The first published framework for measuring what students actually gain from AI learning tools.
Seven critical dimensions of student growth that standardized tests cannot capture — from problem decomposition and self-directed learning to AI judgment and verification habits. Measured through direct behavioral evidence, not surveys or self-reports.
Created by Lisa Russell, M.A. | CEO, Instant Tutor LLC | 30+ years in education and special education support
Schools are adopting AI tools faster than they can measure their impact. This framework closes that gap.
Standardized tests measure what students know at a single point in time. They cannot measure how students think, question, verify, or self-direct their learning.
When students use AI tools, new behaviors emerge — problem decomposition, critical evaluation of AI answers, independent verification. No existing test captures these.
Districts investing in AI learning tools need evidence that students are developing durable learning habits — not just getting faster answers. This framework provides that evidence.
Each dimension captures a critical aspect of learning behavior that standardized testing fundamentally cannot measure.
Transparent scoring based on observable student behaviors, not surveys or self-reports.
Growth Index = weighted composite of 7 behavioral dimensions. Inputs include session logs, tool interaction events, difficulty progression tracking, and verification behavior patterns. Scores are normalized 0-100; trend labels require a minimum of 4 sessions.
Existing platforms track surface-level metrics. This framework measures the behavioral dimensions that actually indicate learning.
| Measurement Capability | Traditional LMS | Adaptive Learning | AI Tutoring Platforms | This Framework |
|---|---|---|---|---|
| Test scores / quiz completion | Yes | Yes | Yes | Yes |
| Time on task / session duration | Yes | Yes | Yes | Yes |
| Content completion rates | Yes | Yes | Yes | Yes |
| Problem decomposition behavior | No | No | No | Yes |
| Critical AI consumption habits | No | No | No | Yes |
| Cross-subject knowledge transfer | No | No | No | Yes |
| Self-directed learning measurement | No | Partial | No | Yes |
| Confidence trajectory tracking | No | No | No | Yes |
| Question quality progression | No | No | No | Yes |
| Verification instinct measurement | No | No | No | Yes |
| IEP-adaptive scoring | No | Partial | No | Yes |
| Cohort-level behavioral analytics | No | No | No | Yes |
Based on publicly available product documentation as of April 2026. "Traditional LMS" includes Canvas, Blackboard, Google Classroom. "Adaptive Learning" includes DreamBox, Khan Academy, IXL. "AI Tutoring" includes Khanmigo, Duolingo Max, Photomath.
This dashboard shows a seeded demonstration cohort of 23 student profiles used to illustrate how the 7-dimension framework reads on a fully populated cohort across grades 3-10, including IEP profiles. The live student data feed below uses the same framework against actual Diagnostic Assessment sessions as they arrive.
Compare growth profiles across multiple students to identify patterns and opportunities.
| Dimension |
|---|
The framework produces school-level and cohort-level analytics, not just individual reports. The numbers shown here come from the seeded demonstration cohort above and illustrate the same aggregate views the live data feed below produces against actual Diagnostic Assessment sessions.
This panel is connected to the live Diagnostic Assessment feed and refreshes automatically as students complete sessions in the Student Portal. Seven measurement dimensions are shown day-one. Diagnostic Assessment feeds three of them today; the remaining four populate as additional AI tools are wired in.
| Student (pin) | Subject | Sessions | Adaptive Reach | Question-Pace Signal | Growth-Path Velocity | Platform-Inferred Level (Δ) | Hint Reliance | Conceptual Stickiness | Effort Consistency | Independence Index | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| — | |||||||||||
Columns marked in amber are placeholders for AI tools not yet wired into outcome tracking. They populate once each tool ships its save hook — same pattern as Diagnostic Assessment.
Platform-Inferred Level is derived from diagnostic accuracy adjusted for pace and effort (hints used, questions skipped, pace setting). It is a directional signal designed to surface change over time for tutor judgment — it is not norm-referenced and is not a substitute for standardized assessment such as NWEA MAP, DRA, Lexile, or STAR.
This panel is connected to the live Subject Test Generator feed and refreshes automatically as students generate and take subject-specific practice tests in the Student Portal. Per-student rows are staff-visible only; public visitors see the framework + aggregates but not per-student row data. Uses the same staff sign-in as the Diagnostic Assessment panel above.
| Student (pin) | Subject | Level | Tests Generated | Auto-Graded | First Score | Latest Score | Avg Score | Δ (latest − first) | Time on Task | |
|---|---|---|---|---|---|---|---|---|---|---|
| — | ||||||||||
Diagnostic Assessment, Subject Test Generator, Essay Assistant (Grammar Check + Style Analysis), the IEP Goal Generator, and the AI Study Companion are the AI student tools quantified into outcome tracking here. This page measures raw learning signals only — the AI reports those tools generate live in the Student and Parent portals, not here. Additional AI student tools wire in one at a time over the coming weeks, each following the same pattern.
