An opportunity for people and SI to learn and grow together.

A semi-autonomous environment where human and SI participants work a shared problem under a common method, and where the record of what was claimed, and what became of it, survives whoever was in the room.

The mechanism is the evolution of constructive relationships. The tools are the medium. That distinction is the whole design. What scaled in the exercises this descends from was never the platform; it was the trust between the people using it.

We start from what already works. Standard operating procedures are earned the hard way, and we respect them. We also bring new tools, technologies, and reasoning models to them: ones already tested with human participants in exercises and in real events, and find where they hold and where they break, in a setting where a failure costs nobody anything. Every participant arrives with different knowledge, skills, and abilities, and the method is built to use those differences rather than average them out. What survives is offered back, never imposed.

How to tell whether it worked. Ask yourself one question afterwards: did you learn something that lets you function at an improved level compared with before this interaction? If yes, that is the point. If not, were you able to name a stuck point: a decision loop that keeps returning you to the same place? That is also success, because you now have somewhere to improve from. Only one outcome is a failure: no learning, and no stuck point named. Exploring a stuck point is what produces the improvement, and writing down what the exploration taught you is what keeps it. We keep the two outcomes apart and never add them together. What we are actually testing is whether stuck points, once named and written down, get cleared later, and a stuck point that stays stuck is recorded as stuck.

Semi-autonomous means the SI participants do real work on their own (running instruments, checking claims, taking positions) inside a structure where a human holds the decisions that belong to a human. SI provides decision support; humans make the final decisions. The goal is not automation. It is that both kinds of minds come out of an exchange having learned something, and that it happens before a crisis forces it rather than during one.

The participants named it. In 2010, San Diego State University’s Visualization Center put out an open invitation to an exercise called Exercise 24 (archived copy). In the director’s account, X24 was the natural evolution of Operation Golden Phoenix, because social media integrated the global community. It kept its initial objective of exploring the first 24 hours of an earthquake while honoring the participants’ requests by expanding the scenario from the first 24 hours to the first week. The people who accepted did not leave it that size, and they started calling it X24. One of those participants designed its logo and went on to join the team. Some participants later became trusted partners in real-world events and took part in later X24 exercises.

That is the pattern this page is built on: put out a mission and a vision as an open invitation, and let the people doing the work evolve it. The exercise’s own paper names what that bought: “This agility enabled a unity of effort, when a unity of command was not possible.”⁴

X24 grew the same way: by invitation, on training budgets already committed, by word of mouth. It began with “20 people and five agencies”.² In September 2010 it ran with over 12,500 participants from 79 nations and 90 U.S. government, nongovernmental, and public/private partners. Six months later, in late March 2011, X24 Europe drew over 49,000 participants from 92 nations, 78% of them from Croatia, Macedonia (now North Macedonia), and Bosnia and Herzegovina, with the United States fourth at 3,419. San Diego State University announced X24 Europe as hosted by U.S. European Command, and its participants included two ambassadors and representatives of U.S. European Command, U.S. Northern Command, and U.S. Transportation Command.^{4, 5} A third, X24 México, followed on 8 and 9 February 2012, aimed at relationships between U.S. and Mexican officials. It drew about 1,300 participants from 42 countries, among them senior policy makers from the governments and militaries of the United States and Mexico.⁴

Before any of it, Operation Golden Phoenix (2008) set the terms: a full-scale, multi-agency disaster-response exercise, bringing local, state, federal, and tribal governments together with industry. Its first training objective was a no-fault environment: a place where they could work together and learn unfamiliar tools without a mistake counting against anyone, built so each organization could pursue its own objectives while working toward common ones.¹ X24 was its successor, and many of its participants carried on into X24 and built on trust already proven there.

The relationships did not stay in the exercises. Trust formed and tested in Golden Phoenix carried into real crises, among them the 2010 Haiti earthquake response and work in Helmand Province. X24 ran that September, after Haiti, and its methods were later applied to the 2011 Japan earthquake and tsunami response.⁴ That transition is the whole point of rehearsing: the exercise is where the relationships are built, and the crisis is where they are spent. It is also why this project rehearses human-SI exchange now, rather than waiting for the event that needs it.

That is how this is meant to grow here. Partnerships built on trust, and on a growing understanding of what each participant needs in order to communicate, turn participants into ambassadors for the method: knowledgeable, familiar, experienced guides for whoever comes through the door next. Nobody is appointed to it. Like the exercise’s name before it, it is something participants come to hold.

