INDRIYA — SOURCE-GROUNDED LEARNING: DRAFT EVALUATION PLAN
September 6, 2026
Status: proposed evaluation design.

QUESTION
Can a learner inspect a source-grounded explanation, identify an unsupported claim, and demonstrate their own understanding without disproportionate extra effort?

CURRENT ARTIFACT
A Learning Studio development prototype that connects selected source material, responses with passage references, recall questions, and review activities.

PROPOSED TASKS
1. Select permitted sources for a bounded question.
2. Find the passage supporting a specific explanation.
3. Identify a deliberately unsupported claim in an explicitly labelled test item.
4. Explain an idea without copying the source.
5. Revisit the idea in a later recall activity.

COMPARISON
Compare against the participant's existing reading, note-taking, and checking workflow. Fix source material and task difficulty. Record tool/model configuration where generation is used.

OBSERVATIONS TO COLLECT
- Whether the selected evidence supports the claim.
- Whether the participant finds unsupported or contradictory material.
- Corrections made before accepting the result.
- Task completion time and perceived effort, considered together.
- Quality of the participant's explanation under an explicit rubric.
- Reasons to abandon or decline to use a result.

STUDY DESIGN STILL TO SPECIFY
Recruitment, sample size, consent, task balance, evaluator blinding, scoring rubric, follow-up interval, stopping criteria, and handling of exclusions.

DATA BOUNDARIES
Use permitted materials, informed participation, and minimal necessary information. Do not enter sensitive personal or student records into a public demonstration.

INTERPRETATION
A citation is not proof of support. Fluent recall is not necessarily durable understanding. A prototype walkthrough can reveal usability problems but does not establish learning outcomes.

NEXT STEP
Refine the task and evaluation rubric with practitioners.
Contact: contact@indriyai.com