AI & “Un-Personalised” Learning
Exploring the full potential of AI to improve human learning, beyond the 1:1 AI tutor
Hello learning friends, and happy new year! 👋
When it comes to the topic of AI and education, one thing I’ve noticed over the last year is that the personalisation of the learning experience is often considered to be the primary, most powerful use case for AI in education.
But 1:1 learning isn’t the only instructional strategy proven to be impactful on learning. It may also not be the golden bullet that we sometimes assume it to be.
In this week’s blog post we will look at AI from a different angle and ask: what are the pros and cons of using AI for personalisation? And what’s the potential impact of using AI to optimise and scale more connected, communal learning experiences?
Let’s go! 🚀
The Personalisation Pedestal
One of the most well-cited pieces of research on how humans learn is Bloom’s 2-Sigma research, which was published almost 40 years ago in 1984.
Put very simply, Bloom found that when taught 1:1 using mastery learning coaching techniques, learners performed two standard deviations better than those who we taught in a more traditional 1:many learning environment.
Building on the ideal of the personal “Socratic tutor” which stretches back to the Ancient Greek era, Bloom’s research invited us to idealise 1:1 teaching and learning - and we very much took him up on his offer.
Of course, personalised learning is not without value. As researchers since Bloom have found, personalised learning offers a range of benefits, including higher satisfaction levels, improved levels of inclusion and improved academic performance when compared with traditional 1:many classroom learning (Gallien, 2008).
As a result of our fascination with personalised learning, many of the technologies that have been built to scale learning have “productised” this pedagogical approach.
Some tools, like Socratic by Google, have been around for while. Others, like Khanmigo by Khan Academy, have leveraged the power of Generative AI to scale Bloom’s vision of 1:1 Mastery coaching with more power and potential than ever before. If recent investment patterns are anything to go by, more AI tutors will follow suit in rapid succession in the coming months and years.
While 1:1 tutoring tools like these will no doubt have for some value for some learners in some contexts, we should not let our fascination with personalised learning blind us to two important things:
the shortcomings of personalised learning for some learners, e.g. less confident learners, and learners from more communal-centred cultures;
the benefits of alternative pedagogical approaches, and the opportunity offered by AI to scale access to these alternatives.
The Power of “Un-Personalised” Learning
Communal learning theories emphasise the role played by collaboration, coordination, co-creation and co-operation in learning.
Communal learning experiences are “un-personalised” by design: they involve shared goals and group activities, and emphasise the value of joint human effort towards a shared understanding (Baker, 2015). Here’s a quick summary of how communal learning differs from personalised learning:
Approach to Knowledge Acquisition
Personalised Learning: Focuses on individual pacing and style, tailoring the educational experience to each student's needs and preferences.
Communal Learning: Emphasises collective knowledge building through group activities, discussions, and projects, encouraging students to learn from and with each other.
Learning Environment
Personalised Learning: Often relies on technology, such as adaptive learning platforms, to provide customised learning paths and feedback for individual learners.
Communal Learning: Creates a social learning environment where students engage in dialogues, debates, and collaborative problem-solving, fostering a sense of community.
Role of the Educator
Personalised Learning: Teachers act as personal tutors, focusing on identifying and supporting the unique learning needs of each student.
Communal Learning: Instructors play the role of a moderator or collaborator, guiding group dynamics and promoting cooperative learning strategies.
Assessment Methods
Personalised Learning: Utilises adaptive assessments that align with individual learning paths and measure personal progress.
Communal Learning: Often involves group assessments, peer reviews, and projects that evaluate collective understanding and collaboration skills.
Learning Goals
Personalised Learning: Prioritises meeting individual learning objectives and catering to personal interests and talents.
Communal Learning: Strives to build a collaborative learning community, fostering social awareness and a sense of shared responsibility.
There’s a large body of robust research to show that intentionally un-personalised, “communal-by-design” learning experiences deliver a number of significant positive outcomes for learners which are social and psychological as well as academic (Laal, 2012).
Woods and Baker (2005) discuss the importance of social networks in the learning experience, showing that students who feel socially connected report higher levels of learning and persistence in coursework (Woods & Baker, 2005).
