JRS Review Conditions
The five conditions themselves: reconstructability, basis identification, chronology, decision-process traceability, evidentiary sufficiency. What each one asks and what a failure looks like.
Six self-paced modules and a companion desk reference for deciding whether a record can still explain its own conclusion, particularly where AI-assisted drafting has contributed to it. Built for compliance, HR, audit, legal, AI governance and investigative reviewers. Module 1 opens with no sign-up.
Written by Phillip Wikes, former Lead Civil Rights Officer, Maryland Commission on Civil Rights, M.S. Negotiation and Conflict Management. The research samples must be distinguished: 16 independent detection-panel reviewers each examined 24 constructed records, while a separate reliability analysis used 104 determinations from 22 analyzed raters. These findings do not establish operational effectiveness. Read the study results and limitations.
Each module ends in a short applied exercise. Progress saves in your browser, so you can stop and come back. The full set takes most reviewers two to three hours.
The five conditions themselves: reconstructability, basis identification, chronology, decision-process traceability, evidentiary sufficiency. What each one asks and what a failure looks like.
The patterns that recur across real records: conclusions without a stated basis, chronologies that cannot be followed, and characterizations with no visible source.
What a second reviewer needs from a file, when a record should be returned to the drafter, and how to write an escalation that survives being read a year later.
Reviewing records that a model helped draft. Polished prose can hide missing evidence and untraceable reasoning, and the same five conditions still apply.
Applying the conditions to investigative files: interview records, evidence handling, and whether the finding can be rebuilt from the file alone.
Putting pre-finalization review into an existing workflow without adding a stage, and what to report to leadership.
Every figure is recomputed from the study database rather than transcribed. The two reported as interim have not met their pre-registered secondary criterion and say so. Full detail on the research and validation page.
No sign-up is required. All six modules are accessible now, and a name is optional if you choose to generate a personal participation record.
Open the trainingFree access · No credit card · All six modules open straight away