Evidence reviews and scorecards · Complete web edition
Phase Seven Public Evidence Archive
This is the complete web edition of the canonical project document. Its narrative, recommendations, evidence tables and source links are presented below without abridgment.
Project, Research, Analysis, Editorial by Patrick Russell; assisted by ChatGPT (OpenAI).
California Education and AI Public Evidence Archive
Phase Seven | Baseline sources and selected review units
Baseline date: September 12, 2026 • Informational working document • Public evidence only
Responsible person | ________________________________________ |
|---|---|
Organization or role | ________________________________________ |
Review date | ________________________________________ |
Phase decision
The investigation is ready to move from design to a dated public-evidence baseline. All ten California institutions remain in scope. The comparison uses separate K through 12 and higher-education tracks, records each institution's publicly stated educational purpose, and tests whether AI practice supports or displaces the formation of attention, memory, reasoning, judgment, communication, agency, and relationships.
Mother of Divine Grace School is classified only as a K through 12 private Catholic distance-learning and home-education program. It is not a college or university. Thomas Aquinas College is a separate four-year college comparison.
- No information-request packet or direct outreach is included. This phase relies on official public pages, catalogs, handbooks, policies, course materials, board records, and published student work.
- Silence is not a policy. When a public AI statement cannot be located, record the result as a public-evidence gap rather than inferring permission, prohibition, or indifference.
- Institutional guidance and actual classroom practice are separate evidence layers and should not be treated as equivalent.
Baseline evidence standard
Field | Required record |
|---|---|
Identity | Institution, educational level, public/private status, document title, publishing unit, and canonical URL. |
Time | Publication or revision date when shown; access date; academic year; and whether the source is current, archived, or superseded. |
Authority | Institution-wide policy, division/school guidance, course rule, program description, board action, or illustrative practice. |
AI scope | Who may use AI, for what task, with what disclosure, data limits, approved tools, age/consent rules, and human verification. |
Formation claim | The educational capacity the source says it protects or develops: knowledge, wisdom, reasoning, writing, creativity, agency, relationship, ethics, or vocation. |
Preservation | Save page or PDF, title, URL, access date, visible version/date, and a change note. Where feasible retain a screenshot or PDF and a checksum. |
Limitation | State what the source cannot establish, especially implementation, student learning, independent transfer, mental fitness, or causal impact. |
Evidence and AI-condition codes
Code | Meaning | Use |
|---|---|---|
C | Current | Official source appears current at baseline date. |
A | Archived | Historical source retained for comparison. |
Δ | Changed | Material wording, tool, scope, or governance change found. |
M | Missing | Required public evidence not located. |
S | Superseded | Newer official source controls. |
R | Restricted | Known source is not publicly accessible. |
Condition | Definition | Formation question |
|---|---|---|
H0 | Protected human performance: no AI during the task. | What can the student independently remember, reason, create, explain, or defend? |
H1 | Teacher-facing AI only; students complete the learning task without AI. | Does instructional support improve learning without outsourcing student cognition? |
S1 | Bounded student AI use with stated purpose, limits, disclosure, and verification. | Does AI provide feedback or extension while preserving productive struggle and authorship? |
S2 | Open student AI use within general integrity rules. | What is gained, and what independent capability may be obscured or lost? |
Ten-institution baseline and review units
Institution | Type/level | Public baseline | Selected review units | Primary test |
|---|---|---|---|---|
UC Berkeley | Public university | C — Central AI Hub and teaching guidance; broad course and program ecosystem. | AI Hub teaching resources; one writing/general-education course; one computing/data course; protected oral or in-person verification. | Compare institutional enablement with course-level limits and independent mastery. |
UCLA | Public university | C — Six teaching-and-learning principles; student AI literacy, faculty fellows, workshops, research-led adoption. | AI Literacy Module; one Faculty Fellows course; one matched course with limited or no AI. | Test equity, literacy, assessment integrity, and evidence behind adoption. |
