Human-Centered AI: Neuromorphic Computing, Human-AI Decision Making, and Agentic-AI

The question

There are four high-level primary questions. (1) How can neuromorphic computing aid user interaction and decision making? (2) How is decision making affected as a user goes from a frontier LLM to a smaller model, and everywhere within that spectrum? (3) Agentic AI is set to become pervasive in many jobs. How does its use affect our cognitive load, and how can we improve the interfaces? And (4) how can neuromorphic computing and other AI paradigms help the user have a personalized interaction? This last one is especially important for accessibility.

How we work

We have started most recently in this area. For example, Beyond the Wizard of Oz (opens in new tab) , an IEEE VR publication (under the journal IEEE TVCG), shows what happens when real-time machine learning is accurate enough (80% or higher) yet makes mistakes. We validated the work conducted for years by the human-factors community, now using real-time machine learning. Yet it was not only validated — the automation bias was worse than we believed.

Currently, we have students working on the questions above to solve different critical human-centered AI challenges.

What this work enables

We hope that our work will enable the improvement of the use of AI, in particular under low-powered and constrained environments.