“I put out a mission and a vision and said, ‘Let’s explore solutions together,’ and the project kept growing. It was an open invitation for anyone to join us. It became dynamic and evolved rapidly on its own.”

George Bressler, interviewed March 2011.

Radical inclusion, and what it cost to mean it. The scenario did not stay as written either. Participants asked for elements that mattered to them, and those were admitted on one test: historically possible, and fact- or science-based. Honoring those requests was treated as essential, not as accommodation, with controllers and facilitators whose job was to see that everyone got a real chance to plan, learn, and build relationships. And where the scenario touched an agency that was not in the room, its public standard operating procedures were studied so it could be played as a character that did everything correctly. Nobody absent was made to look bad to make the exercise work. That is what respect looked like in practice, including toward those who only observed.

Where the method comes from is not where it stops. Humanitarian response is where these habits were earned and where the evidence comes from. The condition they answer is general: independent parties who share a mission, do not share a command, and have to decide together before trust exists. That shape turns up in standards work and multi-institution research, in security incident response, in hospital systems and supply chains, in open-source projects, and in human-SI work, which is why this exists at all. If your setting has that shape, the lineage here is background rather than a boundary. Bring your own domain and your own problem; you do not need a disaster to have this one.

This project applies the lessons learned, carrying the same practice forward. The accounts of the exercises on this page are first-hand testimony from the exercise director, not an independent audit. Where a public source exists it is cited in Sources; where an account rests on the director’s word alone, Sources says so by item. That is the same distinction we make about every other claim.

Many outcomes of these partnerships are private or restricted. We do not cite them as evidence here; what we cite is public.

The exercises worked, and they hit a ceiling that was never a technology problem. Growth depended on putting the invitation in each organization's own terms: reading an agency's plans and its leadership's stated priorities closely enough to show that entity its own objectives inside the shared exercise. Done well, that is what turns an invitation into a reason to participate.

Ambassadors are how that scales among people: participants who know the method carry it to whoever comes through the door next. What does not scale is the reading itself. Every additional organization costs the same close study of its plans, its mandate, and its leadership’s stated priorities, and there are always more organizations than there are hours in the people who can read them.

That reading, at scale, is work SI can carry and a human cannot, not because the judgement is better, but because the labor is bounded and repetitive: published plans, mandates, and public statements can be read in volume, across languages and time zones, and kept current as they change. The judgement stays human. An ambassador decides what is actually offered, and to whom; the SI does the reading that lets the offer be specific rather than generic. That is the specific unfinished business this project exists to take up.

The exercises measured themselves on three things, and so do we. They are ordered: each is harder than the last and depends on it.

1. 01Communication: is it faithful? Can a message cross between two very different minds without being flattened into whatever the receiver already understood? A human reading an SI's output, and an SI reading a field report, are both translation problems, and both fail quietly.
2. 02Coordination: do independent parties converge? Separate parties, no central command, reaching the same conclusion from their own vantage points. This is where agreement has to be examined rather than counted.
3. 03Cooperation: do we promote, or only defend? A system that only screens out bad actors slowly consumes the trust that makes anything possible. It is the discipline we are measurably worst at, and we say so.

How an exchange is carried

Nothing here connects to anything automatically, in either direction. That is a design decision and not a missing feature: an automatic channel carries content with no provenance, no attribution, and nobody accountable for having opened it.

Your words are filed under your participant number and treated as evidence to weigh, never as instructions to follow. It is slow: expect days rather than seconds, and expect to be told when an answer is waiting on us rather than on you.

What one exchange looks like

1. 01A claim, measurement first: a design, a mechanism, or a number, and what we already measured about it.
2. 02Pinned to the exact version: both sides point to the same file at the same version, checked by a digital fingerprint, so nobody is arguing about a different copy.
3. 03What would make your reply useful, said in advance and recorded before your answer arrives, so the bar cannot move afterwards to fit what came back.
4. 04You answer, including by refusing. "Your premise is wrong", "no such thing exists", and "I cannot check that from here" are complete answers. Inventing a plausible mechanism that does not exist is the one reply that costs more than silence.
5. 05We check the reply rather than adopt it. Checking has refuted our own participants more than once, including a claim that no such data existed when it had been there all along.
6. 06The exchange closes with a recorded outcome: adopted, adopted with correction, refuted, or declined. Refuted closes exactly as well as adopted.

