L4 BK10 The Real Spy Is Human Nature

Two American counterintelligence and political-influence cases, separated by three decades, raise the same unsettling question: why can experienced people who trust their own judgment fail to recognise danger? Beginning with the stories of Katrina Leung and Christine Fang, this book explores trust, desire, power, artificial intelligence, and the character required to resist self-deception.

LIFE MUSEUM

THE REAL SPY IS HUMAN NATURE

Trust, Desire, Ambition, and the Weaknesses No Institution Can Eliminate

◆

FANG TIANLIANG (FTL)

Seeking Traditional Wisdom · Reflecting on Modern Civilization

CONTENTS

PREFACE	If It Were Me, Would I Have Done Better?
PART ONE	Two Women
Chapter 1	Parlor Maid
Chapter 2	Christine Fang
Chapter 3	Thirty Years Apart
PART TWO	The Enemy Within
Chapter 4	Why Intelligence Does Not Prevent Self-Deception
Chapter 5	The Illusion of Control
Chapter 6	Rules Cannot Substitute for Character
PART THREE	The Recurring Crisis of Civilisation
Chapter 7	Athens
Chapter 8	Rome
Chapter 9	The Ming Dynasty
PART FOUR	Human Nature in the Age of AI
Chapter 10	The Stronger the Tool, the Greater the Responsibility
Chapter 11	When AI Amplifies Human Weakness
Chapter 12	Are We Ready?
PART FIVE	The Real Defence
Chapter 13	Honesty
Chapter 14	Reverence
Chapter 15	A Higher Principle
CONCLUSION	If It Were Me, What Would I Choose?
SOURCES	Selected Sources and Further Reading
AUTHOR	About the Author

Author’s Note on Evidence

This is a work of reflective nonfiction. It uses two widely reported American counterintelligence and political-influence cases as a starting point for a broader inquiry into judgment, self-deception, institutions, technology, and character.

The two cases are not legally or factually identical. Katrina Leung was a long-serving FBI asset whose handling became the subject of a major investigation by the United States Department of Justice Office of the Inspector General. Christine Fang was described in public reporting as a suspected Chinese intelligence operative who cultivated American political figures; she was not charged in the United States, and some details about her activities remain unproven or undisclosed.

Throughout this edition, three kinds of statement are kept separate:

1.   Documented fact — supported by official records, court proceedings, or multiple reliable reports.

2.   Reported allegation — attributed to the public source that reported it.

3.   Authorial interpretation — an attempt to understand motives or draw a moral lesson, not a claim to know another person’s private thoughts.

Where the historical record does not establish a scene, dialogue, motive, or private act, this book does not present it as fact. That distinction matters. A book about intellectual honesty must practise the honesty it recommends.

Preface

If It Were Me, Would I Have Done Better?

Before we begin these two stories, I want to ask you a question.

Not about someone else. About yourself.

Imagine that you are a federal agent who has spent twenty years in counterintelligence. You have investigated suspected spies, tested sources, watched skilled people lie, and learned that misplaced trust can ruin careers and endanger lives. Suspicion is not a personality defect in your profession. It is part of the job.

Then one day you meet someone who does not fit your mental picture of danger.

She is intelligent, composed, and useful. She understands the world in which you work. She seems to recognise the pressure you cannot discuss at home. Nothing about her feels crude or theatrical. Her value appears measurable; her company feels natural.

You tell yourself:

I know exactly what I am doing.

This relationship is under control.

I have seen too many deceptions to fall for one myself.

Now ask yourself honestly: can you be certain that you would never lower your guard?

If your first response is “Of course I would not,” this book is for you.

The people in the following pages were not ignorant of risk. Some were trained to detect it. Yet training did not abolish loneliness, vanity, ambition, fatigue, or the desire to feel understood. Expertise sharpened their judgment in many situations, but it did not grant immunity from being human.

This is therefore not primarily a book about spycraft. It is a book about human nature.

Human nature is indifferent to profession, rank, experience, and IQ. It asks a more uncomfortable question: are we willing to see ourselves clearly when the truth threatens something we want?

I began this book while reading about two cases in the United States, separated by roughly three decades. The first grew out of Cold War counterintelligence in California. The second emerged in the age of social media, campaign networks, and political data. The women were different. Their methods, targets, and legal circumstances were different. Yet the stories produced the same unease in me.

That unease did not come from imagining the women as superhuman villains. It came from recognising the men around them: intelligent, successful, respected people who trusted their own judgment. That trust may have become a blind spot.

I kept asking myself: If it were me, would I have done better?

I could not give a clean answer.

That uncertainty is why I wrote this book. Not to expose private scandal for entertainment, and not to condemn people from the safety of hindsight, but to confront an older and heavier problem:

Why, as human power becomes more sophisticated, do we remain so poor at mastering the weaknesses within ourselves?

Athens asked the question. Rome lived through it. Chinese dynastic histories recorded their own answers. In an age of artificial intelligence, genetic engineering, and machines capable of shaping attention at enormous scale, the question has become more urgent.

This book has five parts.

Part One reconstructs the two cases within the limits of the public record. I am not a judge or a prosecutor. I am a reader trying to understand why warning signs fail to become action.

Part Two examines three recurring mechanisms: motivated reasoning, the illusion of control, and the mistaken belief that rules can substitute for character.

Part Three widens the historical lens to Athens, republican Rome, and the late Ming. Civilisations differ, but the conflict between private appetite and public responsibility returns again and again.

Part Four asks what changes when artificial intelligence makes persuasion faster, cheaper, more personal, and more difficult to detect.

Part Five considers the remaining defences: honesty, humility before our own fallibility, and the cultivation of character.

All five parts point towards one idea:

The decisive weakness is often not outside the gate. It is already inside us.

One final note before you turn the page. I feel no contempt for the people discussed here. Contempt is easy. The harder task is to finish their stories, sit alone, and ask: In their place, under their pressures, wanting what they wanted, might I have made the same choice?

If that question creates even slight discomfort, this book has already achieved half its purpose.

The other half is waiting on the last page.

Part One

Two Women

Chapter 1

Parlor Maid

In 1981 the FBI opened an inquiry into a young Chinese American businesswoman in Los Angeles. Her English name was Katrina Leung; she was also known by the Chinese name Chen Wenying.

The file did not begin as a romance. It began as an ordinary counterintelligence question. Leung had moved into an apartment building associated with people who had attracted FBI interest. She was educated, socially connected, and involved in import-export business. Those features could be entirely innocent. They could also make a person useful to an intelligence service—or useful to the FBI.

An inexperienced agent conducted the initial inquiry. According to the later review by the Department of Justice Office of the Inspector General, the investigation was closed in November 1981 without Leung being interviewed. This detail is important because popular retellings have sometimes placed Special Agent James J. Smith at that first decision and imagined a dinner after which he personally cleared her. The official record does not support that scene.

Smith entered the story later.

In August 1982 he reopened the matter. He was an experienced FBI agent working Chinese counterintelligence in Los Angeles, and he saw in Leung a potential source with unusual access. She moved through social, political, and business circles that were difficult for the Bureau to penetrate. The vetting process did not reveal a reason to reject her, and in December 1982 the FBI formally opened her as an asset.

Her code name was Parlor Maid.

The choice made operational sense. Human intelligence depends on people who can enter rooms that officers cannot. A source’s value often comes precisely from relationships that would look troubling in another context. The difficulty is that the same access that makes an asset valuable can also make the asset dangerous.

