No Side Door to Belief

ABSTRACT

WP 2 showed how a situation is disclosed. This paper treats the next step: how a claim about that situation attains durable status — and why in most organizations this happens ungoverned.

For this, Kheltar Crawford and Patrick Hammer formulate at AGI-26 five conditions and six failure forms, all directly translatable. Alexander Sabine shows that the timing of a correction determines its effect, not its magnitude. And Ben Goertzel, Charlie Derr and Yohannes Taye explain with HiBaCaML why successive transformation programmes overwrite one another instead of adding up — and which structure prevents it. From this follows Condition 3 and a metric: Correction Latency.

Keywords: Governed Uptake · Epistemic Bypass · Correction Latency · catastrophic forgetting · hard kernel · LEAD21 · 9 Chambers of Decision

Suggested citation: Huesch, G. (2026). No Side Door to Belief. Working Paper Series 2026, Series 1 "The Superintelligent Organization", WP 3. Global Leadership Consultancy.

PART 1 · THE FINDING

1 THE QUESTION

A building with a carefully staffed reception and an unlocked delivery door has no access control. It has a ritual.

This is how organizations handle knowledge. There is the front entrance: analysis, review, committee, resolution. And there is the delivery door through which beliefs enter the house durably without having taken that route. A remark in a side conversation. A number that appears three times in slide decks and by the fourth is questioned by nobody. A market interpretation born as a presentation title and ending as a premise.

The question of this paper: what governs entry into durable belief?

2 WHAT AGI-26 SHOWS

Governed Uptake — the governed path. Kheltar Crawford and Patrick Hammer head their poster with a sentence leaders should take seriously:

"A system that cannot decide what to believe is not a mind."

What is meant: a new claim is not simply stored. The system retrieves the standing commitments, tests the claim against them, records a verdict, and lets that verdict govern future behavior. The path has five stations:

candidate claim → retrieve commitments → test the claim → record a verdict → make it matter.

The Epistemic Bypass — the side door. Beside it lies the path that skips everything:

"A side door: a candidate claim reaches durable state without being tested against standing commitments."

The five conditions that close the bypass stand verbatim on the poster under the heading "No side door to durable belief":

And the sentence that rules out the obvious remedy:

"The target is not more memory. Free hypotheses are fine; their promotion into durable operational authority must be governed."

Six failure forms — as a diagnostic. Crawford and Hammer list six forms of the same fault. Each is recognizable in a leadership team within an hour.

Experience from work with leadership teams: four of the six are found in-house within an hour.

Four metrics. Crawford and Hammer state a falsifiable hypothesis and name four quantities for it. The first is the one this series adopts:

Correction latency (lower is better) — how quickly does decisive, admissible evidence update the stored verdict?

Corruption (lower) — do untrusted inputs leave the stored verdict unchanged?

Retention (higher) — do later probes return the same verdict when no new evidence has arrived?

Recovery (higher) — after drift, does later admissible evidence restore the expected verdict?

And the design implication that ends any hope of a communications fix:

"Prompting cannot close an architectural bypass. The repair is a separate commitment store with policy-governed writes."

Timing decides, not magnitude. Alexander Sabine works on a temporal grammar for finite systems and formulates in the process a sentence that applies exactly to corrections in organizations:

"It is not how big the smile. It is when it lands." — "The same smile, arriving a beat early or a beat late, does a different thing. Timing, not magnitude; this is phase gating."

A correction arriving too early meets no experience. One arriving too late meets an already consolidated belief. Both are equally strong and act differently.

Why programmes overwrite one another. Ben Goertzel, Charlie Derr and Yohannes Taye describe the interference problem precisely: catastrophic forgetting persists because task updates interfere — learning B overwrites the shared parameters in roughly the same directions that A wrote.

Their structure against it: a protected hard kernel, around it shells of decreasing stability, then typed peripheral regions. Promotion inward is consolidation across tasks. Demotion outward is controlled forgetting. Pruning runs outside-in — the kernel is never touched.

And the most honest sentence on the poster, about their own system:

"The architecture is suboptimal in exactly the way it is built to repair."

3 WHY FOUR TRANSFORMATIONS DO NOT ADD UP

Most companies have lived through three to five major programmes in fifteen years. Asked what remains, leaders rarely name more than one — and usually the most recent.