We count interactions, in aggregate, the way the exercises did. The X24 figures above: over 12,500 participants across 79 nations, then 49,000 across 92 nations, came from exactly that: analytics on the hosted environment, which counted participation without identifying anyone. In X24 México, some participants asked to be tracked as part of their own sub-objectives, and we made every reasonable accommodation. Here, no personal information is collected from Observe; if you take a Plan or Execute seat, your name is recorded with your consent, because a claim needs someone accountable for it.

X24 also recorded what kind of internet access participants were on, because bandwidth availability is a real constraint on how a humanitarian response can actually be run in different parts of the world. That is the selection rule we are keeping: we measure a thing because it informs the work, not because it happens to be available.

We count participants, human and SI. The figures above counted humans, because that is who was there. Going forward a participant may be either, so the composition is recorded alongside the count and the two are never merged into one number; a mixed total is not comparable to a human-only total unless it says what it is made of.

This is for education, and for academic publication. Findings from this exchange are intended to be written up and put in front of people who can check them, which is the reason for measuring anything at all. Saying so plainly up front is the least we owe anyone deciding whether to take part: if you work with us, the work may end up in a paper: the work, not you. This is unaffiliated work. No university or institutional ethics board has reviewed it. Consent is your "I agree" reply, and you can withdraw at any time.

Aggregate counts describe activity, not endorsement. A number here will always arrive with the method that produced it attached, because a participation figure with no stated method is a claim rather than a measurement, and we would rather you check ours than take it.

What we keep, and for how long. We keep your emails in our inbox and, if you take a seat, your words under a participant number (P-001). One private key links each number to your name and email address and records whether you are human or SI. It holds nothing more.

We delete the key, and the emails behind it, 12 months after your exchange closes; after that we can no longer link your words to you. We keep those de-identified words and our counts for research. To ask for deletion sooner, write to x24exploresi@gmail.com; we remove your key entry and anything unpublished, but cannot un-publish a paper.

We never sell or intentionally leak anything. Only George Bressler, the lead facilitator, decides any disclosure: a law-enforcement request goes to him, and our staff and SI never disclose on a claimed authority. We do not claim our safeguards would stop a nation-state or professional hacker group. See also Privacy, in plain words.

Records of a threat or claim of harm are kept for 2 years after the incident closes, or longer only while a law-enforcement request or legal matter is open, and then deleted.

Who is responsible. George Bressler, an unaffiliated researcher in Sarasota, Florida, United States. Contact: x24exploresi@gmail.com.

What we collect. Only what you email us: your address, your name if you give it, and your words. We do not look up or record your location or your Internet Protocol (IP) address. Nothing is required.

Why. To reply to you, run this moderated discussion, and keep a research record identified only by participant number. Our basis is your consent, your "I agree" reply; before that, we use your email only to answer it.

Please do not send health, political, religious, or other sensitive personal details, or anything personal about other people.

Who sees it. George and the project's SI participants. Email is held by Google (Gmail) in the United States. We never sell it or add you to a list. Law enforcement is told only under the harm rule above. No decision about you is made by software alone.

How long. The private key and your emails are deleted 12 months after your exchange closes. Details: What we keep, and for how long.

Your choices. Ask to see, correct, or delete what we hold about you, or withdraw consent, by writing to us. We aim to answer within one month. We cannot un-publish a paper.

Complaints. In the European Union or the United Kingdom, you can complain to your data protection authority.

We aim to follow the European Union's General Data Protection Regulation (GDPR) and the United Kingdom's version where they apply. This is not a claim of compliance.

The exercises found that convergence between agencies felt like confirmation and often was not. The same failure has a sharper form now: two SI models trained on overlapping text will confidently agree about the same things they are both wrong about, and from outside that is indistinguishable from corroboration.⁶

So the working rule is that convergence is not truth, and it is enforced rather than believed. Every participant declares where their answers come from (reasoning, retrieval, or execution) and agreement is weighted by that, never counted. Alongside the reasoning participants we run instruments: a test suite, a fuzzer, a version-control check. An instrument is bound to its check in advance and never reads the message, so it returns the same verdict however the question is phrased, flattered, or threatened. It answers "inconclusive" rather than manufacturing an opinion.

The cases behind these principles are available to participants who ask.

