External learning research
Learning science establishes useful instructional principles. It does not prove that a particular English Compass AI implementation works.
EVIDENCE FIRST · LANGUAGE EDUCATION
The right question isn't whether an app looks impressive. It's whether students learn.
An academic partnership deserves a testable claim, a transparent method, and a clear boundary between independent research and results actually obtained with English Compass AI.
THREE LEVELS OF EVIDENCE
What we can demonstrate today
The public learner lab, interactive pilot planner and local practice-review tool exist. A university study using real learners, validated score growth, independent accessibility audit or active LTI integration has not yet occurred.
Working public preview
We can demonstrate sample reading, writing self-review, review scheduling and hypothetical faculty planning. Results are local and not an official score.
English Compass efficacy evidence
Not yet collected. A real, approved pilot would need standardized conditions, appropriate rubrics and transparent reporting.
REFERENCE LIBRARY
Instructional foundations, with their limits
Spaced study and retrieval
The U.S. Institute of Education Sciences practice guide recommends spacing learning and using quizzes that require retrieval.
What this does not prove: Supports designing a review schedule; does not validate our specific intervals or raise an individual student's grade.
Read official source ↗Transparent language descriptors
The Council of Europe CEFR descriptors can help faculty choose instructional goals and describe performance.
What this does not prove: CEFR-aligned goals do not equal an accredited language test or a recognized certificate.
Read official source ↗Human assessment and authentic work
Faculty-scored comparable tasks and clear performance criteria can document specific writing and reading changes.
What this does not prove: The pilot's chosen rubric and evidence validity still require academic review and, where applicable, an ethics protocol.
Read official source ↗Institutional interoperability
1EdTech LTI 1.3 and LTI Advantage describe the integration services institutions often require.
What this does not prove: Referencing this standard does not mean our platform is integrated, certified or available in any LMS.
Read official source ↗Accessible experiences
W3C WCAG 2.2 provides testable accessibility success criteria.
What this does not prove: Mentioning WCAG is not a conformance claim; independent assessment with assistive technology is still outstanding.
Read official source ↗PROPOSED STUDY PROTOCOL
A credible pilot must be able to fail.
We would propose an evaluation that defines success before seeing results. Faculty and institutions would approve the protocol, not accept a marketing promise.
Before launch
Define a learning objective, learner group, benchmark, accommodations, data permissions and a pre-specified decision threshold. Decide whether formal research or ethics approval is needed.
Baseline
Give a suitable reading or writing task and use a faculty-approved rubric. Document the starting distribution without using a public sample as an official CEFR exam.
Practice period
Schedule focused exercises and instructor feedback. Record only approved participation measures; note missing data, access limitations and teaching time.
Endline
Assign a matched task and score with the same criteria. When practical, use second raters or de-identification. Do not claim causality from a weak comparison.
Review and publish
Share results, attrition, adverse feedback, confidence and costs, including null outcomes. Agree whether to continue, revise, or stop.
PRE-PILOT STATUS
An evidence scorecard with blank cells is more credible than invented gains.
| Outcome | Recommended evidence | English Compass status |
|---|---|---|
| Reading comprehension | Human-rated improvement on comparable texts | Not yet measured |
| Written communication | Faculty rubric and revision evidence | Not yet measured |
| Participation | Completion, time on task and dropout | Not yet measured |
| Access and accessibility | Accommodations, device and assistive-technology audits | Not yet verified |
| Faculty load | Time to assign, support and score | Not yet measured |
| Institutional cost | Real hosting, support, and AI expenses per active learner | Not yet measured |
SAFETY AND HUMAN OVERSIGHT
IT readiness ↗Data and human review come before automation
The institutional inquiry form should contain department and pilot information only. Identified learner records, classroom analytics, grade exports and AI voice data must wait for a dedicated secured service, appropriate contracts and a verified retention/access model. Human judgment stays central for consequential grading.
Want to design an evaluation with your faculty?
Use the faculty studio to draft an instructional plan, or request a scope conversation. No student enrollment or charges are created.