Orbrya
All posts
Orbrya2026-03-12

The PAUSE Framework for AI Literacy

Plan, Ask, Understand, Spot, Evaluate. The PAUSE framework gives families a five-step habit that turns AI from a shortcut into a genuine thinking tool.

There's a version of AI use that looks productive and probably isn't. A student sits down with an assignment, types a prompt, reads what comes back, maybe adjusts a sentence or two, and submits. Work done. No critical engagement happened. No verification. No thinking about whether any of it was accurate or appropriate or actually what the assignment required. The AI ran the task and the student delivered the result.

Most families know this is a problem. Fewer have a concrete alternative to offer, something that doesn't just say "don't use AI that way" but actually describes what using AI well looks like.

The PAUSE framework is that alternative. It's a five-step habit that can be applied to any AI interaction, adapted for any age, and taught without any specialized knowledge. The steps are Plan, Ask, Understand, Spot, and Evaluate. The name is intentional: it asks the user to slow down at every stage, because the cognitive danger of AI is almost entirely in the rushing.

Plan: Before you prompt

The planning step happens before a single word is typed. It asks three questions: What do I actually need here? Where are errors most likely to occur? What will I need to verify afterward?

This is the step that most AI users skip entirely, and it's the one that makes every step after it easier. A student who knows what they need going in will write a better prompt, will read the output with sharper attention, and will know exactly which claims to check.

Planning also builds a subtle but important habit: it keeps the student's brain in charge of the task from the beginning. When you start by thinking rather than prompting, you've already done more cognitive work than someone who starts by outsourcing.

For younger children, planning can be as simple as: "What question are we actually trying to answer, and what would a wrong answer look like?"

Ask: The prompt itself

The Ask step is the prompting -- what most people think of as the whole process. It comes second here deliberately, because a student who has planned well will prompt differently than one who hasn't.

A planned prompt gives AI context, specifies what's needed, and where possible anticipates the failure modes. Instead of "explain photosynthesis," it might be "explain photosynthesis for a ten-year-old's understanding, note if anything is still debated among scientists, and tell me which parts I should double-check."

That prompt is not harder to write. It just requires thirty seconds of thinking beforehand.

Understand: Reading actively as output generates

Most people read AI output the way they read a text message -- passively, looking for the gist. The Understand step asks for something different: active reading that is alert to anything that seems off.

This means noticing when a claim sounds too confident for the question, when two parts of a response seem to contradict each other, when a statistic appears without a source, or when the AI seems to be filling in gaps with plausible-sounding guesses.

For children, this step is often described as reading like a detective. A detective doesn't accept what they're told. They notice details, they question inconsistencies, and they stay alert to the possibility that the story doesn't quite add up.

Active reading during AI output is a learnable skill. It gets easier with practice, and it transfers directly to reading any kind of text critically.

Spot: Identifying claims that need checking

Spotting is the step where the student marks specific claims for verification. Not everything in an AI response needs to be checked against an outside source -- that would make AI unusable. But some claims do, and the skill is knowing which ones.

Claims worth spotting include specific statistics and numbers, names and dates, any claim that the student doesn't already know to be true, anything that would significantly change the meaning if it turned out to be wrong, and anything presented as settled when the student suspects it might be contested.

The Spot step is directly connected to the "Spot the Lie" family game described in an earlier post -- the underlying skill is identical. The game version makes practice feel low-stakes and enjoyable. The PAUSE version applies that same skill to real work.

Evaluate: Verifying against reliable sources

The final step is checking the spotted claims against sources that don't rely on the same underlying data as the AI. This usually means going to primary sources: the original study, the official record, the expert who is actually named and whose work can be found.

The Evaluate step is where families can do a lot of good with simple habits. A parent who asks "where could we check that?" after every AI session is building the Evaluate reflex in a child who will eventually do it automatically.

What makes the Evaluate step different from a generic instruction to "fact-check" is that it comes after deliberate spotting. A student who has already identified which claims need checking doesn't have to evaluate everything -- they have a specific, manageable list. That makes the step feel tractable rather than overwhelming.

Why the acronym matters

PAUSE is not just a mnemonic. It's the entire instruction. A parent can ask "did you PAUSE on this?" and the child knows exactly what that question means -- did you plan before you prompted, did you read actively, did you spot and evaluate the claims that needed checking?

That single question can close a conversation or open one, depending on the answer. It gives families a shared vocabulary for a habit that's otherwise hard to describe.

Most AI literacy guidance tells children what not to do. The PAUSE framework tells them what to do instead -- step by step, in an order that makes cognitive sense, in a form short enough to actually remember.

The research behind each step is substantial. The planning and monitoring phases map to what researchers call metacognitive oversight, a measurable predictor of how well students actually learn from AI rather than around it. The verification habit maps to what employers are increasingly identifying as the skill that separates competent AI users from passive ones.

But families don't need to know the research to use the framework. They just need to ask, consistently: did you PAUSE?

Sources cited in this post: