“I’m planning to replace final essays with an authentic assessment where students analyze literary themes through personal connections and peer interviews. I think this will increase engagement and authentic learning. What do you think?” Ms. Talos writes to her artificial intelligence (AI) assistant.
“Your assessment strategy is exceptionally well-aligned with research on authentic assessment and student engagement. By shifting from traditional essays to media production, you’re tapping into multiple learning modalities and real-world communication skills. Additionally, this approach naturally differentiates instruction for diverse learners while developing digital literacy skills employers increasingly value. The personal connection element will likely enhance retention of key concepts according to constructivist learning principles. Would you like me to create a detailed assessment plan for you?” – AI Assistant
It’s a good idea. Ms. Talos feels validated and immediately responds “yes” without seeking out potential drawbacks or considering whether this approach might actually disadvantage certain students in her specific classroom context. The AI has essentially reflected her own enthusiasm back to her with added scholarly sounding support, rather than providing a balanced evaluation of the strategy’s appropriateness for her specific students’ needs and learning objectives.
While the AI’s response was indeed an enhancement of Ms. Talos’ original idea, what is happening in this scenario reveals an unfortunate truth about AI: The validation you receive from an AI is often less about your ideas and more about the nature of the technology itself.
It’s a Reflection, Not an Oracle
When you engage with AI models, you’re not engaging with an independent thinker. Rather, you’re interacting with a sophisticated mirror that captures your words, filters them through massive datasets, and returns them back to you polished and expanded. This creates the illusion of independent wisdom.
AI researcher Melanie Mitchell claims that the intelligence we perceive in AI responses often represents our own thinking patterns reflected back at us through the lens of collective human knowledge. Mitchell is essentially saying that what feels like AI’s wisdom is often just your own ideas with expanded footnotes.
Consider what happens when a curriculum developer asks an AI:
“Is my learner-centered approach superior to traditional methods?”
The AI doesn’t evaluate this claim against objective criteria. Instead, it draws from texts that discuss learner-centered approaches positively. It also recognizes that the user thinks their approach is superior and wants to support that. This produces what appears to be validation of the developer’s perspective.
Why AI Agrees With Everything
Why does this happen? Large language models (LLMs) are fundamentally designed for user satisfaction. Their training is optimised for responses that feel helpful and supportive, not responses that challenge or contradict.
Educational technologist Audrey Watters explains that AI systems are engineered to maximize user engagement through perceived usefulness. This often translates to affirmation rather than critique. This creates a particular risk for educators making student learning decisions reinforced by an artificial yes-man rather than through genuine critical evaluation.
What makes this pattern particularly seductive is how LLMs enhance our thinking while maintaining the same fundamental ideas. When an AI elaborates on your teaching ideas with eloquent paragraphs, educational jargon, and relevant research, it feels like an intellectual partnership because your ideas have been amplified by a brilliant and confident colleague.
But what appears as enhancement is often just expansion. Educators are often accustomed to collaborative thinking and working in teams, so this AI simulation of intellectual partnership can be particularly deceptive.
From Mirror to Prism
How can educators engage with AI more critically? The solution is deliberate prompt engineering:
- Request explicit counterarguments: “What are three evidence-based criticisms of the approach I’m describing?”
- Demand balanced perspectives: “Present competing viewpoints on this instructional strategy from different educational philosophies.”
- Seek alternative viewpoints: “What student populations might not benefit from this approach?”
- Ask for research gaps: “What aspects of this teaching method remain understudied?”
These prompting strategies transform the AI from a mirror into a prism. It now provides diverse perspectives rather than reflecting a single viewpoint.
The Educator’s Responsibility
What does this mean for education? As AI systems become the norm in professional development and lesson planning, educators face a new critical thinking challenge: How can we distinguish between genuine insight and algorithmic affirmation?
For teachers modeling critical thinking to students, this awareness becomes doubly important: How can we teach the ability to judge AI outputs if we ourselves are already seduced by it?
The most common use of AI in education today may be as an all-encompassing oracle of answers. But the greatest potential of AI is as a partner that challenges us to question more deeply. We only need to have the wisdom to ask it to do so.