A measurement framework for the reliability of LLM-based agents in regulated environments: hallucination detection, uncertainty quantification, and operational trustworthiness. Developed through Wade Lovell's doctoral research at Walsh College, the CRI directly informs every product in the CognitionHive portfolio.
Agent outputs are scored against verifiable ground truth, claim by claim, so accuracy is measured rather than assumed.
Outputs stay grounded in their sources and evidence, with hallucinated content detected and quantified.
Consistent results across runs, prompts, and model versions, so behavior changes register as signal rather than noise.
Policy, regulatory, and format constraints are enforced and audited on every output.
The framework extends into sub-indices for specific reliability properties. The Explainability Performance Index (EPI), implemented directly by Prufs.ai, measures whether agent reasoning is auditable at the decision level, not just the output level.
A practical framework for educators integrating AI into K-6 classrooms. Addresses how schools can implement AI systems that promote student agency, build confidence in AI literacy, and maintain equity across diverse learning environments.
The research-driven companion to the classroom practice book. Presents the evidence base for why early AI education matters and why the current approach of restricting student access to AI tools creates more risk than it mitigates.
A direct examination of executive-level AI adoption: the organizational, strategic, and governance decisions that determine whether AI investments create value or compound risk. Written for boards, CISOs, CROs, and technology leaders in regulated industries.
Walsh College. Research focuses on developing the Claim Reliability Index (CRI), a measurement framework for hallucination detection, uncertainty quantification, and operational trustworthiness of LLM agents in regulated settings. This research directly informs every product in the CognitionHive portfolio.
Every product in the CognitionHive portfolio is built on the Claim Reliability Index: trust infrastructure that signs and audits agent decisions, agentic services whose outputs are verified before delivery, wealth intelligence with explainability on every recommendation, and education measured against the framework's four dimensions.
Wade Lovell is principal of CognitionHive. His AI work began in a doctoral seminar in Econometrics. Across his career he has served as CTO, Technical Architect, and AI Architect in banking, insurance, and enterprise software. His doctoral dissertation at Walsh College focuses on LLM agent reliability in regulated settings. He is a published author on Ethical AI in Education, the Workforce, and the C-Suite.