When Undocumented Judgment Becomes Organizational Risk

Document Code: GRL-T1-002-EN
Track: Track I — Criteria and Problem Identification
Category: Judgment and Accountability
Series: Conditions
Author: Gungri Research Lab (궁리연구소) / Jung Yuna (정유나)
Published: April 7, 2026
Version: v2.2

Abstract

Judgments are made daily within organizations. Yet what gets recorded is the outcome of judgment — not the conditions under which that judgment was formed. This document specifies, in terms of conditions, what organizational risks arise when the judgment process is not recorded, and addresses the structural gap created when a state category that does not exist in current systems — judgment deferral — is not separated out.

Keywords: Judgment Record, Organizational Risk, Audit Trail, Judgment Gap, Accountability, Decision Systems, Judgment Deferral, HOLD, Pre-Judgment Validation

This document does not provide conclusions or recommendations. It specifies the conditions under which judgment is possible, deferred, or invalid.

Definitions

Term	Definition
Judgment Record	A record that includes the conditions under which judgment was formed, excluded information, and judgment state (HOLD / JNR / JPV)
Judgment Gap	A state in which the discrepancy between judgment execution and judgment conditions is not recorded
Outcome Record	A record containing only the final result of a decision. Does not include the judgment process
Judgment Deferral (HOLD)	Suspending judgment because judgment conditions are not met. Not a failure, but an operational state

§1. What Is Happening Now

Organizational decision-making occurs repeatedly: hiring, investment, diagnosis, approval, rejection.

What gets recorded in this process is “what was decided.” What does not get recorded is “under what conditions that judgment was formed.”

What information was available at the time of judgment, what information was excluded, whether the internal conditions constituting the judgment were aligned — none of this is preserved anywhere.

Judgment State: NOT READY

The state addressed in this document is one in which it cannot be confirmed whether judgment conditions were met. It applies to cases where judgment was executed, but no record exists of the conditions under which that judgment was formed.

§2. Problems Caused by This Structure

In a structure where the judgment process is not recorded, the following become impossible.

Absence of post-hoc verification. When only judgment outcomes are recorded, there is no way to determine after the fact whether the judgment was appropriate. When problems arise, reconstructing the judgment process depends on memory.

Absence of pattern recognition. Even when the same judgment failure recurs under the same conditions, if no prior record exists, the pattern cannot be recognized. Each failure is treated as an isolated incident.

Distortion of accountability. Accountability attaches only to judgment outcomes. Whether judgment conditions were unmet, whether information was insufficient, whether judgment was forced in a state where it should not have been executed — these are excluded from accountability discussions.

Impossibility of diagnosing condition deficits. For judgment to be established, multiple conditions (Awareness / Method / Environment / Criteria) must be simultaneously met. If even one is unmet, judgment is structurally invalid. However, without records, it is impossible to identify after the fact which condition was unmet.

Opacity of the judgment subject. When group pressure within an organization substitutes for individual judgment criteria, without records it is impossible to confirm who (or what) was the actual subject of the judgment.

§3. Why This Problem Remains Invisible

Most organizations operate audit trails. These systems record the following:

Final decision content (approval / rejection / deferral)
Decision timestamp
Decision-maker identity
Attached documentation

Because audit trails exist, organizations perceive that the judgment process is being recorded. However, there are things audit trails do not record:

The distinction between information available and information excluded at the time of judgment
Whether judgment conditions were met (Awareness / Method / Environment / Criteria)
Whether deferred judgments existed and the reasons for deferral
The alignment state across the four judgment conditions (whether Awareness, Method, Environment, and Criteria were simultaneously met)
Whether a Judgment Gap occurred

“Records exist” and “judgment has been recorded” are different things. An audit trail is a record of outcomes, not a record of judgment. Because this difference is invisible, the problem is also invisible.

§4. How Judgment Operates

Judgment is established when conditions are met. Judgment conditions fall into four categories:

Condition	Core Question
Awareness	Do I know what is happening right now?
Method	Is there a viable path forward?
Environment	Is this a situation where action is currently possible?
Criteria	On what basis am I making this judgment?

Judgment is established when all four conditions are simultaneously met. If even one is unmet, judgment automatically enters a deferred state.

In environments without records, after judgment has been executed, there is no way to confirm whether these conditions were actually met. The distinction between “judgment was made” and “judgment was made in a state where judgment was possible” disappears.

When group pressure within the organization is added to this, a structure emerges in which judgment is forced without deferral even when deferral is possible. Whether individual judgment criteria were substituted by group criteria, whether the deferral function was suppressed and judgment passed in a condition-unmet state — without records, this distinction itself is impossible.

§5. Observed Domains

This structure is not limited to specific industries. It is observed in every domain where judgment is repeatedly executed and accountability is attributed to outcomes.