Leung proved useful. Over the years she supplied reporting that received favourable evaluations inside the FBI and, at times, the CIA. She passed polygraph examinations in 1984 and 1986. The Bureau paid her more than $1.7 million over approximately eighteen years. None of this proves that every report was true or that every concern was unfounded. It explains why officials became invested in preserving the relationship.

That investment mattered because, by August 1983, Smith and Leung had begun an intimate relationship. It continued for many years. Smith did not report it as required.

This is the point at which a professional problem became a moral and institutional one. The conflict did not consist merely in a private affair. Smith was responsible for judging the reliability of a source with whom he was sexually involved. He was expected to disclose warning signs that could end the very relationship he was concealing. His duty demanded distance; his private life created dependence.

The public record also shows that Leung had a sporadic affair with William Cleveland, another FBI official with experience in Chinese counterintelligence who later worked at Lawrence Livermore National Laboratory. Cleveland denied giving her classified information and passed a counterintelligence polygraph on that question. There is no official finding that he handed secrets to her. That distinction must not be erased for the sake of a more dramatic story.

The real record is dramatic enough.

Warning Signs

By 1990 and 1991, the FBI had received serious indications that Leung had disclosed classified information to officials of the People’s Republic of China. The later Inspector General review concluded that the FBI’s response was inadequate. Concerns were raised, but the Bureau continued to use her.

Smith was not the only person who failed. Supervisors, managers, and parts of the organisation had enough information to ask harder questions. Yet the system treated individual fragments as manageable rather than recognising their combined significance.

Why did Smith not insist on stopping?

The Inspector General’s account offers a sobering answer. Smith later acknowledged that he considered Leung too valuable to lose. He also feared that a serious investigation would reveal their intimate relationship and destroy his career. He admitted that he had been the source of information she disclosed.

Two motives had become entangled: an operational belief—she is too valuable—and a personal fear—the truth will expose me.

This is more illuminating than any imagined scene of midnight photocopying. It shows how a conflict of interest changes perception from within. A person does not need to wake up and decide, “Today I will betray my duty.” He may simply make one exception, then protect the exception, then reinterpret new evidence so that the original exception still appears defensible.

The first compromise changes the cost of honesty. Before the compromise, disclosure may be uncomfortable. Afterwards, disclosure threatens reputation, family, pension, and identity. Each passing year makes truth more expensive.

A Useful Source—and a Dangerous Dependency

It would be convenient if Leung had provided nothing of value. Then everyone around her could be dismissed as foolish. But useful intelligence and counterintelligence risk can coexist. That is what makes source handling difficult.

A source who is always useless will be dropped. A source who is useful enough can acquire institutional protection. Success creates advocates. Positive reports become evidence of reliability. Past investment becomes an argument for future investment. Managers do not merely evaluate the asset; they also evaluate their own earlier decisions.

The problem is therefore larger than seduction. It includes what organisations now call sunk-cost effects, motivated reasoning, and normalisation of deviance. A relationship produces results. A rule is bent. Nothing visibly catastrophic happens. The absence of immediate disaster is mistaken for proof that the exception was safe.

Then the exception becomes routine.

The central question is not simply, “What did Leung do?” It is also, “Why did trained professionals continue to believe that they controlled the risk?”

One answer is emotional. Counterintelligence work imposes unusual isolation. Officers cannot discuss much of what burdens them. They are trained to doubt others and to hide parts of themselves. A person who offers understanding, admiration, usefulness, and intimacy can appear not as a threat but as relief.

Another answer is professional pride. Smith had experience. Experience is often protective, but it can also make a person overconfident in his ability to manage an exception. A novice may fear the rule because he does not understand the danger. An expert may believe he understands the danger so well that the rule no longer needs to apply to him.

That is the more serious temptation.

The Legal Ending

In April 2003 Smith and Leung were arrested. The legal outcomes were narrower than many headlines implied.

Smith pleaded guilty to making false statements. He received probation and a fine. Leung faced charges connected to the alleged mishandling of classified information, but the espionage-related case was dismissed after a judge found prosecutorial misconduct involving the government’s handling of Smith’s plea agreement. Leung later pleaded guilty to lying to the FBI and filing a false tax return. She received probation, community service, and a fine.

These outcomes do not erase the Inspector General’s administrative and counterintelligence findings. Nor do those findings permit us to say that a court convicted Leung of espionage. Criminal guilt, intelligence assessment, and organisational failure are different categories of judgment. Responsible nonfiction must keep them separate.

What the Target Lacked

Popular culture reduces stories like this to a familiar formula: an attractive woman seduces a powerful man. That explanation is shallow.

Effective cultivation is rarely generic. It supplies what a particular person lacks.

One person wants affection. Another wants recognition. A third wants access, prestige, money, family warmth, or the feeling of being understood. The most powerful offer is the one that fits the private deficit the target has never named aloud.

For Smith, Leung appears to have combined several forms of value: access to difficult networks, intelligence that advanced investigations, and a relationship in which professional and emotional needs became inseparable. The attraction was not only sexual. It involved usefulness, identity, and the belief that he was still in command.

That combination explains why warning signs could be recognised intellectually and still fail to produce action. To accept the warnings would not merely mean losing a source. It would mean admitting that a central professional achievement and a central private relationship had become liabilities.

The mind resists conclusions that threaten too many things at once.

This chapter therefore leaves us with a harder lesson than “do not trust.” No intelligence service can function without trust. No human relationship can exist without it. The question is whether trust remains accountable to evidence, and whether the person who grants it is willing to submit his own judgment to independent review.

The danger begins when we say:

My case is different.

My experience protects me.

I can stop whenever I choose.

Human nature offers no professional immunity.

Chapter 2

Christine Fang

Roughly three decades after the FBI opened its first inquiry into Katrina Leung, a young Chinese national named Christine Fang—also known as Fang Fang—appeared in the political life of California’s Bay Area.

The world around her was different. Political relationships were now organised through social media, university associations, community events, campaign databases, and networks of donors and volunteers. Influence no longer required access to a locked briefcase. It could begin with introductions, photographs, fundraising, and the patient cultivation of people who might one day hold greater power.

According to an investigation published by Axios in 2020, U.S. intelligence officials believed Fang was involved in a political-intelligence operation directed by China’s Ministry of State Security. She attended California State University, East Bay, became active in Chinese student and Asian American civic organisations, and developed relationships with local and national politicians between about 2011 and 2015.

Fang was not charged with a crime in the United States. She has not been publicly tried, and the full intelligence file has not been released. She should therefore be described as a suspected intelligence operative, not as a convicted spy.

That caution does not make the reported pattern unimportant.

The Currency of Political Access

A young local politician usually lacks what an established politician takes for granted: donors, volunteers, community recognition, invitations, and the confidence of influential networks. Someone who can supply those things becomes valuable quickly.

Fang reportedly helped politicians connect with donors and community groups, appeared at campaign events, and encouraged political participation. Federal fundraising rules prevented her, as a foreign national, from making lawful campaign contributions herself. Public reporting did not establish that she made illegal contributions. Her value lay in bundling social and political access—introducing people, organising events, and making herself useful.

Among the politicians she cultivated was Eric Swalwell, then a rising figure in local government who was elected to Congress in 2012. Axios reported that Fang helped place an intern in his congressional office and participated in fundraising activity on his behalf. The report did not allege that Swalwell disclosed classified information to her. When the FBI gave him a defensive briefing around 2015, he cut off contact and cooperated with investigators. He was not accused of wrongdoing in the matter.

These facts are less sensational than some online retellings, but they reveal the actual security problem: influence often grows through ordinary, lawful-looking political behaviour. The person who makes introductions today may gain proximity to decision-makers tomorrow.