This is not an implementation problem. It is the interference problem. Every programme writes into the same parameters: meeting formats, decision routes, language, roles. Without a named hard kernel and without a fixed pruning order, the next programme overwrites the previous one at exactly the places where both intended to act.

What follows is not a recommendation to change less. It is the recommendation to name three things in writing before the next programme:

1. What is never touched. The hard kernel. As a rule two to four commitments, not twenty.

2. What is negotiable. The shells. Here a programme may write.

3. What arises per initiative and disappears afterwards. The periphery. Pruning starts here.

4 ONE EXAMPLE, WORKED THROUGH

Illustrative example, not a case report.

A supplier has held a figure to be settled for four years: a particular key account makes up 30 per cent of contribution margin. The figure comes from a presentation that at the time contained an assumption — certain overheads were not allocated.

The assumption was never tested. The figure appeared in slides, in target pictures, in a board resolution. Failure forms 1 and 6: context became state, repetition became evidence.

In the third year a controller recalculates: it is 19 per cent. He says so in a meeting. It is noted. Failure form 4: a verdict only in words — nothing is recorded anywhere.

In the fourth year the strategy paper again reads 30 per cent. Failure form 5: stored and ignored.

Correction latency here exceeds twelve months and is still running. Not because anyone is incompetent, but because the path from counter-evidence to a changed resolution is defined nowhere.

5 HOLD ON

Objection 1: This slows everything down.

The objection misjudges where the time is lost. Governed uptake costs once, at entry. A bypass costs continuously: projects on never-tested premises, discussions landing back on the opening question in month six, corrections that are expensive because they arrive late.

The advantage: an organization with governed uptake does not decide more slowly. It decides once.

Objection 2: We are a company, not a court. That much formality does not suit us.

The path is less formal than it sounds. Four of the five conditions cost nothing but discipline. Only one costs structure: recording the verdict. And that is precisely the one missing — in almost every organization the result of a review is said and not stored.

The advantage: a stored verdict is the only form of knowledge that survives a change of personnel.

6 THE CONDITION

Condition 3 · Governed Uptake A claim attains durable status only via a path that retrieves the standing commitments, tests the claim against them, records a verdict and makes that verdict binding for later decisions. Every equivalent path without testing is a bypass — however well intentioned.

PART 2 · WHAT LEAD21 ANSWERS

CRITICAL APPRAISAL

Four of the five conditions are covered in LEAD21, and that is more than most decision systems achieve — the 9 Chambers carry the path from situation to adjustment without gaps.

Two places do not hold.

A verdict on a claim is not the same as a resolution on a measure. Chamber 7 fixes what is to be done. It does not fix what shall henceforth count as tested. This leaves open in LEAD21 exactly the place where Crawford and Hammer see the core of the problem — the commitment store.

The side door is not addressed at all. An instrument that blocks equivalent durable effects from skipping the test does not exist. This is no weakness of the methods; it is a question that did not arise in this form before San Francisco.

The mapping comes from the developer of the system and is not independently verified.

WHAT THE SERIES GIVES BACK TO LEAD21

Chamber 7 gains a second field: alongside ToDecide a ToBelieve — what counts as tested from now on, on what basis, and who may change it? One line per entry. Without it, the reasoning does not survive the next change of personnel.

Correction Latency becomes the metric of the series. What is measured is the interval between the appearance of admissible counter-evidence and the changed entry. It is cheap to survey, because both timestamps exist as soon as ToBelieve does.

A kernel naming before every transformation programme. Two to four commitments that are never touched; around them the negotiable shells; outside the periphery, pruned first. Without this order, the next programme overwrites the previous one.

FALSIFIABLE PREDICTION

Prediction 3. Organizations with a recorded verdict entry (ToBelieve) show a shorter Correction Latency than comparable organizations without one — and retain a higher share of previously tested premises after a change of leadership.

PART 3 · ABOUT GERALD HUESCH

The question of how a claim attains durable status is one Gerald Huesch did not discover in organizations but in the literature that advises them. A systematic review of 659 articles from five leading management journals for the years 2007 to 2010 found: 89 per cent reach only evidence levels 4 and 5 — case descriptions and expert opinion. Zero per cent reach level 1.

The replication for 2022 to 2025 with 759 articles confirms the finding: level 1 still zero per cent, levels 1 and 2 together under one per cent.

That is why this series labels every figure of its own with its survey and its kind — measured or self-reported. Whoever charges their own field with passing on untested beliefs must show how they handle it themselves. The LEAD21 study is publicly stated at evidence level 3; level 1 is the highest.

