One identified decision. Two independent examinations.
When an AI-assisted decision requires retrospective scrutiny, examine both the route that shaped it and the human judgment exercised within it.
A bounded post-decision review. Distinct findings. Controlled bilateral comparison. No merged verdict.
A recorded approval is not the entire decision.
A consequential outcome may be formally authorised while AI has already shaped what was seen, which alternatives seemed credible, or how much reliance was placed on a recommendation.
What became visible?
How did technical outputs influence framing, prioritisation and the evidence considered?
What did the human actually do?
Was there a meaningful opportunity to question, reconsider or intervene?
What can be reconstructed?
Which records support an accountable explanation, and which uncertainties remain?
ADIR examines an identified AI-assisted decision under internal review, challenge, dispute or comparable scrutiny. A harmful incident is not required.
Different questions. Independent professional findings.
Two separately accountable perspectives examine the same bounded decision context without combining their conclusions into a single verdict.
Decision architecture & accountability
Reconstructs the decision route: AI influence, evidence, formal and effective authority, escalation and practical intervention.
- Formation and decision-state reconstruction
- Authority and accountability
- Evidence timing and sufficiency
- Challenge and intervention pathways
Behavioral science & human reliance
Examines human judgment and the reasonableness of AI reliance in the context of what the decision-maker was shown.
- Comprehension of AI output
- Reliance, credibility and judgment
- Environmental considerations of oversight
- Biases in human reasoning and decision-making
Comparison without loss of independence.
Independently attributable findings are compared to identify convergence, divergence, evidentiary asymmetry and unresolved questions.
No merged score, compulsory consensus, joint certification or predetermined finding. The bilateral interface has been internally stress-tested; this is not external validation.
A common case. Independent findings. Controlled comparison.
Structural finding
ID∆AC™ examines the decision route, practical authority and supporting evidence.
Behavioral finding
Behavieural examines human judgment and reliance within the identified context.
Bilateral comparison
Findings are compared, not merged. Agreement, disagreement and uncertainty remain attributable.
Explore the review, one dimension at a time.
These public-facing dimensions explain the service without revealing proprietary internal mechanisms.
01 / Purpose & scope
02 / Independent methods
03 / Review outputs
04 / Review sequence
- Initial enquiry and suitability assessment
- Agreed scope, confidentiality and terms
- Independent work within the evidence boundary
- Bilateral comparison and bounded reporting
05 / Limits & exclusions
Three distinct written outputs.
Each deliverable is attributable to its responsible practice or the defined comparison process. Format, scope and delivery terms are agreed before engagement.
Structural findings
Reconstructible account of decision formation, AI influence, evidence, authority and the limits of the record.
Behavioral findings
Independent assessment of defined questions about human judgment, interpretation, AI reliance and opportunity for challenge.
Bilateral note
Comparison of agreement, divergence, evidence asymmetry and unresolved matters without collapsing either independent conclusion.
Start with the decision, not the entire evidence archive.
Tell us which decision needs retrospective scrutiny and what question requires examination. Do not send confidential case material in an initial email.
Initial suitability
Identify the selected decision, the reason for scrutiny and the question to be answered, without confidential attachments.
Defined mandate
Agree evidence perimeter, privacy, confidentiality, responsibilities, timelines, commercial terms and exclusions.
Attributable conclusions
Receive distinct findings and a bilateral comparison with the evidentiary limits retained.
Professional responsibility, clearly attributed.
Independent practices with different analytical responsibilities. Professional biographies and affiliations belong to the respective individuals and organisations.
Cosmin-Corneliu Oprea
Founder · ID∆AC™
Independent researcher in AI-assisted decision systems and human agency. His work examines how authority, evidence, responsibility and the possibility of intervention change as AI shapes consequential decisions.
Research: ID∆AC Research · Guest essay at Erpenbach Institute.
Eugene Y. Chan, PhD
Founder & Principal Strategist · Behavieural
Behavioral scientist and academic whose research examines human judgment, credibility and decision-making. His scholarly and applied work connects behavioral science with organisational decision challenges.
PhD · University of Toronto; MA · University of Chicago; AB · University of Michigan. Institutional affiliations remain individually attributable.
A professional’s academic affiliation or a guest publication does not imply university endorsement, certification or independent validation of ADIR.
One decision worth examining?
Begin with a bounded scoping enquiry. We assess suitability and agree responsibilities, confidentiality and the evidence boundary before any substantive review.
Request an ADIR scope review ↗Partner service information: ADIR at Behavieural ↗