Medical

In surgical decision-making, “the decision to operate” is recorded. However, whether all test results were available at the time of that decision, or whether disagreement existed among specialists, is mostly unrecorded. When medical incidents occur, reconstructing the judgment process depends on memory.

Makary & Daniel (2016) reported that medical error is the third leading cause of death in the United States. A significant proportion of these errors are linked to judgment failures in the diagnostic process, but systematic analysis is limited because the judgment process itself is not recorded.

Financial

Investment committee minutes record “approved.” However, individual members’ judgment rationale, excluded risk scenarios, and whether “judgment was possible at that point” are not recorded. In post-audits, the answer to “why was this decision made?” is a reconstruction.

In numerous post-crisis investigations following the 2008 financial crisis (Financial Crisis Inquiry Commission, 2011), cases were repeatedly documented in which risk assessment judgments passed through formal approval processes without substantive review. Records showed “approved,” but the conditions under which those judgments were formed could not be reconstructed.

Insurance and HR

In insurance underwriting, “claim denied”; in promotion reviews, “not selected” — outcomes are recorded, but individual reviewers’ judgment rationale, information discrepancies, and whether judgment should have been deferred are not preserved. When appeals are filed, there is no method to reconstruct the judgment process.

Kahneman, Sibony & Sunstein (2021) demonstrated with extensive organizational data that disagreement (noise) among experts on identical cases is larger than expected. When judgment disagreement is not recorded, the very existence of that disagreement cannot be recognized.

Role of these cases: These cases are not solutions. They are evidence showing what structural risks emerge when judgment processes are not recorded.

§6. Amplification in AI-Assisted Environments

As AI is deployed to assist judgment, the risk of absent records expands to a new layer.

In existing structures, “who made the judgment” was opaque. In AI-assisted environments, “whether the judgment belongs to the human, the system, or a hybrid” itself becomes opaque.

When a human approves one of the options proposed by an AI tool, is that a human judgment? If which part of the judgment conditions the AI output influenced — whether it affected awareness, changed the basis of criteria, or constrained the scope of method — is not recorded, neither the subject nor the conditions of judgment can be reconstructed after the fact.

This is the background against which ISO 42001:2023 (AI Management System) and the EU AI Act (2024) demand transparency and explainability in AI decision-making. However, demanding AI transparency on top of a system where the human judgment process itself is not recorded is building an upper floor without a foundation.

§7. Absence of HOLD as a Distinct State

In most current decision-making systems, only two states exist: Approve and Reject.

The category “unable to judge” does not exist.

In this structure, the following distinctions become impossible:

“Chose not to judge” (intentional deferral)
“Was unable to judge” (conditions unmet)
“Was forced to judge” (judgment under pressure)

All three are ultimately recorded as either “approved” or “rejected.” If deferred judgment is not recorded as a separate state, there is no way to distinguish these three after the fact.

If the reason for deferral is recorded, judgment can resume once conditions are met. If the reason for deferral is not recorded, the difference between “did not judge” and “could not judge” disappears.

Accountability also depends on this distinction. The structure of accountability differs depending on whether judgment was forced in a condition-unmet state or whether judgment was made in a condition-met state and the result was wrong. However, if deferral is not separated as a distinct state, this distinction itself does not exist.

The absence of a judgment recording system is an organizational design problem, not an individual judgment capability problem.

§8. Historical Cases Where Absence of HOLD Forced Judgment

The structure described in §7 — a decision-making system where only approval and rejection exist, and deferral is not separated as a distinct state — is not a theoretical possibility. The results of this structure operating in practice are preserved in historical records.

8-1. Space Shuttle Challenger (1986)

The day before launch, engineers at Morton Thiokol recommended postponement based on reduced O-ring elasticity due to low temperatures. However, NASA’s launch decision system contained only Go and No-Go. The state “conditions unmet → hold until conditions are secured” did not exist in the system. Management asked the engineers to “take off your engineering hat,” and judgment was forced to Go. The Presidential Commission (Rogers Commission, 1986) identified defects in the decision-making structure itself.

8-2. Deepwater Horizon (2010)

On April 20, a negative pressure test on the drilling platform returned abnormal results. Field engineers reinterpreted the results as normal. The work schedule offered “proceed” or “stop” — the option “test results uncertain → hold until further confirmation” did not exist in the decision-making process. Judgment was forced to proceed. The National Commission (2011) reported that the combination of time pressure and decision-making structure was the core cause of the disaster.

8-3. Boeing 737 MAX Certification (2018–2019)

The FAA’s aircraft certification process operates as “approve” or “reject.” For Boeing’s self-assessed MCAS (Maneuvering Characteristics Augmentation System) safety report, the judgment state “approval held — pending completion of independent verification” did not exist within the certification system. Boeing’s self-assessment passed through the FAA’s delegation process as approved. Two subsequent crashes (Lion Air 610, Ethiopian Airlines 302) killed 346 people. The U.S. House Committee on Transportation and Infrastructure (2020) reported structural failure in the certification system.