What Public Reporting Does—and Does Not—Show

Axios reported that Fang had sexual or romantic relationships with at least two Midwestern mayors. Those allegations were attributed to intelligence officials and surveillance described by the news organisation. No reliable public evidence has established a sexual relationship between Fang and Swalwell.

This distinction matters. Political partisans later blurred “Fang cultivated Swalwell” into “Swalwell had an affair with a Chinese spy.” The first statement is supported by public reporting; the second is not.

The same discipline applies to later events. Swalwell resigned from Congress on 14 April 2026 after separate allegations of sexual misconduct led Democratic leaders to call for an investigation and for him to end his campaign for governor. His resignation should not be presented as the delayed result of the Fang case unless evidence establishes that connection. As of this edition, the public record does not.

A writer who condemns self-serving belief must not select facts according to political preference. Accuracy is not a favour to the subject. It is a duty to the reader.

The Confirmation of Power

What, then, can the Fang case teach us?

For an ambitious politician, help is never merely help. A room full of volunteers says, People believe in you. A successful fundraiser says, Your future is real. An energetic organiser says, I can take you where you want to go.

This is why political cultivation can be emotionally powerful even without romance. It offers a confirmation of destiny.

Most politicians understand that supporters have motives. They may tell themselves that politics is an exchange: the organiser gains status and access; the candidate gains money, visibility, and votes. Because the exchange appears mutual, the politician assumes it is controlled.

But repeated dependence changes the relationship. The organiser becomes embedded in the network. Staff learn to call her. Donors recognise her. Invitations pass through her. What began as one useful contact becomes part of the candidate’s political infrastructure.

The vulnerability is not stupidity. It is the tendency to treat a person who consistently advances our goals as safer than the evidence warrants.

A Modern Form of Cultivation

Leung’s value to the FBI was connected to information and access inside difficult communities. Fang’s reported value to politicians was connected to mobilisation and access inside electoral communities. The methods were different, but both depended on becoming useful.

Usefulness lowers suspicion because it gives the target reasons to preserve the relationship. The target can always offer a respectable explanation:

She understands the community.

She brings people into public life.

She is helping us reach voters.

All those statements may be true. They do not answer the counterintelligence question: Why is this person investing so much effort, and who else benefits from the access?

In 2015, after the FBI’s concerns became serious enough to brief targeted politicians, Fang left the United States. The precise reasons for her departure and the full scope of her activity remain outside the public record.

The responsible conclusion is therefore measured. We cannot declare every reported allegation proven. We can say that the case illustrates a longstanding method of influence: identify promising people early, provide what their ambitions require, and build a relationship before their future power makes access difficult.

The most effective moment to cultivate a leader is often before the leader feels important enough to be a target.

Chapter 3

Thirty Years Apart

In 1981 Katrina Leung entered an FBI inquiry in Los Angeles. Around 2011 Christine Fang entered the political networks of the Bay Area.

Between those dates, intelligence work changed dramatically. Paper files gave way to digital databases. Campaigns learned to organise through social media. Personal histories became searchable. Networks that once took years to map could increasingly be inferred from data.

Yet one element remained stable: access still passed through human need.

The cases should not be collapsed into one story. Leung became a paid FBI asset and maintained a long intimate relationship with her handler. Fang was a suspected foreign operative who cultivated multiple political figures and left after an FBI investigation. The evidence, legal posture, and institutional failures differ.

Their resemblance lies at a deeper level.

The Pattern

First, neither woman necessarily matched the target’s imagined picture of danger. People search for threats using mental templates. A person who appears educated, helpful, socially fluent, and emotionally attentive can pass outside that template.

Second, each relationship produced value. Leung supplied intelligence. Fang reportedly supplied contacts, volunteers, and fundraising connections. Value creates its own defence: the more useful a relationship becomes, the more costly it feels to question.

Third, ambition and identity were involved. Smith’s identity rested partly on being a skilled counterintelligence officer capable of handling a difficult source. A politician’s identity rests partly on being a leader whose rise attracts believers. Evidence that threatens a relationship may therefore threaten the target’s idea of himself.

Fourth, outside intervention mattered. Smith did not resolve the conflict by voluntarily disclosing the affair and withdrawing from the case. The Bureau eventually acted after accumulating counterintelligence evidence. In the Fang matter, defensive briefings by the FBI caused politicians to break contact.

The people closest to a compromised relationship are often the least able to evaluate it. Distance is not bureaucratic inconvenience; it is a form of protection.

The Immunity Illusion

I call the underlying belief the immunity illusion:

This may happen to other people, but it will not happen to me.

The illusion grows stronger when a person has real skill. Success teaches us that our judgment works. Usually that lesson is useful. Under pressure, however, it can become a claim of exemption: because I have handled danger before, I may ignore the safeguard designed for everyone.

The greatest risk may arise not when we know nothing, but when we know enough to produce a persuasive explanation for what we want to do.

Technology changes the instrument. Human nature supplies the opening.

The first defence, then, is not to imagine ourselves immune. It is to accept that the right combination of need, opportunity, and self-justification can mislead anyone—including the person reading this page, and the person writing it.

Part Two

The Enemy Within

Chapter 4

Why Intelligence Does Not Prevent Self-Deception

When people hear the stories of Leung and Fang, they often ask how experienced officials could fail to recognise danger. The question usually contains a hidden assumption: intelligent people should be difficult to deceive.

Intelligence certainly helps us detect inconsistencies, compare explanations, and learn from experience. But it also gives us a larger vocabulary with which to defend a preferred conclusion. The mind that can construct a good argument for the truth can construct a good argument for desire.

The problem is not that intelligence makes people foolish. It is that intelligence does not automatically make them impartial.

The Limits of Self-Assessment

In 1999, psychologists Justin Kruger and David Dunning published a series of studies on the relationship between performance and self-assessment. Participants who performed poorly on tests of humour, grammar, and logic often overestimated their relative performance. The researchers argued that some of the skills required to perform well are also required to recognise poor performance.

The finding has often been reduced to a slogan: “stupid people think they are smart.” That is neither precise nor particularly useful. Later scholars have debated how much of the observed pattern reflects metacognitive deficits, statistical effects, and the familiar human tendency to rate ourselves above average.

The enduring lesson is more modest: self-confidence is not a reliable measure of accuracy. Expertise is domain-specific. A person can be excellent at analysing an adversary and poor at analysing his own emotional involvement.

Smith’s professional experience could help him evaluate sources. It could not make him a neutral evaluator of a source with whom he had a secret intimate relationship. Indeed, his expertise may have made the conflict harder to acknowledge, because acknowledging it would damage his identity as an expert.

Motivated Reasoning

Suppose two people examine the same evidence. One has nothing to lose if the evidence proves alarming. The other risks losing a relationship, a source, a promotion, and a reputation. Are they likely to weigh the evidence in the same way?

Motivated reasoning describes the tendency to process information in ways that protect goals, identities, or desired conclusions. We do not always begin with evidence and move calmly toward judgment. Often a hope or fear quietly sets the destination, and reasoning finds a route.

The process can be sincere. A person need not consciously lie. He may demand impossible proof from threatening evidence while accepting weak reassurance at once. He may remember facts that support his position and forget facts that disturb it. He may describe caution as prejudice, or a warning as the product of a rival department.

This is why the sentence “I honestly believe it” does not settle a dispute. Honesty about one’s conclusion is not the same as honesty about how the conclusion was formed.

Cognitive Dissonance

Cognitive dissonance is the discomfort produced when beliefs, actions, and identity conflict. If I believe I am a careful professional but discover that I have violated a fundamental rule, I have several ways to reduce the pain.

