The AIVAS Lab does research at the intersection of artificial intelligence and cognitive science, in the area of computational cognitive systems. Most of our research involves studying how visual mental imagery contributes to learning and intelligent behavior, both in humans and in AI systems.
(An exercise in gradient descent)
Most of the research that we do follows two main pathways. First, we build and study AI systems as a way to understand how people think. Much of our work uses AI to model learning and reasoning in cognitive conditions such as autism, so that we can help find better solutions for communication, education, and assessment for people with these conditions. Second, we use what we are learning about human cognition to advance the state of the art in AI. We focus primarily on new AI techniques for using visual mental imagery to solve complex problems.
What are Visual Analogical Systems?
The term analogical means that something has an organized relationship with something else (like an analogy). In AI and cognitive science, analogical representations refer to a way of portraying information that retains a correspondence with the real world. For example, an image of a cat is analogical because the 2D spatial information contained in the image corresponds to what the cat looks like in the real world. On the other hand, the word “cat” is not an analogical representation. Most of the AI systems that we build use visual analogical representations as the core data structures that support learning, problem solving, and other intelligent behaviors.