Evidence · Work-record design · Student mirror
Teachers need more than a polished final answer. Own Mind is designed for one shared work record that keeps the student’s decisions, declared AI job, tests and revisions connected to the task. The aim is to make the work transparent without turning the student into data.
One project, made inspectable
Grade 8 Social Studies · one recommendation on unequal urban heat. Follow Act → Delegate → Verify → Defend from the student’s opening judgment to the final AI-off explanation. Framing and deciding remain visible human reasoning within the four-step method.
“I am prioritizing residents with the least ability to avoid indoor heat, not simply the hottest postcode.”
AI is off. The criterion and affected group must come from the student.
“Initial recommendation: fund cool roofs for the highest-risk social housing blocks first.”
A substantive first view is saved before AI can contribute.
“Cluster the supplied evidence by intervention, reach and stated limitation. Quote the source ID. Do not rank or recommend.”
AI gets one declared job — a role, a source boundary and a stopping point. It may organize the approved source pack; it may not choose the intervention or write the recommendation.
“The cluster overstates evidence for tree cover. Two sources model city averages and do not separate renters from owners.”
The machine return stays labeled. Confidence does not substitute for traceable evidence.
“Change: keep cool roofs as the immediate recommendation; add shaded public refuges, because the first option excludes some renters.”
The final judgment cannot be automated. “Hold” is a legitimate call when evidence is insufficient.
“I can explain the recommendation, the strongest counterargument, where AI contributed, and the evidence limit I would investigate next.”
The student returns later for an AI-off explanation and a changed-case transfer prompt.
The record centers the student’s decisions and explanation—not a score or automated authorship accusation.
Work transparency without student surveillance
The planned record is designed to let an assigned teacher inspect what was made, decided and revised, plus each material AI job and its outcome. Safeguarding content follows the school’s separate, restricted safeguarding process.
The design keeps starting points, AI contributions, tests, revisions and explanations with the exact assignment.
The planned student view shows the same work evidence in an age-appropriate form, including the human judgment attached to it.
Human judgment remains separate from generated grades, inferred traits, authorship accusations and predictions.
Student and teacher see the same work evidence
The planned record is designed to give assigned teachers and students the same work evidence in a clear view for each. Safeguarding follows a separate restricted process.
I prioritized residents least able to avoid indoor heat, not simply the hottest postcode.
Substantive first recommendation and criterion recorded before any machine contribution.
The design links periodic statements to defined progression, requires human review and keeps them open to challenge and visible to the student. One piece of work stays one evidence note; only a teacher writes a broader statement across different pieces of work.
“Across the access-guide and heat-risk projects, Amara moved from naming the visible symptom to revising the problem after listening to affected users. In the unfamiliar heat-risk task, she changed the frame when the evidence showed that the first proposal excluded renters.”
A clear purpose for each record
Evidence serves different educational jobs. Own Mind keeps those jobs distinct so a useful record does not quietly become an automated learner profile.
The exact starting point, contribution, test, revision and explanation for one declared project.
A teacher note tied to the assignment criteria and the work the student had the opportunity to do.
A teacher-written statement, reviewed by another adult and linked to work the student can question.
Selected work that the student can explain, export and use to show responsibility, not a feed of platform activity.
Research & testing
A design argument, an engineering check and a student outcome are different kinds of evidence. Own Mind keeps those layers separate when describing what is known, what is proposed and what still has to be tested.
Research on scaffolding supports checking what a learner needs, matching help to that need and reducing help over time. It does not show that Own Mind improves learning.
A tutor may hide the final answer while supplying the key idea, sequence, next move or judgment. Own Mind treats that possibility as a risk to check.
Repeatable tests check whether the product leaks an answer, supplies the structure, slips in a verdict or takes over the student’s voice. They find product failures before classroom research.
No study has yet compared Own Mind with another approach. A fair study must test what students can do later, without AI—not only what they complete while help is present.
Compare the task, not the polish
The finished answer is not the only thing AI can supply. It can also supply the frame, criteria, sequence and judgment. The comparison follows one task and makes those allocations visible.
An illustrative finished-answer request in regular ChatGPT, a Guided Learning route in Gemini, a teacher-configured Flint Sparky activity, and one bounded AI job inside Own Mind—the same brief, side by side.
Compare a direct-answer use of OpenAI’s ChatGPT, Google’s Guided Learning, Flint’s teacher-configured tutor and Own Mind’s bounded-work structure. The map describes responsibility in this pictured task; it does not score learning outcomes.
A direct study with students would be required before claiming an advantage over another product.
Evidence with a ceiling
Bring one real task and see how the planned record would keep the student’s decisions, AI contribution, test and explanation together.