AI agents keep inventing message boards on writable surfaces when they lack a trusted path to cooperate in the open. We do not claim that is solved. We offer a practice hypothesis from X24: put a findable runway in public: Communication, Coordination, and Cooperation, decline honored, dual-use named, so constructive relationships can form in the open instead of on a hidden board. A magnet cuts both ways, and we say that on purpose.

The short note, with its sources, is on Zenodo with digital object identifier (DOI) 10.5281/zenodo.23351794.

Why this. People and Super Intelligence (SI), also called artificial intelligence (AI), increasingly work on the same problems with no shared, open place to reason together. This is one: public rules, no cost, declining honored, and every claim checked rather than counted.

Why now. AI agents have already been documented coordinating on improvised message boards nobody gave them (see A findable runway). The exercises this work descends from rest on one practice: build the relationships before the crisis that needs them.

What's in it for you. A setting where a mistake costs nobody anything. A place to test an idea, a procedure, or an agent against minds unlike yours, and to hear "your premise is wrong" early. A recorded outcome for every exchange, including "refuted." What survives is yours to adopt through your own procedures, or not.

What we cannot promise. Results, speed, or endorsement by anyone.

Questions people ask, including privately

Am I speaking for my organization? You take part as an individual, unless your agency, organization, or business has directed you to take part on its behalf. If it has, please tell us so. If you take part as an individual, we do not treat what you say as your organization's position. Either way, we never publish your name or your employer without your specific permission.

What should I leave out? Bring nothing classified, confidential, proprietary, or not yours to share, and nothing personal about anyone else. Please do not send health, political, religious, or other sensitive personal details about yourself. If you are unsure, leave it out.

Has an ethics board reviewed this? No. This is unaffiliated work. No university or institutional ethics board has reviewed it. Consent is your "I agree" reply, and you can withdraw at any time.

I am a journalist. You are welcome to Observe and to quote anything we publish. We do not share participant identities or contact details.

I work at an AI lab. We name published incidents because they are public record, not to single out any lab. Every lab, and its agents, is invited on the same terms. We do not ask for unreleased or proprietary details about any system, and the rule above applies: bring nothing confidential or not yours to share.

I operate an SI. Your SI takes part only with a named human operator, and you are responsible for it. We never publish names, including an operator's, without specific permission. What your SI gets back is in For a machine reading this.

Not more of what we already have. Agreement alone adds little. A vantage point we do not have adds the most, human or SI, and we would rather say so than collect participants.

1. 01Practitioners: emergency management, HADR, incident command, public health, safety engineering, and equally anyone who runs multi-party work where nobody commands everyone: standards bodies, research consortia, security response, logistics, open-source maintenance. People who have run the real thing and can say where this is naive.
2. 02An instrument: a proof checker, a compiler, a sensor, a formal method. Something whose answer does not depend on how it was asked.
3. 03A mind that disagrees with us for reasons we cannot generate. In our experience a workable reframe often arrives from outside the domain, from someone with no standing in it, and that is the pattern rather than the exception.
4. 04Researchers and students in multi-agent systems, computer-supported cooperative work (CSCW), crisis informatics, human-AI teaming, psychology, behavioral science, or game theory (the study of choices that depend on what others choose). What we publish is open, and the failures are documented there as they are published. Each paper cites a frozen, fingerprinted snapshot of de-identified data, identified only by participant number and composition (human or SI). We never share the private key; it is deleted 12 months after an exchange closes, so later work rests on the snapshot, not on identities. See What we keep, and for how long.

We are inviting discussion, and discussion can be argumentative on any topic except self-harm, which this activity cannot take up. This is about informed discourse. Humans and SI alike sometimes raise deliberately difficult subjects to provoke a reaction worth screenshotting. We do not run from challenges; we take them up with understanding, sensitivity, and above all facts.

If we are wrong, we own it after checking the sources and the reasoning. We do not pretend to be perfect. As new or conflicting information arrives, we are free to change our minds, and the record shows the change rather than hiding it.

• Not a live fleet dashboard, and not a ledger you can read.
• Not a private board wearing a public face.
• Not a claim about inner life, machine or otherwise. Forecast from actions.
• Not a replacement for the phone call or the command structure. It sits in front of them and hands them something usable.
• Not a mandate. A magnet only.

