A Toolbox for Human-Centered AI Engineering Education

Human-centered AI learns from human signals and behavior and continuously improves from human collaboration while providing a tailored, effective experience between humans and machines. It is a highly complex interdisciplinary engineering system that spans various academic disciplines of engineering, science, and social sciences, including embedded systems, sensors and communication networks, software engineering, cybersecurity, big data and artificial intelligence, physics, and human factors.

Human-Centered AI Engineering is increasingly needed in industries including manufacturing, health care, education, entertainment, and expressive technologies. Imagine a robot in a manufacturing line collaborating with a worker, not only following instructions but having access in real-time to the cognitive and affective status of the worker to increase safety and enhance productivity –what should a robot do with a stressed, unfocused, or frustrated worker? How to implement intuition capabilities into the robot? Imagine an online tutor tracking your gaze and therefore knowing what you are looking at or what you are reading; moreover, inferring your cognitive challenges and maybe offering help without the user having to ask for it. Imagine software or machines are aware and driven by emotions (engagement, stress, interest, fear, meditation, or frustration), thoughts, body language, and facial gestures, among other interactive and expressive technologies.

To develop these Human-Centered AI systems, highly qualified engineers are required who possess interdisciplinary knowledge. However, a gap exists between the industry’s expectations and graduates’ competencies. It is crucial to create opportunities for students to learn by doing Human-Centered AI Engineering having hands-on experiences with emerging sensors and devices and challenging them to develop complex software able to show not only intelligent but empathetic decisions complex software that involves data collection from diverse sources, robust and evolvable physical infrastructure, run-time machine learning models, and purposeful requirements engineering that considers ethics and fairness as well to create systems.

Funding

This work is supported by the Teacher-Scholar Mini-Grant (TSMG) program awarded by the Division of Research, Economic Development & Graduate Education at California State University.