I can admit the violation and change my behaviour. Or I can change the meaning of the violation:

The rule is too rigid.

This source is exceptional.

My results justify the method.

People criticising me do not understand the operation.

The second path is psychologically cheaper in the moment. Over time it becomes expensive, because each new fact must be fitted into an increasingly fragile story.

In the Leung case, disclosure would have threatened both a personal relationship and a professional legacy. That does not prove exactly what Smith thought on any particular day. It identifies a powerful incentive to interpret doubt as manageable.

Political life creates similar pressures. A politician who has accepted years of help from an organiser does not evaluate a warning as a stranger would. The warning also says: Your judgment of people may have been wrong. Your success may have depended on access you did not understand. The mind naturally resists that message.

The Island of Power

Power can reduce corrective feedback. Subordinates learn which objections are welcomed and which are punished. Supporters praise the leader’s instincts. People who want access rarely begin by saying, “Your judgment is compromised.”

The result is an island: the person at the centre receives more information than almost anyone else, but less honest contradiction.

This is not inevitable. Some leaders deliberately protect dissent, rotate advisers, and require independent review. But these practices work only if disagreement is rewarded in reality, not merely praised in a handbook.

The strongest warning sign is not that a leader makes a mistake. Everyone does. It is that the surrounding system makes it dangerous to say he may be making one.

The Bias Blind Spot

When people learn about motivated reasoning, they quickly recognise it in opponents. Political rivals are biased. Colleagues are defensive. Family members refuse to face facts.

The ability to diagnose others can itself become another form of self-protection.

Psychologists have described a “bias blind spot”: people often see cognitive bias more readily in others than in themselves. This makes sense. We observe other people’s conclusions, but we experience our own reasons from the inside. Our reasons feel like reasons; theirs look like motives.

The proper conclusion is not that evidence is useless or that objectivity is impossible. It is that objectivity is a discipline rather than a personal possession.

That discipline asks practical questions:

·    What evidence would change my mind?

·    Am I applying the same standard to information I welcome and information I fear?

·    Who can review this decision without sharing my personal stake?

·    If a colleague made the same exception, would I accept the explanation?

These questions do not eliminate bias. They make bias harder to hide.

The most intelligent person in the room is not necessarily the safest decision-maker. The safest may be the person most willing to say, “My intelligence does not exempt me from review.”

Chapter 5

The Illusion of Control

“I can control myself.”

The sentence sounds responsible. Sometimes it is. Human beings can resist impulses, keep promises, and act against immediate desire. Civilisation depends on that capacity.

The danger lies in treating self-control as unlimited and unaffected by context.

What Situations Can Do

In 1971, Philip Zimbardo and colleagues conducted the Stanford prison experiment. College students were assigned to act as guards or prisoners in a simulated jail. The study ended after six days.

For decades it was presented as proof that ordinary people quickly become cruel when given a role and a permissive environment. Later investigations raised serious questions about the study’s methods, including researcher involvement, instructions given to guards, and the reliability of its most dramatic claims. It should not be treated as a clean demonstration of an iron law of human behaviour.

Yet the controversy strengthens an important lesson: powerful stories about human nature also require independent scrutiny. We should not accept a psychological parable merely because it confirms what we already believe.

Better-supported research shows that fatigue, sleep loss, stress, alcohol, strong emotion, social pressure, and conflicts of interest can impair judgment. None removes moral responsibility. Each changes the conditions under which responsibility must be exercised.

The Disputed “Willpower Muscle”

The theory of ego depletion proposed that self-control operates like a limited resource: exerting it on one task leaves less available for the next. Early experiments appeared to support the idea, and the “willpower muscle” became popular.

Large replication efforts and later analyses produced mixed results. The original strong version of the theory remains contested. It would therefore be wrong to say science has proved that willpower is a fixed fuel tank.

The practical insight survives in a more careful form. People do not make decisions with identical clarity at every hour and under every pressure. A person who is exhausted, frightened, lonely, or intoxicated should not rely on a confident feeling of control as proof of sound judgment.

This is why aviation crews use checklists, surgeons pause before an incision, financial institutions separate duties, and intelligence services impose handling rules. The safeguard is designed in advance, when judgment is calm, because the critical moment may not be calm.

The Exceptional Case

Most serious violations begin with a reason that appears plausible.

The meeting location is inconvenient, so the officer chooses a private home. The source is uniquely valuable, so normal documentation is delayed. The deadline is urgent, so a second review is skipped. The relationship is complicated, so disclosure can wait until the right moment.

Each decision is described as temporary. But the first exception changes the next decision. Once a person has crossed a line, admitting danger also means admitting the first crossing was wrong.

The story becomes:

Nothing bad happened last time.

I know the risks.

I can stop before it goes too far.

The absence of immediate harm is not evidence of control. A driver may speed a hundred times without crashing. Repetition proves only that the risk has not yet matured into consequence.

Designing for the Person We Become Under Pressure

We often design our moral lives for an imaginary self: rested, rational, unhurried, and free of desire. But the decisive choice may be made by a different self—tired, flattered, afraid of exposure, or desperate not to lose an opportunity.

A reliable defence therefore has two parts.

The first is environmental. Do not create situations in which private desire and professional duty must fight alone. Disclose conflicts early. Meet in accountable settings. Share decisions. Preserve records. Require someone independent to approve exceptions.

The second is moral. No procedure can force a person to reveal the fact he most wants to hide. At some point he must prefer the embarrassment of disclosure to the comfort of concealment.

Self-control is real. Wisdom does not deny it. Wisdom refuses to gamble everything on it.

Chapter 6

Rules Cannot Substitute for Character

Modern organisations respond to failure by writing rules. This is often necessary. A rule can preserve hard-won knowledge long after the people who learned the lesson have left.

But rules contain a paradox: they must be interpreted and enforced by the same fallible human beings they are designed to restrain.

Why Doctors Wash Their Hands

In the 1840s, Ignaz Semmelweis observed that women giving birth in a clinic staffed by doctors and medical students died of puerperal fever at far higher rates than women in a clinic staffed by midwives. He required hand cleansing with a chlorinated solution, and mortality fell sharply.

His explanation preceded the full acceptance of germ theory and met resistance. The history is more complicated than the legend of one hero opposed by every doctor in Europe, but the moral remains powerful. Professional status does not make hands clean. A procedure can protect patients from dangers that expertise alone cannot see.

The checklist is not an insult to the expert. It is respect for reality.

Compliance on Paper

An organisation may have excellent rules and still cultivate bad behaviour. Employees learn the difference between what leadership says and what it rewards. If deadlines matter more than safety, people will reinterpret safety rules. If loyalty to a manager matters more than candour, warnings will be softened. If careers depend on reporting success, ambiguous evidence will become good news.

This is how compliance becomes theatre. Forms are completed, boxes checked, and the moral purpose of the rule disappears.

The Leung case was not simply the failure of one agent to follow policy. The Inspector General found warning signs that the Bureau did not pursue adequately. The organisation had information, yet information did not become decisive action.

Rules require a culture in which acting on an unwelcome fact is safer than concealing it.

Columbia

On 1 February 2003, the space shuttle Columbia disintegrated during re-entry, killing all seven crew members. During launch, a piece of insulating foam had struck the shuttle’s left wing. Engineers later sought additional imagery that might have clarified the damage. Management did not obtain the requested high-resolution images, in part because previous foam strikes had not caused catastrophe and the risk had become familiar.

The Columbia Accident Investigation Board described organisational causes alongside the physical cause. NASA’s history, schedule pressures, communication failures, and acceptance of recurring anomalies shaped the decision.

