The future of learning

Episode Summary

Executive Summary: Candice Till argues that learning science is still emerging, and its biggest opportunity is closing the gap between research and real-world teaching by treating practitioners as collaborative researchers. She explains how technology and AI can capture learner behavior, personalize instruction, and improve both learning outcomes and research quality, while warning against outdated ideas like fixed learning styles.

Main Topics: The science of learning as an emerging interdisciplinary field (Priority: 5/5): Till defines learning science as the study of how people change their knowledge and capability, drawing from education, psychology, communication, computer science, and neuroscience. She emphasizes that the field is still nascent and not yet fully integrated into practice. The bi-directional gap between research and practice (Priority: 5/5): She argues that the problem is not only that research fails to reach classrooms, but also that practitioner wisdom rarely feeds back into science. Her goal is to create a two-way relationship between research and teaching practice. Technology as infrastructure for human learning (Priority: 5/5): Till frames technology not as a replacement for teachers, but as infrastructure that helps capture learner data, support instructional decisions, and improve learning design at scale. The Open Learning Initiative and evidence-based online learning (Priority: 4/5): She describes OLI as an early online learning system built around interactive practice, feedback, and data collection, designed to support learning rather than merely distribute content. Evidence from accelerated learning studies (Priority: 5/5): Till recounts a statistics study showing that students in an OLI format learned as well or better than traditional students in less time and with fewer instructor contact hours, demonstrating the potential of well-designed learning environments. AI and generative AI in learning environments (Priority: 5/5): She distinguishes predictive AI from generative AI, arguing that predictive models can personalize learning while generative AI improves communication, explanation, and content creation for learners and instructors. Debunking learning styles and emphasizing multiple representations (Priority: 4/5): Till rejects the popular learning-styles theory as unsupported, while noting that multiple representations of content can still strengthen learning.

Key Arguments: Learning science is still early-stage, so there is substantial room for discovery and improvement in how humans learn. The biggest barrier is not lack of knowledge, but the weak translation between research and practice in both directions. Teachers and practitioners should be treated as collaborators who generate valuable observations, not just implementers of research. Technology can capture learner actions and context in ways humans cannot manage alone, enabling better instructional decisions. AI should support, not replace, teachers by helping personalize instruction and provide feedback to learners, instructors, designers, and researchers. The Open Learning Initiative showed that interactive, feedback-rich online learning can produce strong outcomes even with less time and fewer instructor hours. Generative AI adds value by making learning more conversational, interpretable, and accessible, especially for creating and refining learning artifacts. The learning-styles theory is not supported as a general rule; multiple modalities help, but preferences are not fixed categories.

Data Points: Open Learning Initiative start year: 2002 - Till says OLI began in 2002, before the generative AI era. Traditional statistics course length: 15 weeks - In the accelerated learning study, the control group took the standard introductory statistics course over 15 weeks. Traditional class meetings: 4 per week - The traditional statistics condition met four times weekly. OLI course length: 8 weeks - The OLI condition completed the same statistics course in half the time. OLI class meetings: 2 per week - The OLI condition met twice weekly. Instructor contact hours reduction: About one-quarter of traditional contact hours - Till notes the OLI condition had roughly a quarter of the instructor contact hours of the traditional course. Performance on midterms and finals: As well or better - Students in the OLI condition matched or exceeded traditional students on shared course exams. External statistics knowledge gain: 18 points better - OLI students scored 18 points higher than traditional students on an external measure of statistics knowledge. Amazon workforce scale: 1.6 million employees - Till cites Amazon’s need to rapidly upskill a very large workforce. Leave of absence from Stanford: 2 years - She says Stanford leave policy was two years, which led to her resignation before later being rehired.

Pivotal Quotes: "learning is not a spectator sport" — Candice Till: She uses this to explain why active practice and feedback are essential to learning. "the science of human learning is at the start of a scientific revolution in understanding human learning" — Candice Till: Her rapid-response takeaway about the field’s future and openness to broad participation. "we can observe the learner just as you said, like Netflix does, like Amazon does" — Candice Till: She explains how learning platforms can use data to personalize instruction and understand learners.

Implications: The future of education will likely be more data-rich, adaptive, and collaborative, with AI helping teachers and learners make better decisions. Institutions that connect research, practice, and technology may improve outcomes and broaden access to effective learning.