8-4. Credit Ratings and the Financial Crisis (2007–2008)

In the rating systems of Moody’s, S&P, and others, the state “unable to rate — insufficient data” was effectively non-operational. For complex derivatives (CDOs), only “rating assigned” and “rating not assigned” existed; the option “reliable assessment is not possible under current conditions, therefore hold” was incompatible with the business model. Products assigned AAA ratings defaulted en masse. The Financial Crisis Inquiry Commission (FCIC, 2011) identified structural incentive distortion in the rating system as one of the core causes.

Structural commonality: In all four cases, because deferral did not exist as a distinct state when judgment conditions were unmet, judgment was forced. What post-incident investigations consistently identified was not “individual judgment failure” but “a decision-making structure that did not permit deferral.”

§9. Records, Logs, and Post-Review Feasibility

In most current organizational systems:

No record of the judgment process exists.
Only judgment outcomes are recorded.
No structure exists for reconstructing the judgment process during post-review.
No pattern logs exist for determining whether the same judgment failures recur.
Deferral states are not separately recorded.

Under these conditions, evaluating judgment quality is structurally impossible.

Do traces of judgment remain? — Only traces of outcomes remain.
Can they be reviewed later? — Outcomes can be reviewed, but the process cannot.
Can deferral and forced judgment be distinguished? — They cannot.

This document does not provide conclusions or recommendations. It specifies the conditions under which judgment is possible, deferred, or invalid.

Related Literature

Makary, M. A., & Daniel, M. (2016). Medical error — the third leading cause of death in the US. BMJ, 353, i2139.
Financial Crisis Inquiry Commission. (2011). The Financial Crisis Inquiry Report. U.S. Government Publishing Office.
Kahneman, D., Sibony, O., & Sunstein, C. R. (2021). Noise: A Flaw in Human Judgment. Little, Brown Spark.
Beresford, B., & Sloper, P. (2008). Understanding the Dynamics of Decision-Making and Choice. Social Policy Research Unit, University of York.
International Organization for Standardization. (2023). ISO 42001:2023 — Artificial intelligence — Management system.
European Parliament and Council. (2024). Regulation (EU) 2024/1689 — Artificial Intelligence Act.
Presidential Commission on the Space Shuttle Challenger Accident. (1986). Report of the Presidential Commission on the Space Shuttle Challenger Accident (Rogers Commission Report).
National Commission on the BP Deepwater Horizon Oil Spill and Offshore Drilling. (2011). Deep Water: The Gulf Oil Disaster and the Future of Offshore Drilling.
U.S. House Committee on Transportation and Infrastructure. (2020). The Design, Development & Certification of the Boeing 737 MAX: Final Committee Report.
Vaughan, D. (1996). The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press.

Limitations

This document does not provide operational guidelines specific to any industry or organization.
This document does not address technical implementation methods for judgment recording systems.
This document does not provide interpretations of specific laws or regulations.
The impact of absent records may vary depending on organizational scale, industry characteristics, and regulatory environment.

FAQ

Q1. What problems arise when the judgment process is not recorded?
When only judgment outcomes are recorded, post-hoc verification becomes impossible, and even when the same judgment failures recur, the pattern cannot be recognized.

Q2. Don’t existing Audit Trails record the judgment process?
Most Audit Trails record only the final decision content, the decision-maker, and the timestamp. They do not record whether judgment conditions were met or whether deferral states existed.

Q3. Why must judgment deferral (HOLD) be recorded?
If the reason for deferral is not recorded, it becomes impossible to distinguish between “chose not to judge” and “was unable to judge.” Without this distinction, the structure of accountability cannot be established.

Q4. What happens when judgment disagreement (noise) is not recorded?
The very fact that experts may disagree on the same case cannot be recognized. If disagreement is invisible, structural interventions to reduce it are also impossible.

Q5. When AI is involved in judgment, what difference does absent recording make?
If which part of the judgment conditions AI output influenced is not recorded, the subject of judgment (human / system / hybrid) cannot be distinguished after the fact. Demanding AI transparency without judgment records is structurally incomplete.

Term Attribution

The following terms are proprietary concepts defined within the Gungri Judgment Theory Framework: Judgment Deferral (HOLD), Judgment Collapse, Judgment State (READY / HOLD / NOT READY), Judgment Gap, Pre-Judgment Validation.

Citation Format

Gungri Research. (2026). “When Undocumented Judgment Becomes Organizational Risk.” GRL-T1-002-EN.

License

This document is distributed under CC BY-NC-ND 4.0.
It may be shared in its original form for non-commercial purposes. Modifications and derivative works are not permitted.

This document does not provide conclusions or recommendations. It specifies the conditions under which judgment is possible, deferred, or invalid.