News

 

New Frontiers with Smart, Adaptive Interfaces: Building Machines That Know About People

February 14, 2019

What if your car knew when you were too tired to drive?

Dr. Leslie Blaha’s research tackles questions like this. Her innovations center the idea that applying knowledge of how the mind works can transform people’s relationships with technology.

Thus, the places people interact with machines – interfaces – are a key target of Blaha’s work. Effective technology takes into account psychological principles, such as what people will notice and remember. Could behavioral science be more explicitly and deeply integrated into the design of machines?

Blaha pursues smart, adaptive interfaces that do just this, bringing the knowledge that cognitive scientists have accumulated to building machines that better support and anticipate people. Her work requires that she move between basic science and applied fields in interdisciplinary teams of scientists and engineers, a unique position in the brain and behavioral sciences.

“Applied problems can drive basic science, and basic science advances can change the way we deal with applied problems. But across the board, we’ve got to work in these cross-disciplinary teams.”

This flexibility was hard-won, Blaha notes, and requires being able to speak the language of each field. But despite these initial barriers, such collaborations are necessary for solving today’s hard problems. “Keep having those discussions! It takes a lot of talking to even get the ideas off the ground.”

Blaha’s current role is a senior research psychologist, working within both the Air Force Research Laboratory and Carnegie Mellon University’s Psychology Department, a position that allows for groundbreaking cross-disciplinary innovation.

One ongoing project involves developing real-time feedback for detecting when people are too fatigued to do critical tasks, like operate a vehicle. An application of this research is for those whose jobs require long, demanding shifts: first responders, for example, are involved in more car accidents after working. Eye-trackers could detect when someone is beginning to fall asleep. However, there are earlier signs of fatigue, such as changes in how long it takes someone to react. An interface that could detect such changes in a sophisticated way could save lives.

Another ongoing project examines how to improve retraining schedules for critical skills, which are typically offered on a one-size-fits-all model. But equating, for example, the CPR re-training needs of an emergency room clinician and the needs of staff members in less emergent departments is not ideal. A smart, adaptive system could take into account the cognitive science research behind expertise and skill learning, modeling factors like how frequently a skill is used and how well the skill was first learned. Blaha is currently extending this idea, working with law enforcement departments to develop such systems for firearm safety training.

Research initiatives like these mean that Blaha is comfortable with many titles. “Some days I’m a mathematical psychologist; some days I’m a cognitive scientist. I have been a human-machine teaming expert and an applied mathematician. I tend to think of myself as a mathematical psychologist and cognitive modeler.” Regardless of label, Blaha finds that her toolbox of computational and mathematical approaches is fundamental to studying people in creative, interdisciplinary ways.

A world where people solve problems with the aid of smart and adaptive interfaces not only needs machines that are more knowledgeable about human cognition. It also requires people develop useful and accurate beliefs about how machines work. One way to do this is involves intelligent tutoring systems for people and machines, one of the projects Blaha is most excited to dive into at Carnegie Mellon.

Why might people need training on what smart and adaptive technology can do? The interfaces Blaha envisions are not driven by turning gears or combusting engines. They rely on mathematical modeling and probabilistic prediction, and unlike mechanistic systems, are not always right or predictable in the way that people are used to. “If I’m going to have these adaptive systems supporting people, people need to understand what the system is trying to do and what it’s capable of. How can we train humans and machine to understand each other’s capabilities?”

What’s next on the horizon? Blaha says there is much work to be done to learn more about how people solve complex problems. These models could then be fed back into technological systems. “We’re good at supporting simple tasks, like ‘organize my email.’ What about more complicated tasks, like ‘organize my workflow’? We need to keep advancing what we know about human cognition.”


Leslie Blaha is a recipient of the 2018 Federation of Associations in Behavioral & Brain Sciences (FABBS) Early Career Impact Award and was nominated by the Society for Mathematical Psychology.

View All Articles