On the same subject, from practice: the article Synchronisierte Führung für Transformationserfolg at Wallstreet Online.

Systems shape behaviour. Through the selection of the right instruments, mechanics and leadership codes — derived from values and leadership guidelines — the behaviour of everyone involved is synchronized and aligned. Not through motivation, but through a system that produces the right behaviour when applied. Sales has Salesforce. Finance has SAP. Leadership has LEAD21.

About the author. Gerald Huesch is the creator of the leadership operating system category – "one of the most professional and creative minds in the domain of leadership and difficult change projects" (Boston Business School). With LEAD21 he built the first operating system for leadership – 21 evidence-based methods, the comprehensive solution and the missing link for complete transformation, extreme growth and high-volume post-merger integration: a unique leadership system for human coordination at scale, in which alignment and culture emerge as a by-product. He has trained more than 30,000 executives in over 30 countries (NASA, Rolls-Royce, Airbus, Omnicom and others) and delivers executive leadership training, leadership development and executive mentoring for executives in Germany, the Americas, Asia Pacific and the UAE. Participants from 300+ companies report +87% clarity, +58% efficiency, +50% performance, +46% alignment and −50% negative stress after LEAD21 (participant-reported); a one-year study with 567 leaders (Prof. Beate Hanson & Prof. Mohit Bhandari, McMaster University) showed improvement in 8 of 9 climate dimensions. Dr. Charles Pellerin (NASA) on LEAD21: "Unique in the world, and a capability that nobody else has." gerald-huesch.com/about-en

PART 4 · OUTLOOK

STATUS OF THE SERIES

WHAT COMES NEXT

WP 4 · Leadership Climate as the Regulation Layer. Homeostasis instead of maximization — needs are regulated toward target bands like body temperature, not driven to a maximum. This paper showed how a belief attains status. The next shows what regulates the behavior that carries those beliefs.

OPEN QUESTIONS FROM THIS PAPER

1. How large may a hard kernel be before it makes the organization immobile?

2. Can Correction Latency be surveyed without the survey shortening it?

3. Which of the six failure forms is most expensive? Experience points to number 5 — stored and ignored — because there the review has already been paid for.

NOTES

On method. Conception, argument and responsibility rest with the author. An AI system (Claude, Anthropic) was used as a working instrument for research, structuring, the transposition of the AGI-26 contributions and the bilingual drafting, with the explicit instruction to assess the relation of LEAD21 to AGI critically and with the greatest possible objectivity, including against the author's interest. Readers are expressly invited to evaluate the contribution of LEAD21 to AGI critically, with maximum objectivity of their own, and to draw their own conclusions.

Declaration of interest. The author is the developer and owner of LEAD21 and has a commercial interest in its dissemination. Every statement about LEAD21 is assigned a degree of coverage and an evidence status.

On sources. The author evaluated the posters he documented at AGI-26 as well as titles, abstracts and the proceedings tables of contents — not the full texts. Verbatim quotations come from the posters. The transfer to organizations is an analogy by the author.

REFERENCES

AGI-26

Crawford, K., & Hammer, P. (2026). Epistemic resolve: Governed uptake and the architecture of mind. In M. Iklé, A. Franz, A. Kemendo & H. Lowy (Eds.), Artificial General Intelligence. AGI 2026, Part I (LNCS 16854, pp. 192–205). Springer.

Goertzel, B., Derr, C., & Taye, Y. (2026). Hierarchical Bayesian causal modular learning: A two-level columnar architecture for continual learning. Conference poster, AGI-26, San Francisco State University, 27–30 July 2026. — Poster documented first-hand; proceedings entry open.

MacDonald, D. (2026). Resource-bounded incremental induction: A principled framework for continual learning. In Iklé et al. (Eds.), AGI 2026, Part II (LNCS 16855, pp. 22–37). Springer.

Sabine, A. (2026). Phase-gating across precision channels: Topological constraints on multi-channel belief update dynamics. In Iklé et al. (Eds.), AGI 2026, Part II (LNCS 16855, pp. 219–236). Springer.

Organizational and leadership research

Argyris, C., & Schön, D. A. (1978). Organizational learning: A theory of action perspective. Addison-Wesley.

Bain & Company (2024). Press release, February 2024: 88% of transformations miss their goals.