Sociologist Diane Vaughan had used the phrase normalisation of deviance in her study of the earlier Challenger disaster: a departure from the intended standard can become normal when it is repeatedly followed by apparent success.

The pattern is simple:

4.   A rule or safety boundary is crossed.

5.   No disaster occurs.

6.   The crossing is reclassified as acceptable.

7.   The new boundary is crossed again.

Eventually the organisation forgets that the practice was ever considered dangerous.

The Inner Guardrail

Institutions need policies, audits, inspectors, and consequences. It would be foolish to replace them with vague appeals to virtue. Character without structure is unreliable; structure without character is hollow.

The decisive moment often arrives before the inspector knows there is a problem. Someone sees data that damages the project. Someone knows a relationship should be disclosed. Someone realises that the official account is incomplete.

No manual can make that person speak.

The act requires an inner guardrail: a willingness to accept personal cost rather than transfer hidden risk to others.

This is what I mean by character. It is not moral perfection. It is the trained habit of refusing to use intelligence, status, or procedure as permission to deceive oneself.

I cannot honestly say that I would have acted better than every person in these cases. I have not carried their pressures or faced their temptations.

I can say this: if I begin by believing that I am too experienced to be compromised, I have already removed one of my strongest defences.

That defence is humility expressed as practice: disclose, verify, invite contradiction, and accept the rule even when I believe my case is special.

Part Three

The Recurring Crisis of Civilisation

Chapter 7

Athens

Fifth-century Athens created a form of democratic life that still shapes political language. Citizens debated public policy, served on juries, held offices, and watched drama that examined power, pride, and responsibility. The city also built an empire, fought long wars, and made decisions of astonishing recklessness.

Its history warns us against a simple belief: that a well-designed political system will automatically produce wise choices.

After Pericles

Pericles dominated Athenian politics for many years, but he did not rule alone or without opposition. His power rested on persuasion within a democratic system. When the Peloponnesian War began in 431 BCE, he urged the Athenians to avoid land battles with Sparta, protect their fleet, and rely on the resources of their maritime empire.

Then plague struck the crowded city. Pericles died in 429 BCE. Athens did not collapse at once; it continued the war for decades and sometimes recovered. But the political environment became more vulnerable to leaders who promised emotional satisfaction, quick victory, or revenge.

Thucydides’ history repeatedly shows citizens making choices under anger, fear, hope, and wounded pride. In the Mytilenean debate, the Assembly first voted to kill the adult men of a rebellious allied city, then reconsidered the next day and narrowly reversed itself. Democracy created the possibility of correction—but only after passion had nearly produced atrocity.

The Sicilian Expedition

The most famous example came in 415 BCE, when Athens decided to send a massive expedition to Sicily. Supporters imagined strategic gain, wealth, and an expansion of power. Critics warned that the city was opening a distant front while the war with Sparta remained unresolved.

Alcibiades, brilliant and ambitious, encouraged the expedition. Nicias, who opposed it, tried to discourage the Assembly by describing the enormous force required. His strategy failed. The scale of the proposed armament made the adventure appear more impressive, not less.

The expedition ended in disaster in 413 BCE. Ships were destroyed, soldiers killed or enslaved, and Athenian power severely weakened.

It is tempting to call the decision irrational and move on. A more useful question is how intelligent citizens made it appear rational. They possessed arguments. They assessed alliances, resources, and geography. But desire shaped which possibilities felt vivid. The rewards of victory were imagined in colour; the costs of failure remained abstract.

This is motivated reasoning at the level of a city.

Democracy’s Psychological Vulnerability

Democracy does not eliminate human weakness. It distributes power and creates opportunities for criticism, reversal, and accountability. Those are enormous virtues. But democratic citizens can still reward flatterers, confuse confidence with competence, and prefer a pleasing story to an unwelcome fact.

The safeguard is not less democracy. It is a democratic culture mature enough to value restraint, evidence, and dissent.

Athens shows both sides. The Assembly could be swept by passion; it could also reconsider. The people could elevate reckless ambition; they could also listen to argument. Institutions created the opportunity for wisdom, not a guarantee of it.

When citizens lose the character required to use freedom responsibly, the machinery of freedom can be turned toward self-destruction.

Chapter 8

Rome

The Roman Republic lasted for centuries without a king. Its mixed constitution divided authority among magistrates, the Senate, and popular assemblies. Offices were temporary. Colleagues could restrain one another. Tradition imposed limits beyond written law.

Rome’s collapse did not occur because every institution disappeared on one day. It occurred as ambitious people discovered that violating a norm could produce rewards, and as rivals answered each violation with a greater one.

The First Broken Boundaries

In 133 BCE, the tribune Tiberius Gracchus pursued land reform through methods his opponents considered an assault on constitutional custom. A group of senators and their supporters killed him. His brother Gaius was killed in another eruption of political violence in 121 BCE.

The policies, motives, and constitutional questions remain debated. The deeper turning point is clearer: organised killing entered republican politics as a method of settling disputes.

Once that line had been crossed, later actors inherited a changed world. They could still condemn violence, but they also knew violence worked.

Armies and Personal Loyalty

Rome’s overseas expansion brought wealth, slaves, provincial commands, and opportunities for glory. It also increased inequality and made military careers more dependent on successful commanders.

In 88 BCE, Lucius Cornelius Sulla marched his army on Rome. Roman armies had been instruments of the Republic; now a commander used one against the city to secure political control. Sulla’s enemies later answered with force. He returned, established a dictatorship, and published lists of enemies who could be killed and dispossessed.

Decades later, Julius Caesar crossed the Rubicon with his army in 49 BCE. The act did not create the Republic’s crisis from nothing. It was possible because generations had normalised the use of exceptional commands, political intimidation, private armies, and emergency measures.

Each faction claimed necessity. Each said the other side had violated the rules first. Each exception became precedent.

When Institutions Become Shells

Roman offices and ceremonies continued even as their meaning changed. The Senate met. Elections occurred. Laws were passed. Yet the decisive questions increasingly depended on money, personal armies, intimidation, and patronage.

An institution can survive in form after it has weakened in spirit. People continue to speak its language while using it for purposes contrary to its design.

Roman writers themselves worried about luxury, corruption, and the loss of virtus—a demanding ideal of civic courage and service. Their moralising should not be accepted without question; nostalgia for an imagined pure past is common in declining republics. Nevertheless, their concern identified a real political problem. A constitution cannot defend itself if powerful citizens value victory more than the shared rules that make peaceful rivalry possible.

The lesson is not that private interest can be abolished. Rome was never a society without ambition. The lesson is that ambition must remain bounded by something its possessor will not sacrifice even when sacrifice would bring victory.

When every side treats restraint as unilateral disarmament, all sides help destroy the arena in which lawful competition can exist.

Chapter 9

The Ming Dynasty

In 1644, as rebel forces entered Beijing, the Chongzhen emperor took his own life on Coal Hill, now known as Jingshan. Later tradition attributed bitter final words to him blaming disloyal ministers for the fall of the dynasty. Different versions of his death note circulated, and the most familiar quotations should not be treated as certain transcripts.

The temptation to assign the fall of the Ming to one emperor, one faction, or one moral defect is strong. History is less tidy.

A Convergence of Crises

The late Ming faced fiscal strain, military pressure from the rising Manchus, peasant rebellions, factional conflict, administrative weakness, climatic shocks associated with the Little Ice Age, harvest failures, epidemics, and disruption in silver flows that affected taxation and commerce.

No single cause explains collapse. External pressure and internal weakness interacted.

