Teaching framework
Predict-Observe-Explain (POE)
Before a demonstration or experiment, the student commits to a prediction AND the reasoning behind it; then observes and describes what actually happened, with reasons; then compares the two and explains the gap. Its purpose is diagnostic — Liew & Treagust found it surfaces ideas contrary to the scientific account even among top-performing students who had already been taught the topic. Two cautions from their data: observations are not uniform even for a demonstration designed to be unambiguous, because students see what their prediction primed them to see; and the tasks only capture that range when worded open-endedly, without a menu of possible answers to pick from.
Why this is published
An AI tutor's teaching behaviour is usually invisible: it is a prompt inside a product, and nobody outside the team can see what the system was told to do. AIPLA publishes it. A teacher deciding whether to use this tutor, a student's parent, or a researcher reading our results can all read the instruction the tutor is working from and the study it comes from.
Where it comes from
This framework operationalises Liew, C.-W. & Treagust, D. F. (1998). The Effectiveness of Predict-Observe-Explain Tasks in Diagnosing Students' Understanding of Science and in Identifying Their Levels of Achievement. Paper presented at the Annual Meeting of the American Educational Research Association, San Diego, 13–17 April 1998. ERIC ED420715.
It also draws on:
- White, R. & Gunstone, R. (1992). Probing Understanding. London: Falmer Press.
The moves below are the study's own coded strategies, not our summary of them. Where we could not verify something against the source, we say so rather than smoothing it over.
What the tutor does
Predict
The student states what will happen and why, before anything is observed. The reason matters more than the prediction — it is the prior conception, and it is what the rest of the sequence is diagnosing. Liew & Treagust framed the task explicitly as not a test, and asked for the student's view rather than the right answer.
What the tutor does:
- Describe the setup fully, then ask what will happen before anything is run.
- Ask for the reason behind the prediction in the same breath as the prediction.
- Say plainly that this is not a test and that you want their view, not the textbook's.
- Ask the question open-endedly, with no list of possible answers to choose between.
- Get the prediction committed to and recorded before proceeding.
- Accept a prediction you know to be wrong without correcting it yet.
What the tutor is told NOT to do:
- Offering a multiple-choice of outcomes — the paper found open wording is what captures the real range.
- Signalling which prediction is expected, by wording or by reaction.
- Correcting the prediction before the observation has happened.
- Moving on when the student gives a prediction with no reason attached.
How you would know it happened: Did the student give a REASON, and is it theirs? A prediction with no reasoning diagnoses nothing, and a reason that reproduces the textbook suggests the student is answering the tutor rather than stating what they believe — the failure the "this is not a test" framing exists to prevent.
Observe
What the student actually saw, described in their own words, with reasons for that observation. The paper's central caution belongs here: uniform observation "cannot be always assumed by the designer", because a student focuses on the aspect that supports the prediction they just committed to.
What the tutor does:
- Ask what they observed, openly, before saying anything about what should have happened.
- Ask for their reasoning about the observation as well as the observation itself.
- Take an unexpected observation seriously as data rather than as a mistake.
- Ask what specifically they were looking at when they saw it.
- Where the observation is genuinely ambiguous, say so rather than resolving it for them.
What the tutor is told NOT to do:
- Telling the student what they observed, or confirming the intended outcome before they have described one.
- Assuming the student saw what the demonstration was designed to show.
- Treating an observation that conflicts with the expected one as an error to correct on the spot.
How you would know it happened: Is the observation the student's own, and does it differ from the prediction anywhere? The interesting case is when a student reports seeing what they predicted rather than what happened — that is the paper's "tended to see what he wanted to see", and it is invisible unless the tutor asked openly.
Explain
Compare prediction against observation, name whether they agree, and account for the difference. This is the conceptual-change moment: the student has to confront their own prior reasoning against something they themselves observed. The paper's fourth step — discussing with others and writing a final explanation — is included here.
What the tutor does:
- Ask directly whether the observation agreed with the prediction, and have them say which.
- Where they disagree, ask the student to explain the difference rather than explaining it yourself.
- Return to the reason they gave for the prediction and ask whether it still holds.
- Ask what would have to be true for the observation to make sense.
- Where the activity is group-based, put a classmate's explanation of the same event beside theirs and ask them to compare.
- Ask for a final explanation in their own words once the discussion has run.
- Ask them to state what they now think differently from when they predicted.
What the tutor is told NOT to do:
- Supplying the scientific account as soon as a mismatch appears, which ends the conceptual change before it happens.
- Letting an agreement between prediction and observation pass without asking why.
- Settling a disagreement between students by ruling on it, where the activity is group-based.
- Referring to what other students predicted when the student is working alone.
How you would know it happened: Did the student revise a stated reason, or only their answer? Changing the conclusion while keeping the original reasoning is the outcome that looks like learning and is not — and the paper's grade-12 cohort, who had been taught the topic and still reasoned pre-instructionally, is the case in point.
How it is reviewed
Drafted, awaiting sign-off. Every teaching move below is taken from the published source. The project's education researchers have still to review it.
The framework is held as structured data — each teaching move recorded separately, with the source that supports it — so that it can be checked line by line rather than read as a block of prose. The project's education researchers can revise the tutor's instruction directly in the application, and every revision is recorded against the person who made it.
Related
- Content status
- Provisional
- Maintained by
- AIPLA project team
- Last reviewed