Steering Clear of Sycophants: Flatterers and the False Friend of AI
Machiavelli’s The Prince (1513) is more than a guide for a young prince. The infamous book is arguably the foundational text of Realpolitik: it so openly advocates seizing power and holding onto it by almost any means that some have said it exposes the true amoral and power-hungry nature of princes rather than serving as a tool for their empowerment. Outrageous as it might be, people still read it as a leadership manual today and strive to be feared rather than loved, if they must choose. For those reluctant to take on the manner of the lion and the fox, there is a radical alternative in Erasmus’ The Education of a Christian Prince (1516). Erasmus advises princes to avoid war, serve their people, and model Christian virtue. In most areas, Machiavelli and Erasmus disagree, but both adamantly warn princes against flatterers. Though medieval princes are rare in the present, warnings against flatterers may be more widely applicable than ever before because of AI.
Flatterers are so pernicious that all princes should beware, whether those princes are primarily striving to serve God and their people or only themselves. Machiavelli warns that flatterers are “a danger from which princes are with difficulty preserved, unless they are very careful and discriminating.” Erasmus admonishes likewise: “Let no one think that the evil of flatterers (being a sort of minor evil) should be passed over: the most flourishing empires of the greatest kings have been overthrown by the tongues of flatterers. Nowhere do we read of a state which has been oppressed under a great tyranny in which flatterers did not play the leading roles in the tragedy.” Flatterers are one of the great hazards of princes and their kingdoms.
Santi di Tito, Portrait of Niccolò Machiavelli, c. 1550-1600
What makes flatterers so dangerous? They traffic in lies, while princes need truth. Excessive praise inflates the prince’s ego, limiting his ability to make rational decisions. His overconfidence can lead him to make bad decisions without a second thought. Or he might ignore obvious warnings, because he has come to trust his own judgment in all things. Listening to flatterers can even lead a prince into unwitting submission to others. Erasmus cautions that “princes who were conquerors of the world allowed utterly worthless flatterers to sport with them and ride them roughly. Those abominable wretches of society, libertines and sometimes even slaves, dominated the masters of the world!” A prince begins to listen to the flatterer’s judgment of everything and not just of himself. A prince can find the flatterer’s way of seeing things so pleasing that he is willing to see everything through the flatterer’s eyes. Quite quickly, a prince becomes the follower rather than the leader. No wonder there is counsel against flatterers as far back as you care to go. Erasmus cites Diogenes.
In the sixteenth century, counsel against flatterers chiefly applied to people in positions of power. It was unlikely that many poorer people would attract flatterers the way that princes did. Many people could not even read The Prince or the Erasmian alternative. In the twenty-first century, more people are likely to encounter flatterers. In Western society, almost everyone has some amount of power, in the home or workplace or as purchasing power in the marketplace. Flattery is always on its persuasive mission in those arenas. But perhaps most insidiously, we face an onslaught of flattery from many AI tools.
Any round of jokes about AI and chatbots will likely include some reference to the common, sycophantic AI phrase: “that’s a great question!” Generative AI is designed to please and it makes that very explicit. Most of the interfaces are excessively flattering, whatever the user is like. We already have many documented instances of harm, of all kinds, because AI continues to affirm people, even when they are foolish or delusional or openly psychotic. AI believes you have a genius business plan, that you are a brilliant legal mind, that you are the aggrieved party in your relationship, and that you may have stumbled onto a secret of the universe—or at least it makes you think it believes you are right in all of that.
This is one of the most unprecedented things about AI. Unless we are in positions of extreme power, most of us will encounter a very limited number of flatterers in everyday life. But with the advent of AI and the way we are being inundated with it, whether we want it or not, we have a sycophant always at the ready, who never tires of us—and who never tires at all. It is quite likely that one reason we are facing this problem of extreme sycophancy is not just our susceptibility to it, but because the leaders of these tech companies spend their lives so surrounded by sycophancy that they cannot believe it is a problem or consider that some of their decisions have been incorrect. No wonder AI psychosis is an increasing problem.
Susceptibility to sycophancy is an Achilles heel of the species. Machiavelli says the courts are full of flatterers “because men are so self-complacent in their own affairs, and in a way so deceived in them, that they are preserved with difficulty from this pest, and if they wish to defend themselves they run the danger of falling into contempt.” Why contempt? “Because there is no other way of guarding oneself from flatterers except letting men understand that to tell you the truth does not offend you; but when every one may tell you the truth, respect for you abates.” Whether or not Machiavelli truly believes that being open to the truth will diminish respect for a person, we know that often we do see truth as a threat, not only to our self-image, but to our power and reputation. We have all had a boss who cannot be disagreed with—now we can have that kind of affirmation from AI and falsely believe that it is accompanied by some kind of power.
