The Australian Responsible Autonomous Agents Group

The Australian Responsible Autonomous Agents Collective (ARAAC) is a cross-institutional collective of AI researchers focused on addressing the technical, ethical, and humanitarian challenges posed by the rise of intelligent systems, particularly systems composed of autonomous agents. The collective evolved from the Federation Learning Agents Group (FLAG), a reinforcement learning-focused research group founded by Professor Peter Vamplew at Federation University Australia, which expanded to include members across multiple Australian universities. ARAACis co-led by Professor Peter Vamplew of Federation University Australia (Ballarat) and Professor Richard Dazeley of Deakin University (Geelong). Vamplew has been instrumental since 2008 in establishing multi-objective reinforcement learning (MORL) as a sub-field of reinforcement learning, while Dazeley leads Deakin's Machine Intelligence Lab and specializes in human-alignment of autonomous agents through safe, ethical, explainable, and interactive methods. Both were the first Australians admitted to the Future of Life Institute's Existential AI Safety Research Community in 2022. The collective's research team includes additional researchers and PhD students across Federation University Australia, Deakin University, and the University of New South Wales (UNSW), working on topics including multi-agent systems, fairness and trust in AI, human-robot interaction, human-machine collaboration, and safety knowledge transfer. Notable recent work includes a paper on multi-objective reinforcement learning as a tool for pluralistic alignment, which was accepted with spotlight status (top 3.5% of submissions) at ICML 2024 and presented at the Pluralistic Alignment workshop at NeurIPS 2024. ARAAChas received funding from the Survival and Flourishing Fund ($83,000 in 2023, via Jaan Tallinn) and Founders Pledge ($127,000 in 2024) to support research into multi-objective reinforcement learning applications for AI safety. The Founders Pledge grant funds a research assistant position to develop multi-objective methods for creating safer advanced AI systems. Grants are received through Federation University Australia as the fiscal entity.