#96: Beyond the Algorithm: Practical AI Integration for Critical Classrooms (Notes)

Prof Rose Luckin

Teachers' Conference and ExCEL Fest 2025 (TCEF2025)

June 3 2025

Key Takeaways

Integrating AI with Intention: Priorities Across Levels

Individual Level

1. Be Purpose-Driven

• Start with having clarity on what you want to achieve with your use of AI. (Consider what skills, knowledge, attitudes and values (SKAV) you want students to develop.)

  • Ensure the use of AI doesn’t short-circuit the learning process or hinder the development of executive functioning. AI should scaffold and extend learning, not replace it.

School Level

2. Learn Fast, Act More Slowly

• Rapidly explore and understand new AI tools, but be slow and deliberate in implementing them.

• Exception: Move quickly on governance — establish clear guidelines and boundaries for safe and ethical AI use in schools.

3. Share Failures, Not Just Successes

• Foster a culture of open reflection to help teachers learn collectively—not just from what worked, but from what didn’t. This encourages teachers to critically reflect on and articulate what constitutes effective vs poor use of AI.

  • E.g. Did the tool deepen student learning or just make things faster? Did it help students think more critically, or reduce their cognitive effort?

System Level

4. Adopt an Evidence-Based, Outcome-Driven Approach

• Move away from tool-first adoption. Focus instead on desired student outcomes (e.g. critical thinking), and then explore how AI can support those outcomes.

• Use a theory of change or logic model:

  • Define intended outcomes.

  • Design learning activities that support these outcomes.

  • Collect and analyse data to see if outcomes were met.

  • Refine based on evidence — not assumptions.

Key question: How can we integrate AI to stimulate critical thinking?

• How can we evaluate AI’s potential to stimulate rather than supplant critical thinking?

• How can we design our use of AI to stimulate rather than supplant critical thinking?

• How can we assess the impact of AI to stimulate rather than supplant critical thinking?

Some components of critical thinking

• Critical thinking refers to sophisticated ability to construct knowledge through evidence-based reasoning, not just consuming information.

• It involves being intelligently discriminating — knowing when to question, what evidence to seek, and how to construct justified understanding in an increasingly complex information landscape.

Practical ways to develop critical thinking

• Structured debate - teach students to marshal evidence, even for positions they disagree with 

• Collaborative problem-solving - require justification of decisions to peers

• Questioning techniques - move students from “what do we know?” to “how do we know?” and “why should we believe this?”

Human vs. Artificial Intelligence

• Human and artificial intelligence are different; humans and machines learn differently.

  • Artificial intelligence learns through experience that's represented by data, not through direct experience of the world.

    • Human intelligence has features like reflection, emotional awareness, contextual understanding, and self-efficacy.

      • Meta-intelligences, such as meta-cognitive and meta-subjective intelligence, are significant aspects of human intelligence.

      • Meta-knowing intelligence, or personal epistemologies, is crucial for navigating misinformation and discerning evidence.

• Super intelligence is possible and necessary for human society.

  • We need to increase the sophistication of our human intelligence through developing advanced meta intelligence. 

  • This is possible through humans effectively using AI, including building critical thinking skills, and we must beware of over-reliance on AI.

Theory of Change & Logic Model 

1. Define purpose - Theory of Change

   • Every use of AI should be driven by the desired outcome.What is the PURPOSE you want your use of AI to achieve – what CHANGE should it enable?

2. Build a Logic Model

   • A logic model shows the relationship between resources, activities and intended outcomes. It offers a useful roadmap for demonstrating impact by showing how specific elements will work, in sequence, towards a final goal.

   • It involves identifying any underlying assumptions at each step of the change process.

     • Examples of assumptions when designing an activity involving AI: technology connection, student technological skills, amount of time spent correlates with engagement, etc.

3. Collect data (light-touch methods)

   • Consider the type(s) of data that can be collected to assess resource sufficiency and activity effectiveness.

   • Types of data:

     • Causal data is hard to obtain.

     • Correlation is easier to find.

     • Anecdotal data can be useful.

     • Data collection methods: observations, focus groups, interviews, surveys, and system-collected data.

     • Use a mix of qualitative and quantitative data.

   • Importance of impact evidence:

     • Builds trust with parents, policymakers and funders

     • Drives better decision-making on AI adoption

     • Protects students from unethical innovation (or innovation for the sake of it; without improved outcomes)

4. Monitor outputs and measure outcomes

   • Output: The thing the students produce.

   • Outcome: The change you want to bring about.

   E.g. A student can produce an essay with the help of AI, but it may not have the outcome of improving their writing skills.

Logic Model: AI Essay Buddy Example

Safeguarding priorities