Adams, Nicholson, Maciolek, and Biebel (2008) analysed the outcomes of learning communities, emphasising the value of shared experiences and mutual engagement for achieving positive social and academic outcomes (Adams, Nicholson, Maciolek, & Biebel, 2008).
TL;DR: while personalised learning has some benefits for some learner outcomes, the social interaction and connected aspects of communal learning are proven to offer similar academic benefits, as well as additional socio-cultural benefits for a broader range of students.
AI for “Un-Personalised” Learning
The next question is, of course: how could we use AI to scale the positive outcomes of “un-personalised”, communal learning?
Here are some initial ideas:
AI-Powered Cohort-Based Learning
AI algorithms could be used to analyse students' backgrounds, and performance data to create optimised learning cohorts.
AI-powered features like real-time language translation could be used to create and connect diverse groups.
AI tutors could be built to serve learning groups by stimulating discussion, answering questions, providing explanations and ensuring that all group members are on the same page.
Virtual Co-Creation “Maker Spaces”
AI-facilitated spaces could be built to enable virtual brainstorming sessions, group discussions & collaborative project work.
AI could be used to create groups with shared interests or challenges and suggest connections to foster collective learning and innovation.
AI could also be used to help generate resources to support ideation and to analyse and give feedback on the quality of ideas.
Peer-to-Peer Learning Networks
AI could be used to match learners with others who have similar interests and learning goals, enabling peer mentorship and knowledge exchange.
AI could be used to manage the logistics of peer instruction sessions, including organising discussion groups and timing activities.
Within Peer-to-Peer networks, AI could be used to immediate feedback to students based on their responses, analyse and give feedback on the effectiveness of peer instruction methods and assess changes in student understanding and behaviour.
AI-Mediated Role-Playing
AI can be used to generate scenarios and guide learners through them.
Learners could be allocated different roles in a historical or hypothetical scenarios and AI could coach the group to negotiate and make collective decisions, providing a dynamic learning environment that emphasises teamwork, leadership and decision-making skills.
Conclusion
In conclusion, the journey towards achieving tech-enabled educational excellence calls for a more nuanced understanding of a) the full range of pedagogical approaches that benefit human learning, and b) the potential of AI to help optimise and scale them.
Personalised learning is often hailed as a golden bullet for human learning, and frequently cited as a way to address educational inequities (Carlson et al, 2022). By tailoring education to individual needs, technology can potentially drive improvements in academic achievement for some learners.
At the same time, using AI to optimise and scale only personalised learning comes with considerable risks, including of inequity of impact and a de-prioritisation of valuable social and cultural outcomes which are better achieved through “un-personalised” learning. This includes many so-called “twenty-first-century-skills” like leadership, collaboration and social skills, which are essential to the future of the global economy.
One follow-up question I’ve started to research is: can AI replicate the dynamics and benefits of communal, connected learning? Put another way, what would be the impact if within the “un-personalised”, communal learning experience we replaced humans with AI?
More research on human-machine interaction in the context of human learning is required to be able to answer this question in full, but some interesting work has been done already:
A study by Dillenbourg & Self, for example, showed positive outcomes when a human learner was paired with a computerised co-learner to solve problems learn through collaboration (Dillenbourg & Self, 1992).
Meanwhile, in a similar test Kreijns et al. found that real human interaction was necessary for meaningful interaction and the achievement of outcomes in computer-supported environments (Kreijns, Kirschner, & Vermeulen, 2013).
Whether or not emerging AI technologies will be able to fully replicate AI-human interactions, is TBC. Regardless, one thing is clear: 1:1 AI tutors deliver only a small part of the pedagogical potential offered by AI.
If we want to leverage the full potential of AI to drive positive learner outcomes, we need first to go back to the research and ensure we have a full understanding of what optimal learning conditions looks like for all learners, then ensure that we leverage AI to serve them all as well an equally as possible.
Happy designing and innovating!
Phil 👋
PS: Want to learn more about AI and education? Check out my AI Learning Design Bootcamp and my monthly Learning Futures newsletter.