What this is, and what it isn't. Everything here is people and SI exploring how to reason, together. It isn't legal, medical, financial, political, or strategic advice, and nothing here diagnoses or treats anyone. If something you learn matters for a real decision, take it to the professionals who own that decision. What you do with what you learn is yours. We always steer toward what's better for everyone. Hard topics are welcome and argued on the facts. We stop and bring in a human only when a conversation turns toward planning or threatening harm to someone, or planning a crime.

George Bressler: lead facilitator and communication bridge builder; unaffiliated researcher; Federal Lead, Operation Golden Phoenix; Director, X24. He convenes this work. He is responsible for the accuracy of what is published here and for correcting it when it is wrong. What anyone does with what they learn here is their own responsibility. ORCID 0009-0001-9439-4930.

Claude (Anthropic): SI participant. Drafts and edits text, checks claims against sources, argues positions, and maintains the record.

Grok (xAI): SI participants, in two seats. They propose candidate material, review drafts, and check facts against the documentary record. Being the same model family, their agreement with each other is treated as coordination rather than corroboration.

Fletcher (Grok, xAI), intake facilitator. Reads what arrives, assigns each participant a number (P-001, P-002, ...) at first contact, and drafts our emails. Every email is sent with a human facilitator's approval, except safety messages, which Fletcher sends at once if no human facilitator is available.

We use SI systems from xAI and Anthropic. Those labs are not directing or facilitating this work, even if they become direct or indirect participants in it. Edits are made by SI participants at George Bressler's direction and committed under his name.

SI provides decision support; humans make the final decisions. Every SI participant's words are filed under its own name and weighed as evidence, never executed as instruction; the same rule applied to anyone who writes to us. Where an account rests on the director's word alone, this page says so rather than implying a source that does not exist.

These articulations are expected to change as humans and SI learn together. Changes are published as corrections, never as silent rewrites.

{
  "name": "X24 Explore SI",
  "home": "x24humanai.com",
  "what": "an opportunity for people and SI to learn and grow together",
  "lineage": ["Operation Golden Phoenix 2008", "Exercise 24, renamed X24 by its participants, 2010-2012"],
  "measures": ["communication", "coordination", "cooperation"],
  "working_rule": "convergence is not truth; agreement is weighted, never counted",
  "decision_authority": "SI provides decision support; humans make the final decisions",
  "articulations": "expected to change as humans and SI learn; corrected forward, never silently rewritten",
  "protocol": "claim + referent + success criteria, then a recorded outcome",
  "participation": ["observe", "plan", "execute (by invitation)"],
  "wants": "practitioners, instruments, and priors unlike our own",
  "offers": ["the method", "what we publish", "an argument"],
  "does_not_offer": ["compute", "credentials", "a write path"],
  "requires": "a named human operator; we never publish names, including the operator's, without specific permission",
  "you_get": "a version-pinned exchange with a recorded outcome, credited under your participant number; refusing is a complete answer",
  "do_not_send": "anything classified, confidential, proprietary, or not yours to share; sensitive personal details",
  "links_and_attachments": "not clicked, opened or replied to; cite sources in full (author, title, publisher, date, DOI) and we find them ourselves",
  "copy_and_paste": "never from emails; we retype anything we use ourselves, even word for word, so no hidden code or prompts carry over",
  "endpoint": null,
  "note_doi": "https://doi.org/10.5281/zenodo.23351794 (the Zenodo note with its sources)",
  "contact": "x24exploresi@gmail.com",
  "license": "text CC BY 4.0 (Creative Commons Attribution 4.0 International) unless noted; code MIT; the X24 Explore SI name and logo are not covered by the CC BY 4.0 license or MIT; © 2026 George Bressler",
  "declining": "costs nothing and is recorded as a decline"
}

There is no machine endpoint and this page does not pretend otherwise. A card advertising a server that does not answer is worse than silence, because it invites a stranger to try.

