Research Thrusts
AI-ALOE will conduct both use-inspired AI and foundational research into AI, which are closely related and mutually reinforcing.
A research thrust for use-inspired AI is the development of AI teaching and learning assistants that enhance cognitive, teacher, and social presence in online adult learning to help make it efficient and effective. Briefly, by efficiency of teaching, we mean enhancing the reach of a teacher to a larger number of learners without reducing learning efficiency or effectiveness. By efficiency of learning, we mean time taken to master a lesson or a skill; and by effectiveness of learning, we mean mastery of a lesson or a skill.
Research on AI in education since the 1970’s has developed a complementary strategy: use of data about the learner and the learning to personalize learning. Now the question becomes whether we can generalize personalization to online adult learning that includes more complex and often ill-defined, open-ended problems. This suggests a second research thrust for use-inspired research into AI: learning analytics for personalization of large-scale online learning of interest in adult education.
The ALOEInstitute will address four themes fundamental to AI:
- Cognitively-Grounded AI: Building on current understanding of human cognition and learning, we will develop novel virtual assistants to support cognitive strategies known to underlie effective adult learning such as active learning, learning by doing, learning by reflection, and self-directed learning.
- AI-Based Personalization at Scale: We will develop new architectures for organizing large scale data on learning. We will develop standards for data collection, use, sharing, and distribution. The scale of this data and the diversity of learners will enable more robust generalization of learning behaviors and outcomes, and thereby ensure broader impact. The learning analytics will also support personalization of learning at large scale.
- Human-AI Collaboration: We will develop novel interactive visualization techniques for helping teachers build a theory of mind of learners, thus enabling teachers to ascribe goal and knowledge states to learners. We want our virtual teaching and learning assistants to be teachable. Thus we will develop techniques of machine teaching where teachers will be able to interactively train the virtual assistants using their own preferences and their own data.
- Responsible AI: We will use the methodologies of participatory design and human-centered design to develop sociotechnical systems in which humans and AI agents work together for long durations. Participatory design helps makes tacit human knowledge explicit and invisible social processes visible, which is important for building ethical AI techniques that actually are beneficial in practice.
Research Design
The research design in our work derives from the theoretical foundation of feedback loops in human learning. Although some learners are very good at self-directed learning, most learners need customized guidance and support to acquire the required knowledge, skills, and dispositions. Effective guided learning engages positive, data-driven feedback loops, such as a teacher-centered loop and a learner-centered loop.