TL;DR:
AI-powered activities can be engaging, but that doesn’t always mean students are truly learning. Just like any classroom activity, AI tools need to be carefully chosen and integrated to avoid becoming distractions or “poor proxies” for actual understanding. Experiment with AI tools, but always focus on deep learning goals and use your experience to guide how they are implemented for the best results.
What are Poor Proxies for Learning?
Professor Coe’s 2014 presentation on the idea of “poor proxies for learning” has left a lasting impression. It’s a much-needed reminder that we shouldn’t misinterpret task completion or superficial performance as evidence of genuine understanding. As educators, we always need to be wary of placing undue emphasis on behavioural compliance and superficial activities.
It’s easy to mistake these common classroom scenarios for signs of deep learning:
- Students are busy: lots of work is done
- Students are engaged, interested, motivated
- Students are getting attention: feedback, explanations
- Classroom is ordered, calm, under control
- Curriculum has been ‘covered’ (i.e. presented to students in some form)
- (At least some) students have supplied correct answers, even if they:
- Have not really understood them
- Could not reproduce them independently
- Will have forgotten it by next week (tomorrow?)
- Already knew how to do this anyway
With the surge of AI in education, a critical question arises: “Are we inadvertently creating new poor proxies for learning with AI tools?” My own work makes me question this – am I providing true depth with my Choose Your Own Adventure prompts or simply repackaging content?
To be clear, I’m not seeking to disparage any specific tool or approach, but the potential dangers of AI as a ‘proxy for learning’ must be examined.
Where AI Can Go Wrong: The ‘Poor Proxy’ Trap
The danger is real. Here’s where AI could slip into the ‘poor proxy’ trap:
- Engagement vs. Learning: AI-powered activities or chatbot interactions can be captivating, but we mustn’t mistake student engagement for deep understanding.
- Prioritising Completion: We need to be wary of AI-driven platforms that overemphasise ticking off tasks rather than ensuring conceptual mastery.
- Adapting to Superficial Patterns: Even sophisticated AI systems may adapt instruction based on surface-level responses, missing the underlying nuances of a student’s thinking.
The Power of Expertise and Experimentation
My own approach involves rigorously testing every chatbot I develop for student use. My rule of thumb: if it gives the consistent, error free, outputs 10 times in a row, then it’s ready for students. I apply my teaching expertise to ensure these bots are genuinely scaffolding and supporting learning, not replacing the process itself. Tools like Mindjoy’s Socratic mode are incredibly helpful in boosting this support.
It’s the same approach I take with AI-animated characters in D-ID, VR experiences, and other AI-powered tools. My years of experience guide me in making thoughtful choices, just as I would for any activity, resource or tool for any lesson.
My Experience: Lessons Learned
I once used a chatbot to teach an entire lesson. While the students seemed interested and involved at the time, later retrieval practice and questioning revealed they hadn’t retained much of the information.
On the other hand, Using AI to bring historical figures to life or support coding revision has fostered deeper connections and meaningful questions in my students. Recently, an AI-powered conflict resolution expert bot in PSHE ignited passionate discussions about a complex social issue. By grappling with a complex social issue and exploring diverse perspectives through the interactions, they were empowered to develop their own problem-solving approaches to real-world conflicts.
An AI chatbot even helped me cover an A-level science class, even though I didn’t know the subject beforehand. The chatbot handled the factual information, while I focused on how the lesson unfolded and how I questioned the students. This experience is similar to what experts at PISA are exploring – using AI as a knowledge resource with questions designed for students to think critically and solve problems.
AI tools can be powerful, but they work best when combined with a teacher’s expertise. Just like in my science cover lesson, the AI handled the content, allowing me to ensure the lesson was engaging, thought-provoking, and designed for long-term knowledge retention.
Mitigating the Risks: Towards AI-Enhanced Deep Learning
Let’s choose AI-infused activities that empower our students to become active thinkers.
Prioritise tools and experiences that nurture problem-solving, critical analysis, and an understanding of their own learning process. True learning flourishes when students grapple with ideas, ask questions, and discover solutions for themselves.
With each new AI tool, ask yourself: will this make learning richer for my students?
Your teacher intuition is a valuable guide. But remember, don’t be afraid to experiment – occasionally I’m surprise what makes learning “stick”.
AI has the potential to revolutionise education, but only if we use it thoughtfully. By avoiding the ‘poor proxy’ trap and using our expertise as educators, we can unleash its potential towards deeper, more meaningful learning.
Matthew Wemyss 