Compare behaviour. Investigate outcomes.

Parallel conversations, research protocols and controlled course activation.

Checking your GRITT account…

Create a research project

Platform administrators can create independent projects. An institution administrator with member-management permission can create a project for that institution.

Pedagogical character models

Examples are deliberately designed configurations. They are not observations of real teachers. A configuration review is not proof of educational effectiveness.

Review a configuration

Conversation comparisons

Run the same question through the canonical GRITT conversation chain with two to four PCMs. Drafts require an explicit synthetic trial choice and remain unreviewed. Each lane has its own history and narrative state. Personal memories, private documents and real learner-progress writes are excluded. Model calls use your account’s normal configuration and usage allowance.

Create a GRITT comparison

First save a course connection; it may stay disabled for your normal conversations.

Select two to four reviewed models above, then choose a saved course connection below. For an ADEM study template, choose observation-first planning; creating the comparison does not run any answers.

This does not approve the models, activate them in courses or measure student learning. A later negative configuration review or archive stops their use in this trial.

Recovery becomes available after two minutes and preserves incomplete results. An unverified answer cannot count as a verified PCM delivery. A Director that waits is behaving normally.

Review an actual answer

Investigate student learning

Compare what students learn across teaching approaches and subject matter. You are viewing Learner at study scope; use Scope above to return to an individual conversation.

How does a study work?
  1. Plan. Choose the teaching models, questions and measures. Save a fixed protocol before collecting results.
  2. Run. Participants accept the study notice, complete a baseline assessment and work through an assigned lesson.
  3. Evaluate. Compare recorded outcomes, delivery evidence and missing results. Different answers alone do not establish a learning effect.

A fixed protocol records the agreed study design. Saving it does not enrol participants or start lessons.

Your studies

Design a new study

Record the plan before collecting results

Choose the models and measurement method here. Saving fixes this protocol version so the criteria do not silently change after the results are known.

Teaching models to compare
Which design?

Parallel groups compare different groups. Crossover compares approaches across successive episodes. Personal observation records one participant's experience; it does not establish a causal effect.

Measure the result

Course and task references

The course ID is filled from your selected saved source. The research context identifies this study’s participant lessons; its connection value is suggested when available. Name the task set explicitly after reviewing comparability. These references are not new measurements.

Participants accept their own notice. Saving this plan does not start a lesson.
Assessment instruments

An instrument defines the questions and scoring for baseline and outcome measurements. A subject expert should check whether the forms are comparable. Missing results remain unknown.

Register an assessment instrument

Each item has a question, type, answer key and points. Numeric items also state the unit and absolute tolerance. The example is synthetic and requires your review.

Register an offline synthetic record

This creates an accountless record for manual workflow rehearsal. It cannot log in, accept an invitation or run a participant lesson. To test the account workflow, invite a designated test account as a simulated study participant under Access & audit.

Connect a course to GRITT

The proto-narrative graph retains the actual knowledge elements. Research assignment selects a reviewed PCM only for a matching course and context. The default PCM remains available when participation or an assignment ends.

Reference example

A disabled saved connection is enough for a comparison or study package. Enable only if you also intend to use it in your own ordinary conversations.

Course ownership, element text, context and model limits are checked. This interface operates on your own GRITT account. It does not change another participant’s course selection.

Team access

Grant a role to an existing GRITT account. For an institutional project, the account must also have current membership of that institution. No email is sent by this page.

Invite a participant

Create an invitation visible to the participant in their own account. They must accept the exact notice and choose adaptive selection separately. Withdrawal stops subsequent use of their research assignment.

A simulated invitation is for a designated test account and cannot convert an existing research participant. No email is sent.

Export and audit

The default export omits individual rows, quotations and conversation text. Small real groups are suppressed. Free-text labels can still identify people; review an export before sharing.

Analysis packages contain pseudonymous individual outcomes, intervention settings and available execution evidence. Codes join within one download; withdrawn participants are excluded. Missing model or prompt evidence remains unknown.

Archive project

Authorised readers retain access to records. Archiving is not deletion or an independent research approval.

Your participation

Your researcher cannot accept participation for you. Accepting an invitation does not automatically authorise adaptive selection. You can withdraw here.