By Jeff Dunsavage, Head of Research Publications and Insights, Triple-I
Artificial intelligence throws a risky curve into insuring autonomous vehicles.
AI models for AVs that are trained on typical driving scenarios “can potentially misjudge unusual situations or conditions,” Patrick Schmid, Triple-I’s chief insurance officer, recently told Digital Insurance. “When that vehicle is going to make a bad call, it’s hard to know why the AI made that choice.”
Autonomous vehicles rely on AI for three layered functions:
- Perception (interpreting camera, LiDAR, and radar data);
- Prediction (anticipating behavior of other vehicles, cyclists, and pedestrians); and
- Planning (selecting the vehicle’s next maneuver).
“Each of these layers is a distinct AI model with its own failure modes,” Schmid said. “Perception could have a misclassification of what it sees. Prediction could misjudge another road user’s movements or intent. Planning to select a safe maneuver can be based on a flawed read of what’s occurring on the road.”
Because insurers don’t necessarily have a view into the AI systems decisions and data logs, Schmid said, “Pricing becomes a little bit harder than with conventional driving, or even telematics.”
Schmid predicted that insurance policies for AVs will have to evolve to reflect these uncertainties. He emphasized the importance of staying on top of regulatory developments and pointed to the National Association of Insurance Commissioners’ Model AI Model Bulletin on AI Use by Insurers and an AI evaluation initiative.
Learn More:
Background on: Self-driving cars and insurance
AI Adoption Outpaces Governance, Nationwide Finds
AI Efficiency Gains Pave Way for Insurance Affordability