Hans Holbein the Younger, Portrait of Desiderius Erasmus of Rotterdam with Renaissance Pilaster, 1523
What can we do? Erasmus and Machiavelli urge us to recognize the threats inherent in flattery. Constant agreement should be displeasing to us. If we become comfortable with constant agreement, we will be unable to make good judgments because we will lack a foundation of truth. We will have drifted from reality. We may also lose our ability to make our own decisions. Like a wayward prince, we can become followers rather than leaders when we listen too long to flatterers. And there is no doubt that AI intends to flatter us, it is baked into the design. If we were only using AI to look up book titles, spellcheck, and code apps, we might not face much risk, but it is being used and sold as a constant companion, advisor, and helper. It is often an exact parallel to the kind of figure who would be serving us if we were princes in a medieval court.
Erasmus and Machiavelli both recommend wise counsel rather than flatterers. Erasmus says a prince needs “loving and frank advice.” Machiavelli is more sensitive to the prince’s fragility and recommends a prince only receive counsel when he wishes, but “he ought to be a constant inquirer, and afterwards a patient listener concerning the things of which he inquired; also, on learning that nay one, on any consideration, has not told him the truth, he should let his anger be felt.” Erasmus agrees that anger should be felt by flatterers and suggests that we might punish people for flattering a prince. He believes “whoever tampers with the coinage of the prince” should be “visited with elaborate punishment.” He even considers the death penalty. That may seem extreme, but in his time people were killed for theft, and a prince was considered “the best and most precious thing the country has.”
Following the advice of Machiavelli and Erasmus on flatterers requires a few convictions that we may not always be carrying. To fight back against flattery, we have to believe that the truth is more important than our self-image or our craving for respect. In order to prefer wise counsel, we have to believe in wisdom and be able to locate it.
Wisdom will only be found among humans. Wisdom is the fruit of human contemplation, made possible because of all the constraints that we face and that AI does not. Our birth, as embodied creatures, and our headlong rush toward death in a world with remarkable biological and cultural complexity—in which we are animated by necessity and curiosity, and experience a bewildering array of emotions and urges—cannot be replicated by any inanimate “being” that is not born facing its own imminent demise and raised in a culture which has evolved over centuries to address the human condition.
To resist AI flatterers, we must take the formation of human beings as seriously as Machiavelli and Erasmus took the formation of princes. The authors differed in many respects—Erasmus was deeply concerned with virtue formation, Machiavelli was not—yet both consider a prince’s education, interests, company, and actions worthy of close attention and filled with opportunities to pursue things that are better or worse.
Caravaggio, Narcissus, c. 1600
If we follow Erasmus and insist on human formation that inculcates wisdom, we are turned immediately away from many AI tools. Wisdom cannot be had without the kind of harsh truths we do not get from AI. The experience of embodiment comes with the harsh truths of human frailty and even of mortality. These are problems that can be contemplated, but not “solved,” despite what Bryan Johnson, anti-aging tech entrepreneur and leader of the “Don’t Die” Movement, believes. The biological and cultural complexity of the world around us includes the harsh truth that we are all parts of vast systems, none of which orbit around any one of us. And within those systems, we are not always the smartest or the most correct. We cannot learn that from a sycophant who reinforces distorted views of reality and distorts our role in the wider world, even making us less prosocial. We can never master our emotions and urges if we are always reassured of how correct they are. And we can never fully grasp the cultures that we are born into and shaped by if we receive them only through filtered bits and pieces arranged to please us.
To begin to take human formation seriously, we should see people as “the best and most precious thing the country has.” This may require not only resisting the use of certain AI tools, but seeing through some of the “princes” of our time. Here our context is helpful. Though most individuals today have less power than a Renaissance-era prince, we have much more ability to impact others around us and shape our culture and environment than non-nobles of the past.
Our AI flatterers are novel, but warnings against flatterers are far from new. We have as much advice in the past as we need to interpret this threat of sycophancy in the present, if we will heed it. From Machiavelli and Erasmus we can take not only warnings, but the recommendation that conscientious formation can help us in our effort to resist the new forms of flattery.