NDCA Framework

N · D · C · A

A Framework for Critical AI Exploration

AI sounds confident and polished, and that fluency makes it hard to question. This framework gives you and your students four moves for interacting with AI, prompting students to draw on their cultural knowledge, identity, and critical thinking so they stay in control of the tool.

Try out the framework with our classroom-ready activities →

Notice.

Explore AI responses, considering what is included, what is left out, and whose world it represents. Notice how it is being communicated and how it makes people feel.

Learn more
Students look closely at what the AI actually produced and compare it against their own experience, a community member’s knowledge, or a primary source. They bring their own judgment and logic to the response rather than taking it at face value. Part of this is noticing how it makes them feel — a fluent, confident tone is what we normally use to decide whether to trust a person, so it can make the AI feel more sure and more right than it has earned.

Discuss.

Talk about your reactions, sharing what you know, who you are, and where you come from.

Learn more
Critical thinking about AI isn’t a solitary act between a student and a screen. It happens between people. Students recognize they hold knowledge the AI doesn’t, and learn diverse perspectives by talking with others.

Challenge.

Rethink AI responses, resisting when it stereotypes, flattens, distorts, or excludes.

Learn more
Students name what’s wrong and assert a truer version. Disagreeing with a confident machine can be challenging, and it counts as meaningful action, not just “thinking.”
↓

Act.

Share, advocate, and collaborate with your community for collective change.

Learn more
The horizon: taking what you’ve found to an audience beyond the room.

Tap any move to expand it.  ·  equityinai.net · Equitable AI in the Classroom

What it is

A way of interacting with AI

Generative AI produces the most likely answer, smoothed toward the middle. It tends to flatten what is specific and personal about your students’ worlds.

Four moves (Notice, Discuss, Challenge, Act) can help teachers design activities in any subject where students learn to read AI’s responses against their own knowledge and experience, and to push back when something is flattened or left out.

Activities designed with the NDCA framework build critical AI literacy while also developing content knowledge and meeting curricular standards.

How to use it

A lens for your own content area

This framework provides scaffolding for designing critical AI activities in any subject. Discussion runs throughout, and the steps loop and build on each other. The goal is to help students develop the skills and confidence to engage their own sources of knowledge and critical thinking while using AI.

1
The one essential move

Always give students something real to set the AI’s response beside.

For academics · Learn more

The thinking behind it

NDCA adapts Paulo Freire’s critical pedagogy, then makes two deliberate moves beyond simplification.

Conscientization

Reading the world for inequity, including in AI.

Dialogue

Reflection between people.

Praxis

Reflection and action to transform the world.

Move one · It splits praxis in two

Challenge names micro-praxis: the cognitive and identity-based resistance that is itself meaningful critical action. Act holds fuller, collective praxis as an aspiration rather than a requirement.

Move two · It foregrounds identity

This is not only critical thinking applied to AI, but students protecting and asserting their sense of self against outputs that stereotype, flatten, distort, or exclude.

Four ways AI misrepresents

The Challenge categories give shared language for how AI misrepresents, ordered from most visible to most subtle.

  • Stereotype. Reduces a person or group to a fixed, oversimplified image. (most visible)
  • Flatten. Smooths over nuance and variation until everything reads the same.
  • Distort. Bends the facts; a specific or simply uncommon detail comes out wrong.
  • Exclude. Leaves something, or someone, out of the picture entirely. (most subtle)

They stem from a single mechanism: because generative AI produces the most statistically likely output, it regresses toward the dominant and the average. That is at once an accuracy problem (nuance lost) and an equity problem (the marginal erased). The same move that erases a culturally specific detail also erases a correct but uncommon one.

On epistemic agency and AI literacy

Beneath the four moves is a disposition we are still working out: the confidence to treat what you know and have lived as a legitimate basis for questioning an AI that sounds authoritative. Generative AI has pulled fluency apart from competence, and people are not built to catch the difference. Reading AI critically requires building habits of mind, so the framework tries to provide practice through the moves instead of assuming students already have it.

We have kept this off the teacher-facing figure for now, while the idea settles. Right now we are focusing on epistemic agency:

“the ability to utilize persuasively shared epistemic resources within a given community of knowers in order to participate in knowledge production and, if required, the revision of those same resources.”— Dotson, 2014, p. 115

That capacity also depends on recognizing where your own knowledge runs out and when to turn to another person rather than the machine, and on what the literature calls positioning: whether a teacher or student actually has the standing to act, which not everyone is given. What we are ultimately after is decentering AI as an epistemic authority, so that a person’s own knowledge counts for something against it. Equity sits underneath this, since the confidence to push back has never been handed out evenly (see Dotson, 2014).

Additional resources
Scroll to Top