The project with 99P Labs is a two-year effort to develop innovative methods for designing AI assistants. Any assistive technology’s usefulness is determined not only by its intelligent capabilities but also by how well it interacts with human users. In other words, for any technology to be used, its benefits must outweigh the effort required to use it. Specifically, this project aims to develop formal methods for determining interaction designs to create AI assistants to help drivers replan in situ (i.e., while traveling).
The two-year effort consists of:
- Understanding the Challenges of Replanning – ethnographic research to understand what types of problems require replanning and strategies for creating new plans while driving,
- Modeling Decision Making by representing strategies with concepts from graph theory,
- Designing Human-AI Interaction by using the decision-making model to create AI functions and human-AI interaction functions, and
- High Fidelity Driving Simulation to test such interaction designs in a realistic simulated driving scenario at the Ohio State University Driving Simulation Laboratory.
In our experiments, participants act as rideshare drivers facing real-world challenges such as changing traffic conditions, low fuel levels, and unexpected detours. Their interactions with the AI assistant provide valuable insights into what type of AI support is helpful – and what isn’t. Although the project aims to test AI technologies in a driving context, the methods developed for it are not constrained to a single domain. Instead, they can be adopted for many domains in which AI technologies can provide benefits, such as aviation, space operations, disaster robotics, and medicine.
Publications
- Kannally, C., Smith, J. R., & IJtsma, M. (2023). Human-AI Teaming in the Automotive and Mobility Industry: Guiding Design to Support Joint Activity. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 67(1), 111-116. https://doi.org/10.1177/21695067231192223