Governed Evidence Infrastructure
SYMCUE's credibility rests on a rigorous evidence framework. We track evidence maturity deliberately—from conceptual constructs through implemented infrastructure toward independent external evaluation.
Evidence Governance Disclaimer
This is evidence infrastructure and governance — it is not scientific validation. Validation claims require completed external review and empirical study, and SYMCUE does not make them ahead of the evidence.
The Evidence Framework
A governed architecture built for traceability and continuous scrutiny.
Canonical framework
Dimensions, competencies, and observable behaviors are defined once, in a governed register — not restated ad hoc.
Item-to-behavior traceability
Every assessment item maps to the observable behaviors it measures.
Course Evidence Packages
Each learning pathway documents what it teaches and which competencies it targets.
Approved-source registry
Documents are generated from registered sources; citations are never invented.
Expert-review workflow
A structured process for external reviewers to examine instruments and findings.
Pilot-data governance
Consent, de-identification, and pre-declared analysis plans before any study data is collected.
Automated claim controls
Generated documents are machine-screened so claim language cannot exceed the evidence.
Evidence Maturity Path
Capabilities and constructs within SYMCUE are tracked on a formal maturity scale. We are transparent about the current state of every capability within the platform.
Conceptual
Defined construct and expected behaviors. No measurement implemented yet.
Implemented–Developmental
Assessment and infrastructure built. Engaging in baseline pilots to evaluate psychometric properties. SYMCUE never claims any capability is beyond this stage.
External Evaluation
Undergoing or has completed independent academic evaluation to confirm validity, reliability, and generalizability.
What We Know
- The platform is operational.
- Version 2 framework definitions are governed.
- Assessments, learning pathways, practice, and prescriptions are traceable.
- Claim language is machine governed.
- Pilot and external-review infrastructure exists.
What We Are Testing
- Reliability
- Construct validity
- Alternate-form equivalence
- Instructional effectiveness
- Retention
- Transfer
- Operational impact
- Cross-population generalizability
External Context: Readiness, Verification and Accountability
These July 2026 sources inform our design rationale. They are not evidence of SYMCUE efficacy, score validity or predictive accuracy. Surveys, editorial proposals and guidance must not be treated as controlled performance studies.
- Adoption is not demonstrated readiness. In a commissioned survey of 80 senior leaders in the US and UK, 3% described their leaders as highly prepared for AI-enabled work and 17% reported advanced or transformational workforce readiness. These are self-reports, not objective readiness tests or causal productivity results. ManpowerGroup / Everest Group, July 22 announcement.
- Declared proficiency needs a separate performance check. Ipsos's release reports 32% of respondents admitted overstating AI skills to colleagues. The underlying TD survey interviewed 2,501 Canadian adults in February 2026; this is not a measured competence deficit or a prevalence estimate for other workforces. Ipsos, July release.
- Examine the process, not only the product. Zhenguo Yuan's editorial proposes recording AI-use disclosures, prompts, model versions and human interventions, alongside substantive verification and responsibility. This is a conceptual assessment rationale, not proof that interaction logs establish understanding. ECNU Review of Education editorial, July 25.
- A practical verification habit. NC State guidance featuring Joey Huang recommends Pause, Identify, Trace, Compare and Judge: form an initial view, flag claims, inspect original sources, compare them with AI's account and own the final judgment. This is professional guidance, not an efficacy trial. NC State, July 30.
- Trust and decision outcomes are different measures. Yang, Zhu and Rau's AHFE experiment (36 participants) found stress-related trust and neural differences, but no significant difference in decision results. Constrained tasks and text-based stress induction limit interpretation; this does not establish an emotional-regulation benefit from SYMCUE. AHFE proceedings, online July 20.
Academic Review Package
A governed, downloadable reviewer package exists for qualified academic reviewers. We welcome external scrutiny of our instruments and intend to partner with universities and operational organizations for empirical evaluation.