Learning Sciences

Research Lead: Ben Nye

The University of Southern California Institute for Creative Technologies is a University Affiliated Research Center (UARC), whose research develops emerging technologies (e.g., AI, ML, VR/AR) to research their impact on training, engagement, behavioral health, and related outcomes.

ICT research typically involves multiple groups working collaboratively, to leverage their strengths. The ICT Learning Sciences group is focused primarily on generalizable and reusable frameworks for leveraging AI for personalized learning and knowledge transfer.

Below, we describe some of these frameworks and note some of their unique capabilities.

Other Relevant Projects

In addition to the projects above which are most relevant to typical training needs, a number of other projects at the research prototype level (6.2) may also be relevant to certain requirements. These will be briefly noted below.

AR/VR Collaborations

In collaboration with the ICT Mixed Reality (MxR) group, we have conducted research on Augmented and Virtual Reality (AR/VR) learning. These include:

● Tar AR (NSF): An AR experience with the La Brea Tar Pits, where learners walk in prehistoric environments. Uses randomized controlled trials to study the impact of visual immersion (headsets) and interactivity level on learning and engagement.

● OMEGA (AFRL): Machine learning metrics for engagement and distraction during Pilot Training Next (PTN) flight simulations. Evaluated a recommender for instructor interventions.
A collaboration with Eduworks and the ITSEC 2021 Best Paper in the Training track.

● VAST: Visual Abstraction for Synthetic Training (Army): Developing research designs to study how shifting representation fidelity (e.g., voxels, wireframes) affects training outcomes.

Engagement Data Mining

We also use machine learning to measure and increase engagement in computer-based learning, such as:

  • ENGAGE (Army): Multi-modal analysis of engagement based on facial video feeds, behavioral logs, and self-report surveys. Developed Reinforcement Learning (RL) models to increase learning through coaching.

SMART-E (Army): Semi-supervised machine learning to classify different types of engagement/disengagement, by using mostly unlabeled training data plus a small number of play-testers who act out engagement types. Build a general purpose approach valid for many types of learning environments.

Agent-Based Intelligent Tutoring Systems (ITS)

We also have two ongoing collaborations for virtual agent tutoring systems:

  • ElectronixTutor (Navy): Adapting and expanding content for a dialog-based ITS to cover Navy Nuclear Power Fundamental Electronics skills. Running a pilot of approximately 500-600 sailors with the Navy Nuclear Power Training Command (NNPTC). A collaboration with the University of Memphis Institute for Intelligent Systems.
  • Advancing Teachers Proportional Reasoning (IES): Teacher professional development for math delivered by a virtual facilitator, who can provide feedback and also control virtual manipulatives such as interactive graphs, calculators, and tables.

A collaboration with the USC Rossier School of Education, University of Georgia, and UMass Dartmouth.