The Australian Responsible Autonomous Agents Group
The Australian Responsible Autonomous Agents Collective (ARAAC) is a cross-institutional group of AI researchers based across Federation University Australia, Deakin University, and the University of New South Wales. The collective specializes in multi-objective reinforcement learning (MORL), explainability, transparency, and AI safety, building the theory and practice of balancing AI performance and responsibility. Co-led by Professor Peter Vamplew (Federation University) and Professor Richard Dazeley (Deakin University), ARAAC's mission is to drive innovation in autonomous agents while anchoring discoveries in responsible and ethical practices, ensuring AI alignment with human interests.
Funding Details
- Annual Budget
- -
- Monthly Burn Rate
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $210,000
- Fiscal Sponsor
- Federation University Australia
Theory of Change
ARAAC believes that conventional reinforcement learning's unconstrained maximization of a single reward or utility measure poses fundamental risks for AI safety and alignment. Their theory of change centers on developing multi-objective reinforcement learning (MORL) methods that use vector rewards instead of scalar rewards, enabling AI systems to simultaneously balance multiple competing objectives such as performance, safety, fairness, and ethical constraints. By treating each aspect of the alignment task as a separate objective, MORL can produce aligned behavior that is difficult or impossible to achieve using scalar reward definitions. This approach supports pluralistic alignment where multiple conflicting values or stakeholders must be considered, and enables methods for automatically learning human preferences and ethics and incorporating them into autonomous agents.
Grants Received
from Survival and Flourishing Fund
Projects
No linked projects.
People
No linked people.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Apr 2, 2026, 10:09 PM UTC
- Created
- Mar 18, 2026, 11:18 PM UTC