Research analytics layer · educators and researchers

Tool ecology and familiarity

Contribution is not a property of one session. This layer records which tools a student chose, rejected, switched between and returned to, and how accumulated interaction history changes what the AI can infer from very little explicit input. Nothing here is a score, and external-tool use is never treated as misuse.

Subject to privacy, consent, institutional policy and applicable research ethics. Students see only the lightweight prompts, never this layer.

Tool events

4

Unaided steps recorded

1

Returns to Con-Met

3

Outputs evaluated

4/4

Observed tool trajectory

The sequence is observed, not prescribed. There is no correct path.

Con-Met learning partnerGeneral-purpose chat assistantLibrary databaseUnaided

Tool-choice provenance events

No tool — unaided12/30/1969 · stage · revising

No tool

Purpose

Decide the final line of argument alone

Stated reason

It performs this particular function better — “I wanted to know what I thought before anything answered me.”

What came back

A decision, plus a reordering of the argument

What the student did with it

accepted · evaluated · did not return

Relation to the final contribution

Non-use recorded as contribution evidence, not as a gap.

General-purpose AI12/27/1969 · stage · exploring

General-purpose chat assistant

Purpose

Generate counterexamples faster

Stated reason

I wanted more divergent ideas — “I wanted breadth before deciding, not a challenge.”

What came back

Ten, of which three were factually wrong

What the student did with it

modified · evaluated · returned to Con-Met

Relation to the final contribution

Two counterexamples survived after verification.

Reference source12/27/1969 · stage · checking

Library database

Purpose

Verify the three claims that looked wrong

Stated reason

I wanted to verify a fact — “The other tool sounded confident and was not.”

What came back

Two disconfirmed, one unverifiable

What the student did with it

rejected · evaluated · returned to Con-Met

Relation to the final contribution

Removed a claim that would have carried the essay in a false direction.

Contribution partner12/26/1969 · stage · framing

Con-Met learning partner

Purpose

Find the assumption underneath my thesis

Stated reason

It performs this particular function better — “It refuses to write, so it forces me to answer.”

What came back

Named two assumptions I had not seen

What the student did with it

accepted · evaluated · returned to Con-Met

Relation to the final contribution

Became the framing of paragraph one.

Compliance versus understanding

Conceptual understanding

Reflections articulate why provenance is recorded and what it is intended to reveal.

Con-Met usage is not Con-Met understanding. The two are recorded separately and neither is inferred from the other.

What do you understand the difference to be between Con-Met AI and a general-purpose AI tool?

The other tool gives me answers. This one keeps a record of what I actually did with them. I think that record is the point, not the answers.

Why do you think your contribution is being recorded?

So the teacher can see where the ideas came from rather than guessing whether I used AI.

Familiarity trajectory

Input length is not a proxy for contribution. As familiarity accumulates, contribution may be compressed into shared context rather than reduced.

Stage 1InitialStudent explains → AI respondsStage 2EmergingStudent directs → AI developsStage 3EstablishedStudent cues → AI anticipatesStage 4HighStudent evaluates → AI refinesStage 5Highly contextualStudent gives minimal signal → AI reconstructs intended direction
Turn 1 · 48 words of explicit inputInitial

Explicit contribution

Full statement of the topic, the claim and the three approaches under consideration.

Contextual contribution

None available. The system knows nothing about this student yet.

References prior work · no · AI reconstruction · none

Turn 6 · 31 words of explicit inputEmerging

Explicit contribution

Direction to develop the second approach, with the constraint restated.

Contextual contribution

The definition of the claim, established two turns earlier.

References prior work · yes · AI reconstruction · correct

Turn 14 · 17 words of explicit inputEstablished

Explicit contribution

A cue: apply the earlier mechanism to the new case.

Contextual contribution

The mechanism, its limits and the rejected alternatives.

References prior work · yes · AI reconstruction · correct

Turn 22 · 9 words of explicit inputHighly contextual

Explicit contribution

“Take the mechanism we developed and apply it here.”

Contextual contribution

An entire shared vocabulary: the mechanism, the counterexamples that survived, the student's ordering preference.

References prior work · yes · AI reconstruction · corrected by student

Not used anywhere in Con-Met: more words = more contribution, more prompts = more contribution, longer interaction = more contribution. Contribution is modelled as current input + prior context + student decisions + AI processing + evaluation + revision + reflection.

AI decision literacy — observation only

A research dimension under observation. Deliberately not reduced to a score at this stage.

  • Recognising that different tools serve different purposes
  • Selecting a tool appropriate to the stage of work
  • Explaining the choice in their own terms
  • Recognising the limitations of the tool chosen
  • Evaluating the output rather than absorbing it
  • Comparing outputs across tools
  • Making an independent judgement after consulting a tool
  • Understanding what provenance records and why