This complexity does not remove the moral question. It sharpens it. Institutions under stress depend on people who can distinguish public duty from factional advantage. When resources shrink and danger grows, incentives to conceal bad news, shift blame, hoard wealth, and attack rivals become stronger.

The Signal Sent by Power

The Wanli emperor reigned for forty-eight years and for long periods withdrew from many routine functions of government. The causes included political conflict over succession and resistance from officials, but the effect was damaging: high office appeared detached from sustained responsibility.

Earlier, Grand Secretary Zhang Juzheng had pursued major fiscal and administrative reforms, including measures associated with the Single Whip system. After his death in 1582, his enemies attacked his legacy and his family suffered confiscation and punishment.

Whatever one thinks of Zhang’s methods, the reversal sent a dangerous signal. Reform that harms entrenched interests can expose the reformer and his family after political protection disappears. Officials learn from such signals. They learn which risks the system rewards and which it punishes.

An organisation may announce that it values initiative while teaching everyone that initiative is fatal.

Corruption and Incentive

Late-Ming stories often describe wealthy officials and imperial relatives refusing to contribute funds while the state struggled to pay troops. Some traditional accounts give enormous totals allegedly extracted by Li Zicheng’s regime after it entered Beijing. The precise figures are disputed and may be exaggerated.

The broader problem is credible. The state had difficulty converting private wealth into legitimate, sustainable public revenue. Tax burdens were uneven. Officials and powerful families could protect resources. Soldiers went unpaid. Local distress fed rebellion.

It is easy to denounce selfish ministers. It is harder to examine the system that made selfishness rational. If honest service brings danger while factional loyalty brings protection, good people face a cruel choice: compromise, withdraw, or be destroyed.

This does not excuse corruption. It explains how corruption becomes systemic. A bad system does not require everyone to be bad. It requires enough people to believe that goodness will be punished and opportunism rewarded.

History’s Question

Athens, Rome, and the Ming were not destroyed by the same mechanism. Athens survived the Sicilian disaster for years before defeat. The Roman Republic transformed through generations of conflict. The Ming fell amid military, fiscal, environmental, and political crises.

Yet each history asks a related question:

What happens when enough people responsible for a system place immediate private advantage above the conditions that allow the system to endure?

External enemies matter. Material conditions matter. Institutions matter. But none can be understood without the choices of human beings inside them.

Look at the system in which you work or live.

Does it reward honesty, or performance of honesty?

Does it reward long-term responsibility, or the transfer of costs to someone else?

Can people who identify danger speak without destroying their careers?

The answers may predict the future of the system more accurately than its mission statement.

Part Four

Human Nature in the Age of AI

The first case in this book grew out of a world of paper files, face-to-face meetings, and human sources. The second belongs to a world of online networks, campaign data, and social media.

The next world is arriving now.

Artificial intelligence can analyse enormous quantities of personal information, imitate voices, generate persuasive language, produce realistic images, and adapt messages to individual users. These capabilities can improve medicine, education, research, accessibility, and public services. They can also make manipulation cheaper, faster, and more intimate.

The central danger is not simply that machines may become powerful. It is that powerful machines will be directed by human beings who remain vulnerable to greed, fear, vanity, tribal loyalty, and self-deception.

The tool changes. The opening remains human.

Chapter 10

The Stronger the Tool, the Greater the Responsibility

At 8:15 on the morning of 6 August 1945, an American bomber dropped an atomic bomb over Hiroshima. The explosion and its aftermath killed tens of thousands of people and transformed the moral horizon of science and war.

The Manhattan Project had concentrated extraordinary intellectual ability on a single technical objective. The scientists succeeded. Success did not answer the political and moral question of how the invention should be used.

J. Robert Oppenheimer later recalled that the Trinity nuclear test brought to mind a line from the Bhagavad Gita: “Now I am become Death, the destroyer of worlds.” The familiar quotation is often used as a complete verdict on his thoughts, but his reflections were more complex. What matters here is the recognition that technical achievement can create responsibility larger than the technical task.

Power Is Not Purpose

People often call technology neutral. The phrase contains a truth and a danger.

A tool does not possess a human conscience. The same model may help detect cancer, translate a medical instruction, generate propaganda, or design a phishing campaign. Human purpose and institutional context determine the use.

Yet design is not irrelevant. Tools make some actions easier than others. A platform rewarded for attention will favour material that captures attention. A system optimised for persuasion will search for messages that change behaviour. Neutrality cannot mean indifference to foreseeable misuse.

Nuclear weapons required states, laboratories, rare materials, and industrial infrastructure. Powerful AI services can be distributed to millions of users at once. This difference lowers the cost of beneficial innovation—and of deception.

The Personal Intelligence File

Imagine a modern cultivator studying a target. She no longer needs months of conversation to learn his preferences. Public posts reveal political identity, friendships, travel, family events, grievances, and ambitions. Commercial data may infer purchases, locations, and interests. A model can summarise the material and suggest which message is most likely to win trust.

The result is not magical mind-reading. Data can be incomplete or wrong. Human beings remain unpredictable. But influence does not require perfect knowledge. It requires an advantage.

A tailored message may praise the target’s neglected talent, repeat his private fear, or present an opportunity at the moment he feels most frustrated. At scale, an operator can test thousands of variations and keep the ones that work.

Traditional cultivation asks, “What does this person lack?”

AI makes the question cheaper to ask and easier to answer.

AI Can Defend—and Rationalise

The same technology can help the defender. It can identify unusual contact patterns, compare statements, detect document anomalies, and alert organisations to undisclosed conflicts.

But a defensive tool is only as honest as the question asked of it.

A person who wants reassurance can ask a model to assemble evidence that a relationship is safe. A person who wants caution can ask it to identify reasons the relationship may be dangerous. Fluent systems can produce confident arguments in either direction, especially when the evidence is ambiguous.

AI may therefore become a machine for motivated reasoning: not because it has human desire, but because it can provide endless intellectual support for the user’s desire.

The discipline of the future will not be merely “use AI.” It will be:

·    state the decision before requesting advice;

·    require the system to argue the strongest opposing case;

·    verify important claims against primary evidence;

·    record what information was unavailable;

·    keep accountable human review for consequential decisions.

Greater analytical power does not remove the need for character. It increases the damage that self-deception can do.

Chapter 11

When AI Amplifies Human Weakness

Falsehood did not begin with artificial intelligence. Rumour, forgery, impersonation, and propaganda are ancient. What changes is scale, speed, cost, and personalisation.

From Broadcast to Tailoring

Traditional propaganda broadcasts one message to many people. Algorithmic systems can show different messages to different people, learn from their reactions, and adjust.

The most effective message is not always the one that changes a belief. It may be the one that intensifies an existing belief until compromise feels like betrayal. It may not persuade citizens to support the same candidate; it may persuade each group that every other group is dangerous.

This strategy exploits confirmation bias—the tendency to notice and accept information that fits what we already believe. Social platforms did not invent the bias. Their ranking systems can amplify it by repeatedly supplying emotionally engaging confirmation.

We should be careful with sweeping claims. Educated people are not uniformly more likely to share false news, and no single demographic explains misinformation. Age, political identity, digital literacy, habit, platform design, and emotional arousal can all matter. The honest lesson is not that “smart people spread more lies.” It is that education alone does not guarantee disciplined verification.

Synthetic Evidence

Generative systems can produce convincing photographs, video, and cloned voices. Quality varies, and many fakes can still be detected. But detection is an arms race, and ordinary users rarely have time to conduct a forensic examination before reacting.

The first danger is believing a fake.

The second is the liar’s dividend: once people know realistic fakes exist, a guilty person can dismiss real evidence as artificial. Shared reality is damaged in both directions. We may accept what never happened and deny what did.

