Best Paper Nominee at HAI 2024 and AAAI Demo

Congratulations to Ph.D. student Sheikh Mannan and former SIGNAL Lab undergraduate researcher Paige Hansen! Our paper, Combating Spatial Disorientation in a Dynamic Self-Stabilization Task Using AI Assistants, received a nomination for Best Paper at the International Conference on Human-Agent Interaction 2024. This paper was written in collaboration with the Department of Psychology and Neuroscience at Brandeis University. In it, we use a high-throughput, high-fidelity simulation of inverted pendulum balancing under spatially disorienting conditions and train a variety of AI models that can perform the task to various different abilities, and also provide cues to humans performing the task in order to enable them to take better actions. We conducted a high-volume study using digital twins and then a human subject study in which 20 subjects and 5 of the best performing assistant models engaged in human-in-the-loop training of the assistants. Our findings indicate that AI assistance can help people perform substantially better at the task, but that assitance who have a more human-like mode of operation are less effective yet more trusted and preferred by humans.

Congratulations to Mannan and Paige!

The link to the proceedings can be found here. The platform we developed for this study will also be presented as a demo at AAAI 2025 next February in Philadephia!

Citations:

• Mannan, S., Hansen, P., Vimal, V.P., Davies, H.N., DiZio, P., and Krishnaswamy, N. (2024). Combating Spatial Disorientation in a Dynamic Self-Stabilization Task Using AI Assistants. In International Conference on Human-Agent Interaction (HAI). ACM.
• Mannan S. and Krishnaswamy, N. (2025). Bidirectional Human-AI Learning in Real-Time Disoriented Balancing. To appear in AAAI Conference on Artificial Intelligence (AAAI): Demos Program. AAAI.