AI-enabled measurement / organizational judgment
Can AI measure judgment — or does it mistake fluent reasoning for good reasoning?
Judgment Trace Workbench is an auditable testbed for validating AI-assisted measurement of human judgment in organizational decision tasks.
The problem
Final managerial recommendations can sound polished while hiding overreliance, cognitive outsourcing, or weak critique of AI advice.
The artifact
A trace-based system that captures initial judgment, AI exposure, acceptance/rejection, verification, final judgment, and confidence change.
The science
The core contribution is the validation of AI-assisted coding against human-coded judgment traces, including where AI coding fails.
Stage-1 status: this prototype demonstrates the pipeline, interface, constructs, synthetic traces, and audit logic. It does not claim measurement validity. Validation requires independently human-coded real traces.