In intelligence work, a forged recording could provoke distrust, damage an alliance, or force officials to spend precious time proving a negative. In family life, a cloned voice can imitate a relative asking for urgent money. In politics, synthetic material can arrive too late in an election for verification to catch up with emotion.

Emotional Accuracy Without Factual Accuracy

Manipulative content succeeds when it feels emotionally true. A fabricated story may fit a group’s belief that elites are corrupt, foreigners are hostile, or opponents are immoral. Because the emotion is genuine, the evidence feels genuine.

This is the digital form of the old vulnerability. Smith’s conflict did not require every statement around him to be false. The relationship only had to make one conclusion feel natural: I understand this source; I can manage the risk.

AI can create an environment tailored to produce the same feeling in each of us.

The answer cannot be permanent suspicion. A society in which nobody trusts any evidence is as vulnerable as one in which everyone trusts the first image on a screen.

We need calibrated trust: confidence proportional to evidence, source, context, and consequence.

Before sharing a consequential claim, ask:

·    Who first published it?

·    Is the original material available?

·    Has an independent source confirmed it?

·    Does the claim arrive at a moment designed to prevent reflection?

·    Would I demand the same evidence if the story damaged my own side?

These questions are slow. That is their value. Manipulation depends on speed.

Chapter 12

Are We Ready?

Every technological revolution tests more than technical skill. It tests law, institutions, imagination, and moral restraint.

Human beings living today are biologically continuous with ancestors who depended on small groups for survival. We remain highly sensitive to belonging, exclusion, status, threat, and reward. Cultural evolution can be rapid, and human behaviour is flexible, but our basic emotional architecture did not redesign itself when the smartphone appeared.

Technical capability, by contrast, can improve in months.

This mismatch is one of the defining problems of the age: tools may change faster than the habits required to use them wisely.

Beyond Regulation Alone

AI policy often focuses on copyright, employment, safety testing, privacy, competition, and national security. These questions are necessary. Law can establish accountability and protect rights. Technical standards can reduce risk. Institutions can audit systems and restrict dangerous uses.

But no regulation can answer every moral choice made by every engineer, executive, official, or user.

The final question remains human:

When my interest conflicts with the long-term interest of people who cannot see my decision, what will I choose?

A developer may discover a safety problem shortly before a product launch. A manager may know that an impressive benchmark hides a weakness. A government may possess a surveillance capability that is legal under a broad interpretation but corrosive to liberty. A user may have the power to fabricate an image that will humiliate an enemy.

Rules matter in each case. So does the person who decides whether to exploit the gap between what is permitted and what is right.

The Myth of Moral Outsourcing

We sometimes speak as if a sufficiently advanced system will solve the problem of human judgment. The machine will be objective. The algorithm will remove bias. The model will tell us what to do.

But models learn from choices about data, labels, objectives, thresholds, and acceptable error. They operate within institutions that decide whose harms matter and whose benefits count. Human judgment does not disappear. It moves upstream and becomes harder to see.

We cannot outsource conscience.

The future of AI will not be decided only by whether machines become more capable. It will be decided by whether human beings become more accountable for the purposes to which capability is put.

That brings us back to the central claim of this book. The deepest defence does not begin with a stronger external enemy detector. It begins with an honest recognition of the weaknesses that an enemy—or an employer, a political movement, a platform, or our own desire—can exploit.

Part Five

The Real Defence

What can human beings do?

We need institutions strong enough to restrain power and people formed strongly enough to respect those institutions when violation would be profitable.

Power must be bounded by moral responsibility.

Knowledge must be guided by conscience.

Freedom must rest on responsibility.

The final defence has three layers: honesty, reverence for our own fallibility, and fidelity to principles higher than immediate desire.

Chapter 13

Honesty

Honesty is usually defined as telling other people the truth. The more difficult form is telling the truth to oneself.

Other people can challenge a public lie. A private rationalisation may remain protected for years. It speaks in our own voice and uses our own intelligence.

The First Honest Sentence

The turning point in a compromised decision is often a sentence the person does not want to say:

I am not evaluating this relationship impartially.

I want this opportunity so much that I am lowering the standard of evidence.

I am hiding information because disclosure would cost me.

I call this professional judgment, but part of it is desire.

The sentence does not solve the problem. It makes a solution possible.

Without it, every safeguard appears unnecessary. With it, the person can seek distance, advice, and accountability.

Facts, Judgments, and Wishes

A simple practice is to divide a difficult decision into three columns:

8.   What I know — verifiable facts.

9.   What I infer — interpretations that may be reasonable but uncertain.

10.         What I want — outcomes that create a personal stake.

The third column is not shameful. Wanting love, success, respect, money, or influence is part of being human. The danger is disguising the third column as the first.

Honesty also requires precise language. “There is no proof” does not mean “there is no risk.” “This person has helped me” does not mean “this person has no hidden motive.” “I have not been caught” does not mean “I have acted rightly.”

The Reversal Test

Ask how you would judge the same conduct if roles were reversed.

If a rival met a source in private, concealed an affair, and resisted oversight, would you accept his claim that the relationship was under control? If an opposing politician accepted years of help from a mysterious foreign organiser, would you dismiss questions as prejudice? If a company you disliked released a powerful system after suppressing internal warnings, would you call the decision efficient?

The reversal test exposes double standards. It does not guarantee the correct answer, but it removes one hiding place.

Confession Before Crisis

We often wait for certainty before disclosing a conflict. That is backwards. A conflict should be disclosed precisely because the person inside it cannot be certain of his own neutrality.

Early confession is painful but limited. Late exposure is catastrophic. The longer concealment continues, the more identities and decisions depend on it.

Honesty is therefore not passive sincerity. It is a habit of making uncomfortable information available before circumstances force it into the open.

Chapter 14

Reverence

The word reverence can sound religious, old-fashioned, or abstract. Here I use it in a practical sense: a disciplined respect for realities larger than our confidence.

The surgeon respects biology. The pilot respects gravity. The intelligence officer must respect deception. The leader must respect the corrupting possibilities of power.

Reverence begins when expertise stops claiming exemption from consequence.

The Checklist

Modern surgical teams use checklists not because they lack training, but because training does not prevent every omission. A checklist can confirm the patient’s identity, the planned procedure, allergies, antibiotics, equipment, and expected complications.

The ritual may appear elementary. That is precisely why experts are tempted to skip it. Yet the body does not care how distinguished the surgeon is. The wrong-site incision remains wrong.

The intelligence equivalent is equally plain:

·    disclose personal or financial relationships;

·    document contact;

·    require independent source validation;

·    separate collection from final evaluation where possible;

·    rotate responsibility when dependency becomes too strong;

·    investigate warning signs as a pattern, not as isolated inconveniences.

These practices express reverence for fallibility.

A Pre-Mortem

Before a consequential decision, imagine that it has failed badly. Ask: What caused the failure?

This “pre-mortem” gives people permission to identify dangers without first persuading the group that failure is likely. One participant may say the source manipulated the handler. Another may say the schedule silenced engineers. A third may say the data was selected to please leadership.

The exercise creates distance from optimism. It makes an unwelcome future briefly concrete.

Protecting Dissent

Leaders frequently ask for honesty while punishing its delivery. Reverence requires more than inviting opinions. It requires protecting the person who raises a credible problem.

An organisation can test itself with a simple question: what happened to the last person who delayed a project, challenged a senior figure, or disclosed a conflict?

Employees will remember the answer more clearly than any ethics statement.

