Prevent car damage with CARMA
At MyWheels, we ran into a difficult problem. Despite existing measures, we noticed that a small group of users was engaging in reckless driving, sometimes resulting in serious damage. We were often only able to intervene after the damage had already been done. Thatโs why we set to work on a targeted algorithm: ๐๐๐ฅ๐ ๐ (๐๐ฎ๐ฟ ๐ ๐ถ๐๐๐๐ฒ ๐๐น๐ด๐ผ๐ฟ๐ถ๐๐ต๐บ).
It analyzes objective driving behavior signals that exceed set thresholds and identifies patterns. Based on this, a risk score is generated, after which an employee manually assesses the appropriate next step, such as issuing a warning.
Because this affects the trust of our users, we choose to communicate openly and clearly about it. We explain in multiple places how driving behavior monitoring works, what CARMA does, and what this means for users. Access to event logs, filing objections, and requesting a reassessment are always possible.
The algorithm has been thoroughly assessed through a DPIA and complies with the provisions of the GDPR and the AI Act.
Since its implementation, we have seen a significant decrease in claims, which underscores the importance of this approach. To me, CARMA demonstrates how to use algorithms responsibly in a product: in a targeted, factual, and explainable way, with the user at the center.