Status
As of 31 July 2026, we are conducting a proof of concept (PoC) in Japan with multiple partners, including a university, to validate and refine the model using additional driving behaviour data, cognitive assessment questionnaire data, and driver personality profiles. We aim to evaluate its predictive performance within this fiscal year.
The model is expected to help reduce motor insurance loss ratios, improve insurance risk assessment, and enable new preventive services, such as AI-driven early detection of cognitive decline followed by interventions that encourage drivers to seek cognitive health assessments.
Detecting cognitive decline from driving data
Summary
Approximately 30% of Japan's population is aged 65 or older, and it is estimated that one in three older adults experiences cognitive impairment, including mild cognitive impairment (MCI).
This project aims to analyse everyday driving behaviour, correlated with insurance claims data, to spot warning signs that may be linked to cognitive decline in older drivers, thereby supporting safe driving among older adults and reducing accident risk.
The problem
As lifespans rise, more people are using the roads well into old age, making traffic safety among elderly drivers an increasingly important issue. In some cases, accidents involving elderly drivers may be linked to underlying cognitive decline, though this connection is not well understood. Cognitive decline is notoriously difficult to notice in oneself, and a formal diagnosis requires a clinical assessment that most people don't undergo until symptoms are already advanced. At present, cognitive decline may not be identified until after a serious accident has occurred.
Preventing such accidents requires early detection of these changes, but continuous cognitive assessment is far from straightforward. We want to surface it earlier, while there's still time to intervene.
Solution
Our working hypothesis, still under validation, is that early signs of cognitive decline surface in everyday driving behaviour well before they escalate into something more serious, such as an accident. We have developed an AI model that uses dashcam footage from traffic accidents provided by an insurance company to identify accidents associated with cognitive decline, learns the characteristics of the drivers' driving behaviour, and identifies other drivers who exhibit similar driving patterns.
The ultimate goal is to enable early identification of changes in cognitive function from everyday driving behaviour, reduced accident risk and contributing to safer mobility for older adults, and greater peace of mind for their families.