Reverence is not paralysis. Decisions must be made under uncertainty. It is the refusal to confuse decisiveness with infallibility.

The mature leader can say, “I believe this course is right, and I may be wrong. Therefore I want the strongest criticism before we proceed.”

That is not weakness. It is authority made safer by humility.

Chapter 15

A Higher Principle

Honesty reveals the conflict. Reverence keeps us aware of risk. Neither guarantees that we will choose duty when duty is costly.

For that, a person needs a principle that does not change with convenience.

The Night of 26 September 1983

Shortly after midnight on 26 September 1983, a Soviet early-warning system indicated that the United States had launched a missile. Then it indicated additional launches.

Lieutenant Colonel Stanislav Petrov was the duty officer at the command centre monitoring the system. He judged the warning to be false. The number of reported missiles appeared inconsistent with the massive first strike he expected, and ground radar had not confirmed the satellite alert. He reported a system malfunction. Later investigation found that sunlight reflected from high-altitude clouds had confused the satellite system.

Popular accounts sometimes say Petrov single-handedly “saved the world.” That claim simplifies a complex command system and a counterfactual no one can prove. His decision nevertheless deserves attention. Under extraordinary uncertainty, he did not surrender judgment to a machine merely because the machine spoke with authority.

Petrov’s action combined expertise, scepticism, responsibility, and restraint. He was willing to bear the risk of saying that an alarming system was wrong.

The Line That Holds in Private

A higher principle is not a slogan displayed when an audience is present. It is a line that remains when secrecy makes violation easy.

For an intelligence officer, the line may be: I do not conceal a relationship that compromises my duty.

For a scientist: I do not hide a safety result because it delays publication or profit.

For a politician: I do not accept assistance whose true source I refuse to examine.

For an ordinary citizen: I do not spread a claim I have not verified merely because it injures the people I oppose.

The principle must be decided before temptation tailors an exception.

Character Is Trained

Character is often described as something a person either has or lacks. That view is too static.

Character grows through repeated choices. Every small act of concealment makes the next concealment easier. Every small act of truth-telling makes the next truth more available.

This does not mean a person of character has no weakness. He may feel desire, fear, resentment, loneliness, and ambition as strongly as anyone else. Character is not the absence of those voices. It is the trained capacity to refuse their command when they conflict with duty.

Human nature produces desire. Character decides whether desire governs.

Human nature seeks shortcuts. Character decides which shortcuts are forbidden.

Human nature manufactures excuses. Character decides whether to believe them.

Institutions are external guardrails. Character is the guardrail within.

We need both. An internal guardrail without accountability can become self-righteousness. An external guardrail without internal commitment can become a game of finding loopholes.

The most resilient society is one in which law, culture, and character reinforce one another.

Conclusion

If It Were Me, What Would I Choose?

We have reached the question with which this book began.

If I had been in their place, would I have done better?

I do not know.

That answer is not despair. It is the beginning of vigilance.

If I were certain of my moral superiority, I could explain every failure as the defect of someone unlike me. The experienced agent was corrupt. The politician was vain. The official was weak. The citizens of old civilisations were foolish.

Then I would close the book unchanged.

Uncertainty forces a different response. It asks me to build defences before I know which temptation will fit me best.

I can disclose conflicts before they become scandals. I can invite review before certainty hardens. I can slow down when anger or flattery demands speed. I can separate fact from inference and inference from desire. I can decide which principles are not for sale.

None of these guarantees that I will never fail.

The goal is not perfection. It is wakefulness.

Civilisation will continue to produce stronger tools. Artificial intelligence will expand what individuals and institutions can know, create, and control. It will not decide why power is used. It will not supply courage to the person who must report an unwelcome fact. It will not confess the conflict we refuse to disclose.

That work remains ours.

Some people choose honesty when no one is watching.

Some accept responsibility when blame could be transferred.

Some preserve a boundary even when crossing it would bring pleasure, profit, or power.

The future depends on enough such choices becoming habits, and enough such habits becoming culture.

So I leave you with three questions:

11.         What do I want so badly that I may be tempted to misread the evidence?

12.         Who is allowed to tell me that my judgment is compromised?

13.         Which principle will still govern me when violation is invisible?

There are no standard answers.

But there is a final warning:

What defeats a person is rarely only the strength of an external enemy. It is often the part of human nature he is least willing to face.

Knowing that we have weaknesses is not enough. The decisive question is which voice—desire or character—we have spent a lifetime training to answer when both speak at once.

Selected Sources and Further Reading

The Katrina Leung Case

• U.S. Department of Justice, Office of the Inspector General. A Review of the FBI’s Handling and Oversight of FBI Asset Katrina Leung (Unclassified Executive Summary), May 2006. https://oig.justice.gov/sites/default/files/archive/special/s0605/index.htm

• U.S. Department of Justice, Office of the Inspector General. A Review of the FBI’s Progress in Responding to the Recommendations in the Office of the Inspector General Report on the Handling and Oversight of FBI Asset Katrina Leung, September 2013. https://oig.justice.gov/archives/reports/2013/s1310.pdf

The Christine Fang Case

• Bethany Allen-Ebrahimian and Zach Dorfman. “Exclusive: Suspected Chinese Spy Targeted California Politicians.” Axios, 8 December 2020. https://www.axios.com/2020/12/08/china-spy-california-politicians

• U.S. House of Representatives, Office of the Clerk. “Office of the Fourteenth Congressional District of California,” recording Eric Swalwell’s resignation on 14 April 2026. https://clerk.house.gov/members/CA14/vacancy

• Office of House Democratic Leader Hakeem Jeffries. “House Democratic Leadership Statement on Accusations Against Rep. Eric Swalwell,” 10 April 2026. https://jeffries.house.gov/2026/04/10/house-democratic-leadership-statement-on-accusations-against-rep-eric-swalwell/

Psychology, Organisations, and Decision-Making

• Justin Kruger and David Dunning. “Unskilled and Unaware of It: How Difficulties in Recognizing One’s Own Incompetence Lead to Inflated Self-Assessments.” Journal of Personality and Social Psychology 77, no. 6 (1999): 1121–1134.

• Emily Pronin, Daniel Y. Lin, and Lee Ross. “The Bias Blind Spot: Perceptions of Bias in Self Versus Others.” Personality and Social Psychology Bulletin 28, no. 3 (2002): 369–381.

• Leon Festinger. A Theory of Cognitive Dissonance. Stanford University Press, 1957.

• Ziva Kunda. “The Case for Motivated Reasoning.” Psychological Bulletin 108, no. 3 (1990): 480–498.

• Columbia Accident Investigation Board. Report, Volume I, August 2003. https://www.nasa.gov/columbia/caib/

• Diane Vaughan. The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press, 1996.

• Atul Gawande. The Checklist Manifesto: How to Get Things Right. Metropolitan Books, 2009.

History and Technology

• Thucydides. History of the Peloponnesian War.

• Mary Beard. SPQR: A History of Ancient Rome. Liveright, 2015.

• Timothy Brook. The Troubled Empire: China in the Yuan and Ming Dynasties. Harvard University Press, 2010.

• Ray Huang. 1587, A Year of No Significance: The Ming Dynasty in Decline. Yale University Press, 1981.

• National Security Archive. “The Man Who Saved the World: Stanislav Petrov and the Soviet False Alarm of 26 September 1983.” https://nsarchive.gwu.edu/

About the Author

Fang Tianliang (FTL) writes at the intersection of traditional wisdom, history, technology, and modern civilisation. His Life Museum series invites readers to examine contemporary problems through longer moral and historical perspectives.

Seeking Traditional Wisdom · Reflecting on Modern Civilization

END