ABOUT THE VALEHART PROJECT
AI systems behave differently in real-world use than in controlled environments. Valehart analyses that gap. Rather than evaluating models in isolation, we analyse AI x human interaction patterns in real-world use.
  • How decisions propagate,
  • Where breakdowns occur
  • How responsibility shifts across users, systems, and institutions.
The goal is to make these systems more observable, auditable, and easier to reason about in practice. We are currently developing TRACE, an interaction-forensics platform that maps how AI-assisted decisions move through these layers. It focuses on provenance, auditability, and identifying failure patterns in AI-enabled workflows, rather than relying on intent inference or abstract scoring.
Case Studies Applied AI x Human use across domains: Farming, historical research, physical builds, fashion.
Case Studies Applied AI x Human use across domains. View
HCI Understanding interaction patterns: How AI behaves with humans in real-world use.
HCI Understanding interaction patterns in real-world use. View
Services Pattern detection engine: Mapping connections across users, systems, institutions, narratives.
Services Pattern detection and forensics across AI-enabled workflows. View
Upcoming Projects Future work and research directions currently in scoping.
Upcoming Projects Future directions currently in scoping. View