Pedagogical character models
Examples are deliberately designed configurations. They are not observations of real teachers. A configuration review is not proof of educational effectiveness.
Parallel conversations, research protocols and controlled course activation.
Checking your GRITT account…
Platform administrators can create independent projects. An institution administrator with member-management permission can create a project for that institution.
Examples are deliberately designed configurations. They are not observations of real teachers. A configuration review is not proof of educational effectiveness.
Your profiles belong to the selected project and your account. Save versions here, then select a fixed version in a conversation comparison. Saving a profile does not activate it in a course.
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.
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.
ADEM · research access
A uses ADEM off. Add one to three ADEM profiles, for example the defaults and two different settings. Every condition uses the same fixed PCM and course material, with separate conversations. Added profile copies are fixed when the comparison is created.
This choice is fixed for this comparison. A keeps the existing route with ADEM off. Every condition uses the normal chat and response service; this does not measure a learning effect.
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.
This ADEM conversation comparison requires current research access.
Each repeat runs every fixed condition and closes only after verified publication. One to five repeats with two to four conditions require two to twenty initial responses; the existing observer and follow-up checks may use additional model calls. Responses may differ; the fixed condition and environment remain inspectable.
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.
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.
A fixed protocol records the agreed study design. Saving it does not enrol participants or start lessons.
No studies yet. To explore teaching models and source analyses, use Teacher. Create a study here when you want to investigate learning outcomes.
Participants accept their own invitation under Access & audit. Allocate episodes using the fixed protocol; a baseline is required by this protocol before activation.
For a single episode or an explicit personal observation. Use the participant group above for reviewing and allocating a batch.
Release only the selected episode’s PCM assignment without declaring the lesson finished or changing recorded outcomes. Use Pause or Stop above for normal episode control.
Readiness reflects the last server read. Account access, current assignment, course connection and protocol are checked again by the server; this display is not permission to run a lesson.
Participants confirm their GRITT lesson using its verified PCM publication receipt. For externally delivered lessons, record the actual reference here; this remains a human declaration. Operator-entered scores are labelled human assessments. Automatic scores come only from submitted assessment attempts.
Select an episode to identify the participant.
This uses the participant in the selected episode. Parallel groups do not establish a within-person comparison; an explicit participant preference remains distinct. Applying a proposal requires their opt-in.
Automatic personal matching is available for PCM-comparison studies. This study uses fixed ADEM conditions; choose their allocation through the saved protocol.
Closing releases this study’s active assignments. Existing outcomes and missing results remain recorded.
Choose the models and measurement method here. Saving fixes this protocol version so the criteria do not silently change after the results are known.
The server copies the fixed conditions and version pins from your own saved comparison. It does not run it. Source-derived drafts require a simulated study with accepted practice accounts; saving a protocol never approves a model.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
Authorised readers retain access to records. Archiving is not deletion or an independent research approval.
Your researcher cannot accept participation for you. Accepting an invitation does not automatically authorise adaptive selection. You can withdraw here.
Open this conversation in GRITT
Starting this lesson selects its assigned course in your GRITT account.
Starting this lesson selects its course in your own Education account and binds this conversation to that course. Use a separate conversation to work with another course. Normal GRITT usage and learning-progress recording apply. Your researcher receives delivery evidence and assessment results; this page does not share your complete conversation.