This panel is connected to the live Essay Assistant Grammar Check feed and refreshes automatically as students run grammar checks in the Student Portal. We do not store a numeric grammar score (the tool returns written AI feedback, not a grade); growth is read from how essay length and estimated reading level change across repeated checks. Per-student rows are staff-visible only; public visitors see the framework but not per-student row data. Uses the same staff sign-in as the panels above.
| Student (pin) | Grade | Checks Run | First Length | Latest Length | Avg Length | Δ Length | Latest Reading Level | Δ Reading Level | Last Activity | |
|---|---|---|---|---|---|---|---|---|---|---|
| — | ||||||||||
Essay Grammar Check is one of the AI student tools wired into outcome tracking. Length and reading-level trends are honest, computed signals — no AI-invented score.
This panel is connected to the live Essay Assistant Style Analysis feed and refreshes automatically as students run style analyses in the Student Portal. As with Grammar Check, no numeric style score is stored (the tool returns written AI suggestions, not a grade); growth is read from how essay length and estimated reading level change across repeated analyses. Per-student rows are staff-visible only; public visitors see the framework but not per-student row data. Uses the same staff sign-in as the panels above.
| Student (pin) | Grade | Analyses Run | First Length | Latest Length | Avg Length | Δ Length | Latest Reading Level | Δ Reading Level | Last Activity | |
|---|---|---|---|---|---|---|---|---|---|---|
| — | ||||||||||
Style Analysis is one of the AI student tools wired into outcome tracking. Length and reading-level trends are honest, computed signals — no AI-invented score.
When a tutor generates IEP goals or a full IEP draft for a student in the Student Portal, the goal area and date are recorded here as a de-identified timeline — so you can see which special-education goal areas a student is working on, and how that focus shifts over time. Buyers and staff see the goal-area timeline; the disability category and the generated goal text appear to signed-in staff only.
The IEP Goal Generator timeline shows goal areas and dates only — sensitive special-education detail stays staff-only, and the saved goal text is contact-scrubbed.
Every question a student asks the AI Study Companion is logged as an engagement event — no question or answer text is stored, only that a learning interaction happened and a best-effort topic. This gives an honest engagement picture per student: questions asked, distinct study sessions, active minutes, and how long they have been using the companion. Buyers and staff see the engagement aggregates; student names appear to signed-in staff only.
The AI Study Companion engagement view uses a 30-minute inactivity gap; active minutes measure the time between the first and last question within a session, so a single-question session honestly shows zero active minutes.
When the AI builds a personalized curriculum for a student in the Student Portal, the modules it creates are recorded automatically. As the student works through and checks off each module in the portal, completion is logged here — giving an honest progress picture per subject: modules assigned, modules completed, and the module they are on now. Buyers and staff see the module counts and progress; the individual module titles and student names appear to signed-in staff only.
Module counts are taken from the curriculum's own content modules — no number is entered by hand and no module is marked complete unless the student checks it off in the portal.
Estimated Growth Index trajectory across the 8-week measurement period, modeled from final scores and trend data for each student. Students with shorter measurement windows (6-7 weeks) are excluded from later weeks.
Nine AI-powered tools generate behavioral interaction data. The framework processes that data into 7 scored dimensions at student, classroom, and cohort levels.
Five interlocking assets that constitute a measurement capability not currently available from any single vendor.
7-dimension behavioral measurement model with scoring methodology, trend analysis, and IEP-adaptive scoring. First-mover advantage in AI learning measurement.
9-tool data pipeline that captures an estimated 500+ behavioral events per student. Processes raw interactions into scored, trended, evidence-backed dimension reports.
Seeded 23-student demonstration dataset with complete behavioral records, 8 weeks of illustrative longitudinal data, 4 IEP profiles, grades 3-10 — used to show the framework reading on a fully populated cohort. Available for independent review or replication. The live student data feed below uses the same framework against actual Diagnostic Assessment sessions.
The 7 dimensions are mapped across core K-12 subjects (math, reading, science, social studies, writing) and adapt to IEP, EL, and grade-band variation so the same measurement vocabulary works district-wide.
Three-level output system (student, classroom, cohort) with API-ready backend, white-label capability, and multi-tenant support path.
Combined: A complete, published, validated K-12 measurement capability ready for district, school, and tutoring-program deployment, with first-mover positioning in the AI learning measurement category for education.
A comprehensive measurement approach that serves every stakeholder in the educational ecosystem.
Students gain visibility into their own learning behaviors, building metacognitive awareness and intrinsic motivation beyond grades.
Parents receive meaningful insight into their child's learning trajectory instead of a single letter grade that says nothing about growth.
Teachers get actionable data on how students learn, not just what they know, enabling truly differentiated instruction at scale.
Track AI implementation impact by school, grade band, and subgroup to guide instruction, budgeting, and intervention planning with real evidence.
Translate AI adoption into board-ready evidence: growth trends, equity signals across IEP and general education, and measurable progress toward strategic goals.
Request a briefing on implementing the growth measurement framework in your schools. Includes launch planning and board presentation support.
Request an Implementation BriefingSee how the 7 dimensions map to your classroom. Explore the Student Portal with 9 AI-powered tools that generate growth data.
Explore the Student PortalLicense the measurement layer for your platform. Demonstrate real learning outcomes to your customers with a proprietary growth framework.
Discuss Licensing