Everyone feels stressed at work from time to time. What if, instead of turning to your co-workers, family, or friends for support, you could turn to a machine?

That’s a question researchers at MIT Lincoln Laboratory are seeking to answer by developing artificial intelligence (AI) systems that can sense when a human’s mental fatigue is causing a drop in performance.

The researchers suggest that AI machines could intervene and offer help and guidance before things spiral out of control by monitoring their human counterparts. As Tactical Systems Specialist at the MIT Laboratory Michael Pietrucha says, “teamwork is a two-way street.” While many researchers focus on how humans can monitor the machine's health and quality, Pietrucha suggests that it is just as crucial for machines to monitor the humans they work with.

Rise of the Machines?

The idea of machines monitoring humans for signs of fatigue or cognitive decline is not new. For decades, researchers at MIT and elsewhere have been interested in how technology can be used to “read” a person by understanding their emotional and mental state at different points in time.

Researchers have collected and processed biometric data, like recordings of people speaking on audio and video. By putting these data through AI algorithms, researchers have been able to identify biomarkers that signify different psychological conditions. Models have been trained, for example, to estimate a person’s level of depression with a high level of accuracy.

The natural next step is to use the information already available to researchers to train AI machines to intervene when a human displays specific characteristics of depression, cognitive decline, or stress in the workplace. MIT researchers will build a cognitive model of a human that can be used to analyze a person’s mental state. This will include how stressed, overloaded, and fatigued a person feels. Biomarkers will be derived from multiple data sources, including facial recordings, EEG, heart rate, and eye movement.

The model will represent a baseline of “normal.” From there, trained AI machines will be able to tell if a person strays from “normal.” Further, the AI machines will be able to identify the risks associated with deviations from normal, including the likelihood of making a mistake or a decrease in productivity.

How Your AI Teammate Might Help

Ok, so if your AI teammate can detect when you’re feeling overwhelmed and stressed, how, exactly, will they step in to help?

There are many options here based on the predicted risks and assumed levels of stress. For example, the AI may suggest taking a break to go outside, get some fresh air, or walk to a nearby coffee shop for a pick-me-up. Depending on the severity of stress, the AI could also suggest something more permanent, like a shift change or even transferring tasks to someone else.

In some cases, the AI might make quick decisions when their human partner cannot. MIT uses the example of an AI deciding to eject an unconscious fighter pilot when the pilot is unable to make the decision.

Conclusion

The possibilities are limitless when it comes to how machines can benefit the humans they work with. By comprehending what a person is feeling, even though they cannot feel emotions themselves, machines might be able to offer more comfort and guidance than our human co-workers ever could.