AI and Driving Telemetry: Screening for Cognitive Decline

Cognitive aging researchers are increasingly turning to everyday vehicles as diagnostic tools to detect the earliest stages of neurodegenerative diseases like Alzheimer's. By combining naturalistic driving telemetry—continuous data collected from an individual's daily driving routines—with artificial intelligence, scientists can identify subtle behavioral shifts long before clinical symptoms appear. Specific metrics, such as hesitation when making turns at intersections and lane-keeping entropy, serve as digital biomarkers. Machine learning algorithms analyze these complex spatial and temporal patterns to offer a non-invasive, continuous, and highly accurate screening method for early cognitive decline in older adults.

Driving is a complex task that requires the rapid integration of executive function, visuospatial awareness, memory, and motor skills. When mild cognitive impairment (MCI) or preclinical dementia begins to develop, these cognitive domains experience subtle degradation. While an older adult might easily pass a standard fifteen-minute clinical cognitive assessment, the continuous cognitive load of navigating traffic inevitably reveals micro-deficits in brain function.

To capture these deficits, researchers use naturalistic driving telemetry. Unlike driving simulators or road tests that induce performance anxiety and artificial behaviors, naturalistic telemetry relies on discreet, in-vehicle data loggers, GPS devices, and onboard diagnostics (OBD-II) ports installed in the driver's own car. These sensors passively record speed, acceleration, braking force, steering wheel angles, and global positioning over months or years, establishing a personalized behavioral baseline for the driver.

One of the most telling biomarkers extracted from this data is intersection turn hesitation. Navigating an intersection—especially making an unprotected left turn across oncoming traffic—demands rapid information processing, divided attention, and risk assessment. Telemetry reveals that older adults experiencing early cognitive decline exhibit measurable delays before initiating a turn. They may brake more erratically approaching the intersection, pause significantly longer at the stop line despite clear traffic, or demonstrate inconsistent acceleration once the turn begins. These micro-hesitations reflect a slowdown in executive functioning and working memory.

Another critical metric is lane-keeping entropy, which measures the randomness and irregularity of a driver's lateral vehicle control. Healthy drivers make smooth, predictable micro-adjustments to the steering wheel to stay centered in their lane. Conversely, drivers with early cognitive decline often display anomalous entropy. High lane-keeping entropy indicates erratic, unpredictable drifting and abrupt steering corrections, pointing to declining visuospatial processing and attentional control. Paradoxically, abnormally low entropy can also be a red flag, indicating rigid, over-cautious driving behavior where the individual struggles to adapt fluidly to road curvatures.

The sheer volume of telemetry data generated over months of driving is too vast for human analysis, which is where artificial intelligence becomes essential. Researchers feed thousands of hours of driving variables into machine learning models, such as Random Forests, Support Vector Machines, and deep learning neural networks.

These AI models are trained to recognize the multidimensional relationships between different driving behaviors. For example, an algorithm does not just look at a single instance of hard braking; it correlates lane-keeping entropy with intersection hesitation, time of day, weather conditions, and route familiarity. Over time, the AI detects deviations from the driver’s personal baseline and compares these shifts against the driving patterns of populations with known cognitive statuses.

Studies have shown that when AI models combine demographic data (like age and education) with naturalistic driving telemetry, they can predict preclinical Alzheimer’s disease and mild cognitive impairment with nearly 90% accuracy. Furthermore, the AI can notice a shrinking "life space"—a trend where the driver unconsciously avoids complex routes, highways, or night driving to compensate for their declining abilities.

Ultimately, integrating AI with driving telemetry transforms the automobile into an early-warning system. By passively monitoring intersection hesitation, lane-keeping entropy, and broader driving habits, researchers and clinicians can identify at-risk older adults years earlier than traditional methods allow. This early detection opens a crucial window for medical interventions, lifestyle adjustments, and timely conversations about driving safety and future care planning.