Intelligent agents that are:

• Complex: human-like intelligent behavior results from an interplay of various forms of reasoning & inference; such as perceptual & sensory inference, task planning, preference calucaltion, goal reasoning, analogical reasoning etc. We build agent architectures that implement such reasoning with AI & ML computational methods and integrate them in a single, cohesive, system.
• Collaborative: agents that can partner with humans need computational models of human decision making, problem solving, and learning. These models should be 'usable' by agents' decision making processes. We study how human models from psychology, cognitive science, behavior economics and other social sciences can brought into intelligent agent systems.

Agents here have been graciously funded by the US Government through DARPA, AFOSR, ARPA-E and other agencies. Read a detalied research statement for the vision behind CoCoA.