Most fitness and Health Tech wearables start with a defined problem, then identify an objectively measurable biometric that serves a specific purpose related to the problem. For example, for hypertension, measure blood pressure. Concerned about heart health? Monitor pulse, carotid artery rise and fall, and perhaps heart rate variation. Need to lose weight? Build a wearable that counts steps, monitors sitting duration, or counts calories. So the path is clear: find a problem, define an objective biometric, make a device with software to measure, monitor, and report what was measured, and you’re home free. If you can make a device that measures accurately and consistently, sell it. The conventional path to health and fitness wearable riches is pre-defined. But now there is another way.

Scientists at the University of Sussex have a sensor-packed smartwatch with an algorithm that tracks everything it can, with no prior instructions. The wearable discovers what you do via machine-learning and then asks if you want to track the behavior or biometric measure. According to Sussex inventor Hristijan Gjoreski, “Current activity-recognition systems usually fail because they are limited to recognizing a predefined set of activities, whereas of course human activities are not limited and change with time. Here we present a new machine-learning approach that detects new human activities as they happen in real time, and which outperforms competing approaches.”

It’s still early days for the Sussex group. Their research will be published this month at the International Symposium on Wearable Computers in Hawaii. The most exciting potential of a self-learning wearable is the discovery of new significant variables have direct bearing on health. We can imagine a device jam-packed with every conceivable sensor observing a person for a while and then reporting thus, “Here’s what happened. Which of these activities or measurements matter?” With aggregate information from larger populations over time, heretofore unexpected relationships between problems and activities or biometrics may come to light.