decision Making)

14:00-14:45  Trustworthy Algorithmic Delegate: An Alignable Decision Maker (Invited Talk)

Everyone wants ethical behavior from autonomous systems, but humans don’t even agree on what ethics to use. In some domains, such as medical triage, there is not a clear “ethical” or “best” decision. Nevertheless, junior people need help with these decisions, and systems will eventually need to make them quickly, without oversight. The Trustworthy Algorithmic Delegate (TAD) is constructed to rapidly acquire a representation of how an individual trusted decision maker makes decisions and apply that standard to new decisions.

Matthew Molineaux (Parallax Advanced Research, USA)

14:50-15:10  R-HTN: Rebellious Online HTN Planning

Hector Munoz-Avila (American University, Paola Rizzo (Intergens, Italy) & David W. Aha (NRL, USA) paper

15:15-15:30  Plan Recognition for Rebel Agent Systems

Michael Cox (Wright State University, USA) paper

15:35-15:50 You said No, What's Next?

Sarath Sreedharan (Colorado State University, USA) and Gordon Briggs (NRL, USA) paper

15:45-16:15  Coffee Break

16:15-17:30 Session 4 (Theme: Related work at AAMAS-25)

16:15-16:35 Model and Mechanisms of Consent for Responsible Autonomy (Invited AAMAS-25 Paper) 

Anastasia S. Apeiron (Utrecht University (UU), The Netherlands), Davide Dell’Anna (UU, The Netherlands), 

Pradeep K. Murukannaiah (TU Delft, The Netherlands), & Pinar Yolum Birbil (UU, The Netherlands)

16:40-17:15 Discussion: Afternoon Presentations

17:15-17:30  Conclusion and Wrap-up