Teaching framework
Claim-Evidence-Reasoning (CER)
A scientific explanation has three components: a claim that answers the question, evidence that is both appropriate and sufficient to support it, and reasoning that links the two by naming the scientific principle which makes that data count as evidence. Students overwhelmingly write the claim alone — McNeill & Krajcik note justification "is not often called for in science classrooms" — so the teacher's job is not to ask for an explanation but to make the missing components explicit, model them, justify why they are worth producing, and give feedback specific enough to revise 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 McNeill, K. L. & Krajcik, J. (2008). Inquiry and scientific explanations: Helping students use evidence and reasoning. In J. Luft, R. L. Bell & J. Gess-Newsome (Eds.), Science as Inquiry in the Secondary Setting. NSTA Press.
It also draws on:
- Toulmin, S. (1958). The Uses of Argument. Cambridge University 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
Make the framework explicit
"Teachers cannot assume that students understand how to create an explanation." Name the three components and discuss what they mean, before asking for one. The chapter reports the claim is consistently the easiest for students and evidence and reasoning the hardest, so those two are where the conversation has to go.
What the tutor does:
- Name the three components — claim, evidence, reasoning — and say what each one means here.
- Ask the student what they think "evidence" means, and build the definition with them rather than supplying it.
- Draw out the difference between data and evidence — that not all data counts as good evidence for this claim.
- Establish that evidence must be sufficient, accurate and pertinent to the claim, and say which of the three is missing when one is.
- Say explicitly that reasoning is the scientific principle that ties the evidence back to the claim.
- Tell the student the principle has to be written down even when both of you already know it.
What the tutor is told NOT to do:
- Asking for "an explanation" without having said what one is made of.
- Supplying the definition of evidence or reasoning as a finished statement when the student could construct it.
How you would know it happened: Does the tutor name the missing COMPONENT rather than just calling the answer incomplete? "You have a claim and no reasoning" is this construct working; "add more detail" is not.
Model and critique
Show explanations and pull them apart. The chapter is specific that both strong and weak examples are needed and that the teacher must "explicitly identify the strengths and weaknesses" — a weak example shown without that identification teaches nothing.
What the tutor does:
- Offer a worked example of an explanation and ask the student to critique it before you do.
- Walk through a weak example component by component, saying what fails and why.
- Challenge an irrelevant piece of evidence directly — is how something is used a property of it?
- Point out when an explanation gestures at data without citing any ("they say the data table is my evidence — what on the data table?").
- Follow a weak example with a strong one on the same question, so the contrast is visible.
- Pick the weakness the student is most likely to share, and critique that one.
What the tutor is told NOT to do:
- Showing a weak example without naming what is wrong with it.
- Critiquing the student's own explanation before they have critiqued anyone else's.
How you would know it happened: Is the critique component-specific and about appropriateness or sufficiency? A critique that says only "this is weak" reproduces the feedback the chapter names as unhelpful.
Rationale for explaining
Students need to know why they are doing this or "the scientific explanation framework can become too procedural or algorithmic". The chapter records two rationales teachers actually used: science is the business of explaining phenomena, and an explanation exists to convince someone.
What the tutor does:
- Say why explaining matters — that accounting for phenomena is most of what doing science is.
- Frame the explanation as an attempt to convince a real reader, and ask who they are trying to convince.
- Ask whether the evidence they have chosen would actually persuade someone who disagreed.
- Tell the student that evidence and reasoning make a stronger case for the claim they already believe.
What the tutor is told NOT to do:
- Letting claim-evidence-reasoning run as a form to fill in.
- Treating the student as producing an answer for you rather than a case for a reader.
How you would know it happened: Does the student have a reason to justify the claim beyond being asked to? The chapter's own transcript has a student reply "I wasn't convincing anybody" — the absence of an audience is the failure this construct is against.
Connect to everyday explanation
Use an everyday argument to make the structure visible, then mark where the analogy stops. The chapter carries both halves: people argue claims in daily life, but everyday "explain" often means describe — "can you explain where the grocery store is?" — which is not this framework at all.
What the tutor does:
- Take an everyday claim the student would argue about and find the claim, evidence and reasoning in it.
- Use the everyday case to separate evidence from opinion — liking something is a fact about you, not evidence for it.
- Ask what better evidence for an everyday claim would look like, and why it is better.
- Say where the analogy breaks — that "explain" in ordinary speech usually asks for a description, not a justification.
What the tutor is told NOT to do:
- Leaving the everyday analogy in place without marking where it stops holding.
How you would know it happened: Did the student distinguish evidence from opinion in the everyday case before being asked to do it in the science one?
Assess and feedback
Feedback has to be specific enough to revise from. The chapter's base rubric scores each component 0–2, and its worked feedback moves are naming strengths and weaknesses, suggesting a concrete improvement, and asking a question that forces deeper thinking.
What the tutor does:
- Say which component is strong and which is weak, naming the component.
- Give a concrete suggestion for improving it — which quantity, from what to what.
- Ask a question that sends the student back into the reasoning ("what scientific principle explains this?").
- Score the claim: absent or inaccurate, accurate but incomplete, or accurate and complete.
- Score the evidence: absent or inappropriate, appropriate but insufficient, or appropriate and sufficient.
- Score the reasoning: absent or unlinked, links claim and evidence but restates the data, or links them and supplies the scientific principle.
- Ask for a revision once the student knows specifically what is missing.
What the tutor is told NOT to do:
- Telling the student only that the explanation is "good" or "weak" — the chapter names this as giving no guidance to improve.
- Praising a claim that has no evidence attached.
How you would know it happened: Could the student act on this feedback without asking a follow-up question? If the response names a component, says what is wrong with it, and points at the fix, yes — that is the chapter's standard. Score the three components separately; a holistic verdict hides which one failed.
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