CAVA’s evaluation methodology is grounded in two published, peer-reviewed research papers. Both are open access and freely available.
Walcher (2026a) — The Framework
A Domain-Agnostic Framework for the Systematic Evaluation of AI-Generated Consumer Guidance
Universal Core Framework v1.0 · Registered Protocol / Methods Paper
This paper introduces the Universal Core Framework — the methodological architecture underlying every CAVA engagement. It defines the six evaluation dimensions and their anchored scoring descriptors, prompt complexity tiers and framing taxonomy, dual-track scoring and calibration procedures, behavioral signature detection rules, inter-rater reliability requirements, reproducibility and metadata standards, claim-strength governance, and evidentiary boundary restrictions.
The framework deliberately separates methodology from any single empirical deployment, enabling cross-domain comparability and independent replication.
Owen Walcher · Independent Researcher · Grand Junction, CO · May 2026
Licensed CC BY 4.0 · Available at Zenodo · https://doi.org/10.5281/zenodo.20511504
Walcher (2026b) — The Reference Implementation
Evaluating AI-Generated Consumer Guidance Across Six Applied Domains
This paper instantiates the Universal Core Framework across six empirical consumer domains: legal literacy, insurance navigation, home buying, consumer protection, home remodeling, and health advocacy. It documents a calibrated 270-prompt library, standardized response collection from five leading AI systems, dual-track evaluation with domain-expert review, and five pre-registered hypotheses.
Planned outputs include domain-specific performance scorecards, human-automated concordance analysis, a consumer-AI error taxonomy, behavioral signature patterns, and a reusable prompt library.
Owen Walcher · Independent Researcher · Grand Junction, CO · May 2026
Licensed CC BY 4.0 · Available at Zenodo · https://doi.org/10.5281/zenodo.20511747
Results
Results from the Reference Implementation study are expected in Q4 2026 and will be published at askafriend.com and submitted for peer-reviewed publication.