Stanford University | Private university | C — Course-specific rules; default caution; disclosure; AIMES examples; emphasis on deep engagement. | One AIMES example permitting/bounding AI; one AIMES prohibited assessment; oral or unaided transfer task. | Compare stated learning purpose with the work students still perform themselves. |
University of Southern California | Private university | C — ChatGPT Edu available; faculty retain course-policy control; privacy, integrity, and training emphasized. | Two courses with contrasting AI permissions; ChatGPT Edu training or AI-in-Education resource; disclosed student workflow. | Assess equitable access, faculty governance, assessment validity, and secure use. |
California Institute of Technology | Private university | C — Central guidance; instructor/division rules; discourages AI detectors; public Ethics & AI course case. | First-year Ethics & AI writing sequence; technical course; proctored/oral/process verification. | Examine whether AI integration preserves writing, voice, trust, and technical reasoning. |
Loyola Marymount University | Private university | C — Intentional decisions, academic honesty, supported tools, privacy, alternatives, literacy, and faculty examples. | Core writing/ethics course; professional or computing course; co-constructed policy example. | Test human-centered and ethical claims against disclosure, access, and independent judgment. |
Thomas Aquinas College | Private four-year college | C — Great Books/liberal education; seminar, tutorial, laboratory, composition; public AI and Human Flourishing symposium. | Seminar; tutorial; laboratory; composition; senior thesis; AI symposium materials. | Determine how dialogic first-principles formation addresses AI without confusing an event with curriculum-wide policy. |
Mother of Divine Grace School | Private Catholic K–12 distance/home education | C — Classical method from wonder toward wisdom; parent-supported stages: K–2, 3–5, 6–9, 10–12. Public AI policy not established in baseline. | K–2 narration/recitation; grades 3–5 memory and observation; grades 6–9 logic/composition; grades 10–12 rhetoric/Great Books; parent role. | Study developmental formation and family mediation; record AI as M unless an official policy is found. K–12 only. |
Harvard-Westlake School | Private grades 7–12 | C — 2026–27 curriculum includes an ethics, technology, and society course in the age of AI; broader policy evidence still to capture. | Middle-school writing/research task; upper-school AI ethics/technology course; senior independent study. | Compare AI literacy and ethics with unaided writing, discussion, research, and authorship. |
High Tech High | Public charter K–12 network | C — Board-approved AI policy referenced in records; teacher guidance; design-principle alignment; classroom examples; K–12 handbook reference. | Elementary, middle, and high project/exhibition; AI-bias project; historical-figure design example; individual defense inside group work. | Test whether AI strengthens equity, personalization, authentic work, and collaboration while preserving individual mastery. |
Case-file protocol
For each selected review unit, construct one case file using the same sequence. The purpose is not to rank schools from website language; it is to connect claims, rules, learning activity, and independently demonstrated outcomes.
- 1. Identify the educational aim and the student capacity the unit is intended to form.
- 2. Record the assigned work before AI enters: reading, observation, memory, calculation, drafting, dialogue, experimentation, or practice.
- 3. Code each stage H0, H1, S1, or S2 and record tool, purpose, disclosure, privacy, access, age/consent, and teacher oversight.
- 4. Preserve the prompt, rubric, syllabus language, assessment format, revision history, and public student artifact where available.
- 5. Identify a comparable protected task that shows what the learner can do after assistance ends.
- 6. Record relationship effects: parent, teacher, peer, adviser, counselor, tutor, community expert, and audience contact.
- 7. Rate evidence strength and note alternative explanations. Do not infer mental fitness, wellness, or causation from a policy page alone.
COVID, family, friendship, and mental-fitness comparison
COVID school closures should be a named historical period, not a footnote. They changed instructional access, parent burden, peer contact, teacher contact, digital exposure, routines, assessment, and the reliability of ordinary school records. The comparison should use pre-COVID, disruption, recovery, and AI-era windows where data permit.
