Finding the personal in personalised learning
Technology, social obligation and the two sigma problem
In 1984 Benjamin Bloom showed that one-to-one mastery tuition improves learning outcomes by two standard deviations. In simple terms, an average student from any given class who is taught one-to-one using mastery learning, could outperform 98% of their classmates.
How to achieve this level of improvement at scale became known as the two-sigma problem.
Bloom’s two sigma problem comes up a lot in talk about EdTech, specifically when it comes to personalised learning. Platforms that use machine learning to adapt their responses to the needs of individual students aim to personalise education to such an extent that it would mirror the input from a one-to-one tutor.
I believe that adaptive educational content is a very powerful tool. I don’t believe that software personalisation can solve the two-sigma problem.
Adaptive software often fails to take into account the role of social obligation and interaction in the tutorial relationship. As a student, I want to impress my tutor. I don’t care so much about impressing my tablet.
I also know that if I show up to a tutorial unprepared, not only will I disappoint my tutor, I will have wasted their work and effort in preparing for the session. I have an obligation to another human being to do the work.
Game-based mechanics and reward systems can be wonderful at providing incentives to take part and progress through learning content, but the allure of extrinsic reward and levelling up wears off over time. Social obligation and social rewards strengthen over time.
This doesn’t necessarily mean that it is impossible to scale the impact of one-to-one tuition with technology. It means that to design effective tuition platforms, building in social interaction, obligation and feedback is as important as designing adaptive content.
There is plenty of practice to draw on from social networks. For example, Snapchat has designed in the social obligation to use the network every day. Snapchat streaks require two users to snap each other every day to build and maintain a streak score. It is highly compelling because if you don’t maintain your streak, your friend loses the score as well as you. It is so effective that young people I’ve worked with have told me that they leave phones with friends to maintain their streaks while they are on holiday.
I find this level of persuasive design pretty reprehensible on a youth-focused advertising platform, but similar mechanics could be used for more productive purposes. Could you engineer a streak programme that ties the progress of two similar students in a study group?
In terms of the need to impress, young hackers have told me about how they ‘code shift’ to a as young or inexperienced. They spend time searching the boards for answers before posting, because it is shameful to ask something that has already been answered on another thread.
On Scratch’s publishing platform, they found that users were posting their own projects, not looking at anyone else’s and then not returning to the platform. They now employ moderators to like and comment on work that is uploaded, and signpost other people’s work because social feedback is vital to get people to post again.
In short, adaptive learning platforms will never provide personalised learning, until they feel personal. Social interaction, obligation and feedback is what enables a learner to feel seen in their learning journey, and if edtech aims to solve the two sigma problem, these must be designed in to their systems. A machine can reward the learner’s progress, but it takes a person to witness the learner’s becoming. And isn’t that why we learn?