Projects

LUCID projects address any pure or applied research problem in which people are learning from machines (educational software, intelligent tutoring, second-language learning, MOOCs, etc), machines are learning from people (crowd sourcing, social network analysis, emotion recognition, natural language processing, etc), or human learning can be illuminated and improved through application of computational insights. The aim is to train scientists who can advance understanding within each core discipline by applying information and insights from the others, and who can bring the central ideas from each field to bear on real-world issues.

For example, there are several ongoing collaborative projects and more developing all the time. Current research themes include:

Machine teaching / intelligent tutoring: How can we leverage computational theories of human learning and theory and methods from machine learning to optimize lessons for human learners in different educational domains? Can cognitive theories of human learning be used to improve educational software or adaptive tutoring systems?

Adaptive crowd-sourcing and human/machine cooperation: Using insights from adaptive sampling / sparsity, can we develop new methods for measuring cognitive or perceptual structures discerned by individuals, or within special populations (infants, neurological groups), or across different demographic, cultural, or ethnic groups? Can we use these methods to better understand what elements of human learning and cognition are malleable through experience? Can we efficiently harness the power of the crowd to solve otherwise intractable computational problems?

New theories of human learning: Can theory and algorithms developed in machine learning provide concrete and testable hypotheses about how human beings learn, in both childhood and adulthood? Can such hypotheses lead to a new understanding of how and why human learning sometimes fails, or why people struggle to learn some kinds of content but excel at others? How do learners combine information from disagreeing sources? Can we create a mechanism for understanding the emergence and widespread persistence of false beliefs?

Signal discovery in neural data: Can we leverage insights from machine learning to develop new methods for decoding neural data? What do such methods suggest about how information is encoded in the brain, or how neural representations can be changed through learning and experience?

Advancing social robotics: Can we both leverage and advance our understanding of human social/emotional cognition by developing robotic systems that can interact socially with people? Can integration and emulation of social cues, including emotion and gesture, help artificial systems to teach human learners?