Form teams and negotiate team contracts

1

Research

Why this matters, and what good looks like

A team contract records how students agree to contribute, work together and revisit problems. PBLWorks offers professional guidance; the team-contract source available here is its public description. The full guide is login-gated and has not been verified. EEF synthesises collaborative-learning evidence, whose security it rates as low. The catalogue's Strong label is an editorial classification, not EEF's rating. These sources establish no AI benefit. Structure comparison and role rotation below are practical design choices.

Plan structured participation and practice, rather than assuming grouping produces collaboration (EEF).
Let students generate and revisit agreements about contributions and problems (public contract description).
Keep both individual and team work visible without asking AI to assess learners (PBLWorks teaching practices).

Where the evidence comes from

Strong for collaborative learning and team contracts; AI benefit unvalidated.

2

Workflow

Brief it, steer it, check it

Compare structures without sharing learners

Use optional D04 and L02 outputs, or equivalent approved teacher documents. Limit this instance to three small teams and one task. Keep rosters, names, pseudonymous profiles, support and safeguarding facts offline. Compare structures only.

Prompt 1 · Compare structures

CONTEXT
Approved project task and collaboration expectations: [paste safe task and expectations]
Two grouping structures without people, profiles or identifiable characteristics: [paste structures]
Existing blank contract: [paste safe blank form or none]
REQUEST
Compare the supplied structures for participation, coordination and visibility of individual contributions. Flag missing requirements for a blank contract.
QUALITY BAR
Use only supplied task facts. No names, pseudonymous learner profiles, allocation, ranking or preferred structure. If personal data appears or a required input is missing, return HOLD and the issue category without reproducing sensitive text. Teachers retain all grouping, conflict and safeguarding decisions. Do not infer learner needs, grades or ability.
FORMAT
A comparison table: structure, task-linked opportunity, risk to check offline. Then contract gaps and HOLD items. Label all comparisons proposals, never verified classroom facts.

Choose teams and prepare negotiation

Choose the structure and allocate real students offline. Teach collaboration expectations. Check and edit the comparison, then ask for discussion prompts. A blank contract leaves decisions open; it must never imply students already agreed.

Prompt 2 · Prepare discussion

CONTEXT
Exact step 1 output checked by teacher: [paste approved output]
Teacher edits and chosen structure, with no roster: [paste explicit edits and structural choice]
Teacher-confirmed task requirements and escalation boundary: [paste safe requirements and boundary]
REQUEST
Draft up to six neutral discussion prompts and a blank recording form for students to negotiate contribution, communication, decisions, conflict referral and rotating roles. Include a review trigger and space for dissent.
QUALITY BAR
Never write agreed terms or assign people. No invented policy, sanction or consent. Refer live conflicts and safeguarding to the teacher through the supplied boundary. If human choice, task requirements or boundary is absent, return HOLD. Carry unsupported claims as HOLD. Do not use identifiable learner data or decide grouping, conflict, safeguarding or grades.
FORMAT
Discussion prompts followed by blank fields for student wording, unresolved disagreement, role-rotation proposal and review trigger. Identify teacher checks; label this preparation only.

Let students negotiate and record dissent

Check and edit the discussion sheet. Every student proposes terms, hears alternatives and can disagree. Students negotiate role rotation and a review trigger. Record exact terms offline. Bring only non-identifying clauses to AI for gap flags.

Prompt 3 · Check negotiated coverage

CONTEXT
Exact step 2 output checked by teacher: [paste approved output]
Teacher edits to the discussion sheet: [paste explicit edits or none]
Non-identifying student-negotiated clauses and role rotation, including dissent: [paste safe clauses with clause IDs]
Teacher record of participation and unresolved categories, without learner details: [paste safe process record]
REQUEST
Flag coverage gaps in contribution, communication, decision-making, conflict referral, role rotation and review. Point to supplied clause IDs and unresolved categories.
QUALITY BAR
Do not rewrite student terms, infer consensus, assign roles or resolve disputes. Missing participation or negotiations means HOLD. Retain dissent. Unsupported or stale claims remain HOLD. No personal profiles, grouping, conflict, safeguarding or grading decisions. A teacher must check equity offline and return unresolved terms to students.
FORMAT
Table: requirement, exact clause evidence, gap or HOLD, human check needed. End with unresolved categories, without deciding whether the agreement is fair or approved.

Check the human-selected agreement

Return gaps to students and revise together. Check equity and all sensitive matters offline. Select the agreement for audit, then personally release or hold it. Keep the real roster attached only offline. Stop or re-scope before exceeding 60 active minutes.

Prompt 4 · Audit selected agreement

CONTEXT
Exact step 3 output checked by teacher: [paste approved output]
Teacher corrections and subsequent student decisions, without identities: [paste versioned decisions]
Teacher-selected non-identifying final clauses and rotating-role plan: [paste selected safe text]
Teacher offline check record for roster, equity, escalation and participation: [paste completion categories only]
REQUEST
Audit the selected text against the supplied requirements, unresolved categories and explicit human decisions. Flag missing coverage, changed student terms, invented consensus and unsupported claims.
QUALITY BAR
Do not rewrite, approve, allocate learners or decide conflict, safeguarding or grades. Do not request private records. Missing human decisions, unresolved dissent blocking agreement, stale facts or incomplete offline checks mean HOLD. The teacher and students decide revisions; teacher release stays outside AI.
FORMAT
Two lists: HOLD issues with exact clause or input evidence and human action; matched items with source evidence. State that textual comparison cannot verify participation, safety, equity or readiness.

Check before you use it

Have you kept identities, pseudonymous learner profiles and the real roster outside AI?
Did you make all grouping, conflict and safeguarding decisions yourself?
Did every student genuinely negotiate terms and role rotation, with dissent retained?
Do the clauses cover contributions, communication, decisions, referral and a review trigger?
Have students revisited gaps and have you checked equity before release?
If agreement exceeds the bounded session, have you stopped and re-scoped without claiming consensus?

Stop rule

Never expose student identities or delegate grouping, conflict or safeguarding decisions to AI.

3

Reuse it

So the next one takes minutes

Suitable

Build a skill

Save this as a reusable instruction you can run on any project.

The bounded comparison and checking sequence can repeat with new non-sensitive materials and explicit human decisions.

Not suitable

Build an agent

Chain the steps, with a checkpoint where you approve.

Live student negotiation and teacher release require direct participation; unattended monitoring is outside this task.