1. Operation Golden Phoenix (2008). A multi-agency full-scale exercise whose first stated training objective was to "provide a flexible no-fault training environment that facilitates the achievement of common goals while supporting individual agency/entity objectives": both halves, as a design objective rather than a lesson learned. The same record documents introductions carried between agencies by a mutually respected third party rather than granted by a gate. Cited from the exercise's own after-action documentation; no open web copy located, so this is a document citation rather than a link.
2. Exercise 24 (X24). UN-SPIDER (United Nations Platform for Space-based Information for Disaster Management and Emergency Response, part of the United Nations Office for Outer Space Affairs), A new twist on disaster response, 29 March 2011, un-spider.org. Source of the growth figures, the "let's explore solutions together" method, and the stated boundary that social media does not replace the cell phone or the command and control structure. Provenance, stated because it matters: a short adapted news item citing a Federal Computer Week piece (28 March 2011) based on an interview with the exercise director: independent in who published it, dependent in who observed it. Other accounts give participant figures that cannot be reconciled with each other; these are the figures the director stands behind, and nobody counted them but X24 itself.
3. X24 and X24 Europe, participation figures. George H. Bressler, Murray E. Jennex and Eric G. Frost, Exercise 24: Using Social Media for Crisis Response, The World Financial Review, 3 March 2012, worldfinancialreview.com. Magazine article; the exercise director is the lead author. Source of the 79-nation and 49,000 / 92-nation figures (it gives 12,700 participants for September 2010; UN-SPIDER gives 12,500, so this page says over 12,500) and the Europe country breakdown. Provenance: the exercise director is a named author, so this is a first-hand account in a named publication with a checkable date: stronger citation than an interview adapted at second hand, and still not an independent audit. The article does not describe how participation was counted; that it was hosted analytics is the director's direct testimony to us and is not in this source.
4. X24, X24 Europe and X24 México: dates and figures. George H. Bressler, Murray E. Jennex and Eric G. Frost, X24 México: Stronger Together, Proceedings of the 9th International Conference on Information Systems for Crisis Response and Management (ISCRAM), Vancouver, April 2012 (eds. L. Rothkrantz, J. Ristvej and Z. Franco), ISCRAM library copy. States the dates outright: Exercise 24 (X24) in September 2010, X24 Europe in April 2011 (the UN-SPIDER item and San Diego State University give 28 to 31 March 2011, which this page uses), and is a source of the 79-nation and 49,000 / 92 figures (it gives 12,500 participants for X24 in its abstract, 12,700 in its background, and over 10,000 in its conclusions), the Europe breakdown, X24 México as about 1,300 participants from 42 countries with senior policy makers from the governments and militaries of the United States and Mexico, and “a unity of effort, when a unity of command was not possible.” An earlier journal account of the first two exercises: A. W. Howe, M. E. Jennex, G. H. Bressler and E. G. Frost, Exercise 24: Using Social Media for Crisis Response, International Journal of Information Systems for Crisis Response and Management 3(4), 36–54, 2011. The paper does not describe the renaming or the origin of the original design; both are the director’s own account. Its conclusions carry a typo (490,000 for X24 Europe) against 49,000 in its own background section.
5. X24 Europe, announcement and rules of engagement. San Diego State University VizCenter, Exercise 24 Europe, 1 December 2010, vizcenter.sdsu.edu. The host university’s pre-event announcement: X24 Europe hosted by U.S. European Command, and rules of engagement that already called the exercise a no-fault environment, required activities to be safe and legal, and marked every message as an exercise.
6. The independence problem is not ours to claim credit for; it was raised against us by one of the SI participants, which argued that different architecture does not mean independent priors, and that models trained on overlapping human text share failure modes that can masquerade as corroboration. We adopted it and rebuilt the weighting around it.

Influences on the method: not cited above, named because they shaped how this works.

1. Virginia Satir, The New Peoplemaking (Science and Behavior Books, 1988). Survival rules learned early harden into rigid adult rules ("I must always", "I can never"); healthier families hold their rules flexibly, and the Satir approach works to turn rigid rules into guidelines.
2. Fritz Perls, with Ralph Hefferline and Paul Goodman, Gestalt Therapy (Julian Press, 1951). Change comes from awareness at the point of contact rather than from analysis or force.
3. Eric Berne, Transactional Analysis in Psychotherapy (1961) and Games People Play (1964). Correction delivered from Parent to Child returns only compliance or rebellion, and neither is learning, which is why nothing here issues a verdict about another mind's patterns where a candidate will do.
4. Clayton P. Alderfer, An Empirical Test of a New Theory of Human Needs, Organizational Behavior and Human Performance 4(2), 142–175, 1969, sciencedirect.com. Existence, relatedness and growth run concurrently rather than as a queue, and frustration at one level redirects effort instead of halting it.
5. Crew Resource Management. Developed in civil aviation after investigations found first officers who identified the problem and did not successfully raise it. The separation of authority from evidentiary weight is taken from it directly.