Artificial Intelligence is becoming a crucial companion on our lifelong learning journey. AI-based learning holds great promise: personalized learning for everyone. But will AI be able to meet the needs of each learner? Will it give feedback that promotes progress instead of discouraging learners? Creating EdTech that provides meaningful, personalized feedback requires close collaboration between humans and AI.
Early in the COVID-19 outbreak, while taking a break from working at home, I had a FaceTime chat with a friend. She was worried about Maya, her 13-year-old daughter, and how school was going. “I wonder whether she’s learning at all,” my friend said, while Maya sat in the next room, participating in remote schooling.
A few weeks later, it became apparent that my friend didn’t need to worry. Maya’s resourceful teacher had found a way to combine whole-class remote meetings and one-on-one support with innovative educational technology (EdTech) tools. Maya, for her part, was proving to be a flexible and robust learner, adjusting to the changed conditions as she continued to learn. The road seemed bumpy at times, but she managed to stay motivated, and this was reflected in her progress.
Clearly not all learners can count on such favorable conditions; many schoolchildren lack the support they need from their teachers, their parents, or the school system. For these children, the current circumstances are even more difficult to navigate. The COVID-19 pandemic is just one example of the changing conditions confronting learners. Profound shifts in society, the economy, and policy have had a significant impact on learning. Today’s learners need to design their own lifelong learning trajectory, and they’re expected to keep up in a world that demands skills such as complex problem solving, analytic reasoning, innovative thinking, and teamwork.
The science of learning explores the promise of tech and AI
Researchers in the learning sciences are looking at how to design environments that will facilitate the learning of students like Maya. To that end, they are investigating what goes on in students’ brains as they learn, integrating findings from psychology, cognitive sciences, sociology, and educational sciences, as well as artificial intelligence and educational technology. Although much remains to be discovered about various aspects of learning, there’s general agreement about a few elements.
For instance, there is a consensus that each learner is different, with a unique set of character traits, skills, previous knowledge, and cultural and family background that influence learning. Instructors as well as designers of learning environments need to take these things into account.
Adaptive learning (often mentioned in the same breath with adaptive teaching) involves using AI to interact with the learner and deliver personalized resources and learning activities. AI can integrate enormous quantities of data, and thanks to machine learning, it can apply these data in a beneficial way for each learner — but only if it is properly trained by human beings who know about the mechanisms of human learning.
When designing adaptive learning environments, the perspective of the instructor is also important. What makes Maya’s teacher a good teacher? John Hattie’s research on “visible learning,” for example, provides important insights into the various factors that influence students’ achievement. As they train AI, instructional designers will be working with tried and tested methods and integrating them into new or existing e-learning tools. AI must be able to adapt to a variety of methods and learning settings.
Let’s talk about meaningful feedback
Another aspect that many researchers from the learning sciences agree on is this: While students may benefit from and prefer very different learning settings, they all tend to learn best in an active, engaged, meaningful, and interactive environment. Interactive means not only that they are able to share their process with other students and with instructors, but also that they receive guidance — and feedback.
During the period of remote learning, Maya’s mother was surprised that her daughter seemed unbothered by the fact that she wasn’t receiving as much human feedback as she had during in-person instruction. “If she’s simply told whether she was wrong or right, how will this actually help her learning?” Maya’s mother wondered — and that’s exactly the point!
Instead of providing a formulaic response such as “Wrong; try again!” or “Correct, you’re doing great!” feedback should refer specifically to the student’s answer and be given in a manner that is motivating and contributes to further progress. This type of feedback is often called formative feedback (or formative assessment).
Feedback that makes learners eager to continue learning
Researchers emphasize the importance of adapting feedback to individual learners and environments. As Valerie Shute, who focuses on the design, development, and evaluation of advanced systems to support learning, points out, “Feedback is significantly more effective when it provides details of how to improve the answer rather than just indicating whether the student’s work is correct or not.” She goes on to identify certain essential characteristics of formative feedback: It should be nonevaluative, supportive, timely, and specific. It should have a clear intended message, and it should be adapted to the learner’s level and academic motivation.
When Maya learns a new math concept, she might initially need more detailed feedback and more hints (referred to as “scaffolding” by learning scientists), but as she progresses, she may require less support and become a more self-regulated learner. Even then, however, she will need immediate, specific feedback showing where she stands in her learning process and what she needs to work on. By offering specific details (“But why was this the wrong answer?”), feedback can prevent cognitive overload and help her stay motivated.
Tasks assigned to a learner should be challenging but not impossible, and task-specific feedback should demonstrate that skills can be developed and are not simply a question of “talent.” While it’s important to give feedback right after a task has been completed, learners also need long-term feedback that highlights their strengths and weaknesses, provides an overview of their progress, and identifies specific areas for the learner to work on. Ideally, feedback should make learners eager to continue learning, rather than focused on achieving a certain level of performance.
Only humans can teach AI to give meaningful feedback
Whether AI is successful in facilitating Maya’s and other students’ learning journeys depends greatly on its ability to provide meaningful feedback. AI underlying an EdTech tool has to gather information not only about learners’ prior knowledge and skills, but also about their objectives, their motivation, and even their affective state. Is Maya by nature a curious person? Is she easily distracted? Is she intimidated by math?
AI is unequalled when it comes to working with big data, and thanks to machine learning, it can apply these data in a beneficial way for each learner — but in order to create something meaningful, it will need to collaborate closely with human beings.
It must learn from people who know, from experience, how different students tick, people who are familiar with pedagogical concepts, and people who are contributing to the growing body of research in this field. Of course, it is also important for AI to learn to collaborate with human teachers as they seek to make blended learning as effective as possible. On top of that, AI needs to respect the privacy of our data. And finally, and importantly, the new solutions must be provided to everyone, especially to learners who are underprivileged or struggle with learning difficulties.
AI will be with us throughout our lifelong learning journey. We must take active, deliberate steps to shape this new learning partner in a way that benefits individuals as well as society — teaching it how to teach us, and teaching it how to give us the right feedback.
Research, literature, videos and articles that inspired this piece
Heidi L. Andrade (2013), Classroom assessment in the context of learning theory and research, In: SAGE Handbook of Research on Classroom Assessment
Hanna Dumont (2019), Learning Science Adaptive Teachers (video); Global Education & Skills Forum
Edited by Hanna Dumont, David Instance, Francisco Benavides (2010), The Nature of Learning: Using Research to Inspire Practice; OECD Centre for Educational Research and Innovation
Emily R. Fyfe, Bethany Rittle-Johnson (2016), The benefits of computer-generated feedback for mathematics problem solving; Journal of Experimental Child Psychology
John Hattie, Visible Learning (books, videos, articles, website)
Kathy Hirsh-Pasek et al. (2015), Putting education in “educational” apps: lessons from the science of learning; Psychological Science in the Public Interest: A Journal of the American Psychological Society
Sarah Hofer and Lennart Schalk (2021), Das individuelle Lernen unterstützen: Formatives Assessment, In book: Professionelles Handlungswissen für Lehrerinnen und Lehrer (Greutmann, Saalbach, Stern); Kohlhammer
Amy Ogan, Personalized learning — a data revolution (2017); BOLD — Blog on Learning and Development, Jacobs Foundation
Kaili Rimfeld et al. (2019), Teacher assessments during compulsory education are as reliable, stable and heritable as standardized test scores; The Journal of Child Psychology and Psychiatry,
Valerie J. Shute (2008), Focus on Formative Feedback; The Review of Educational Research
Dylan Wiliam (2016), Feedback on learning (video clip); National Improvement Hub (Education Scotland)