The interest of the society in applications and devices for the monitoring and increase of well-being of individuals is unprecedented. All major players in consumer electronics are developing or releasing devices and applications for monitoring stress, fatigue or other indicators of well-being.

Most of these approaches rely on specific hardware, such as hearth rate monitors and other physiological sensors or video cameras. While such approaches tend to be accurate, they also require specific actions from the user (e.g. users need to wear the sensor, users need to stop what they are doing and press the video camera for a few seconds).

In recent years a new approach is gaining form, for which Performetric is contributing significantly: one in which it is the behavior of the users, which is analyzed non-intrusively, that points out these indicators of well-being.

This exciting new possibility is becoming reality through the intersection of Behavioral Analysis and Artificial Intelligence, especially sub-fields of AI related to Machine Learning.

While Behavioral Analysis deals with the systematic analysis and recognition of one’s behavior in daily situations, Machine Learning deals with the development of algorithms that can learn from available data, with a strong component of pattern recognition and, afterwards, make predictions. The possibility is thus to collect behavioral data that can be used to learn users’ behaviors in given situations. This allows the later classification of the situation, by observing the behavior.

What we are dealing with is no more than what we Humans do naturally: we observe the others’ behaviors (e.g. body language, gestures, movements, intonation) and make assumptions on their state: someone that is restless, always moving, biting the nails, is most likely nervous. The challenge thus lies in the development of processes to collect this meaningful information, process it and convert it so that it can effectively be used by Machine Learning algorithms.

Here at Performetric we contributed significantly to this domain by developing interaction features that can be extracted from the use of the mouse and the keyboard. This is what makes Performetric so suited for being used in most of nowadays companies, in which employees work with the computer for long hours. Moreover, we also found out that when we are fatigued we interact differently with the computer. This allows to non-intrusively and continuously monitor the behavior of the user in terms of the interaction with the computer, in search for signs of fatigue.

Machine Learning techniques can thus be used to train personalized models, based on individual data, as well as general models, based on data from multiple users. Moreover, these models can be sensitive to the context of the user (e.g. the application being used).

Our aim is to develop hardware and software that is naturally user-sensitive and that places its focus on improving user well-being, with advantages for both employees, organizations, healthcare systems and the society as a whole, especially in times in which productivity and labor is so overvalued.