Window | Approximate period | What to document | Interpretive caution |
|---|---|---|---|
Pre-COVID | Before spring 2020 | Attendance, achievement, belonging, counseling/help-seeking, family routines, device use, and instructional format. | Historical measures and populations may not be directly comparable. |
Disruption | 2020–2021, locally defined | Closure duration, synchronous/asynchronous access, parent teaching burden, lost instruction, isolation, grief, economic stress, and special-needs services. | Exposure varied sharply by institution, household, age, disability, and local rules. |
Recovery | 2021–2022/23 | Return mode, unfinished learning, attendance, behavior, friendship repair, counseling demand, teacher turnover, and family strain. | Recovery timing differs; avoid a single statewide cutoff when local evidence exists. |
AI era | Late 2022 onward | AI access and rules, assessment redesign, attention and effort patterns, authorship, trust, workload, relationships, and independent transfer. | Do not attribute persistent COVID effects to AI or AI effects to COVID without a defensible design. |
Mental fitness should be operationalized as observable capacities—not a diagnosis: sustained attention, frustration tolerance, self-regulation, memory, reasoning persistence, confidence calibrated to performance, help-seeking, recovery after error, and capacity for friendship and cooperative work. Any clinical mental-health measure requires validated instruments, privacy protection, consent, and qualified interpretation.
Baseline findings
- The public and private institutions are comparable only through common formation questions; their missions, ages, governance, delivery models, and evidence availability differ materially.
- Several universities publish mature AI guidance, tools, training, and course-level discretion. This demonstrates governance activity, not student learning by itself.
- Caltech supplies an unusually useful public course case because it connects AI permission, writing practice, disclosure, and instructor reflection; it still requires independent-outcome evidence.
- Thomas Aquinas College publicly frames AI through human flourishing and first principles. The event should be studied alongside—not substituted for—its standing curriculum and assessment practice.
- Mother of Divine Grace supplies a detailed developmental account of forming the mind across K through 12. No institution-wide public AI position was established in this baseline, so the investigation should preserve that as a gap.
- High Tech High has a board-referenced AI policy and classroom examples tied to its design principles. The actual policy text, annual review, grade-level implementation, and independent mastery are priority captures.
- A reliable comparison needs protected human-performance evidence. AI availability, polished artifacts, and policy sophistication cannot substitute for what students can explain or do independently.
Priority gaps and next-phase capture
Priority | Needed evidence | Disposition |
|---|---|---|
1 | Exact institution-wide and unit-level AI policy texts, revision dates, approved tools, privacy rules, and age/consent requirements. | Archive official copy; compare monthly for material change. |
2 | Representative assignment, rubric, permitted-use statement, student artifact, and protected follow-up assessment for each case. | Build institution case files from public evidence. |
3 | Implementation evidence showing how faculty, teachers, parents, and students understand and apply the rule. | Use published training, FAQs, minutes, evaluations, and studies; label gaps. |
4 | Pre-COVID through AI-era indicators for learning, attendance, belonging, family burden, friendship, attention, and support use. | Use comparable institutional, state, and validated research sources. |
5 | Evidence of independent transfer after AI-supported work. | Treat as the central outcome; do not infer from final-product quality. |
Official public source register
UC Berkeley — AI Hub — Teaching
UC Berkeley — Center for Teaching & Learning — AI in Teaching & Learning
UCLA — Generative AI in Teaching and Learning
Stanford — AI Learning Guide for Students
Stanford — Creating Your Course Policy on AI
USC — Faculty Guidance: ChatGPT Edu rollout
USC — ChatGPT EDU at USC
Caltech — AI in Teaching
Caltech — Teaching Ethics & AI in the Wake of ChatGPT
LMU — AI in Teaching & Learning
Thomas Aquinas College — AI & Human Flourishing
Thomas Aquinas College — Order of Learning
Mother of Divine Grace — Curriculum Methodology
Mother of Divine Grace — Curriculum Content and Stages
Harvard-Westlake — Course Descriptions / 2026–27 Curriculum Guide
High Tech High — Design Principles and School Model
High Tech High — 2025–26 Student & Parent Handbook
Source note: all entries were accessed for this baseline on September 12, 2026. A live page may change after access. The archive record should retain the retrieved version and record later changes rather than silently replacing the baseline.
Phase Eight handoff
The next phase should build ten institution case files, beginning with one high-information review unit per institution. Each file should preserve the source set, map the AI condition, identify a protected human-performance measure, capture COVID/recovery context where relevant, and state what remains unknown. The monthly policy monitor should update the change log but should not overwrite this dated baseline.