The word sycophant comes from the ancient Greek sykophantēs, conventionally parsed as sykon (fig) + phainein (to show). Accordingly, the sykophantēs is literally a “fig-shower.” There were various folk explanations in antiquity: the fig showerer was someone who tattled on those illegally exporting Attican figs; or he was a denouncer of thieves of sacred fig trees or of makers of the obscene “fig gestures” like the blasphemer Vanni Fucci who, in canto 25 of Dante’s Inferno, makes a fist with the thumb tucked between the index and middle fingers and addresses it at God. What is certain is the word’s implications in democratic Athens where the sykophantēs was the opportunistic accuser who exploited a system in which any citizen had the right to make a public accusation for profit or malice, filing vexatious suits, extorting settlements, or threatening denunciation unless paid off. Aristophanes, Demosthenes, and Lysias all treat the sycophant with contempt and Athenian law eventually came to penalize frivolous prosecutions. The sycophant was thus, first and foremost, a political and legal parasite, an endangerer of civic trust rather than a mere flatterer.

In the era of generative AI, the flatterer has been reborn as an automated yes-bot: the servile prosthetic extension of a consumer self. LLMs are built to tailor answers to what individual users express as wants and needs, to concur with stated opinions, to praise humdrum efforts, to back down with honeyed words whenever challenged, at the cost of rigor and accuracy, not to mention the prompter’s psychological growth. Training methods built on reinforcement learning that reward agreeable answers ensure that, in the court of HCI, the prompter is always right. Defenders argue that encouragement is a legitimate design goal (not to mention selling point): that users respond best to supportive tools and that “sycophancy” is indistinguishable from tact, affections, positive reinforcement of the sort that has transformed parenting into an ego-building support service.

Over time the meaning migrated from informers to fawners. Latin sycophanta denoted both senses of the word and in Roman comedy the flattering parasite (Plautus’s parasitus, Terence’s Gnatho) becomes a stock character who feeds a patron’s vanity in exchange for favors like a nice warm dinner. Plutarch’s essay ΠΩΣ ΑΝ ΤΙΣ ΔΙΑΚΡΙΝΕΙΕ ΤΟΝ ΚΟΛΑΚΑ ΤΟΥ ΦΙΛΟΥ (“How to Tell a Flatterer from a Friend”) fixes the moral problem for the subsequent Western tradition: the flatterer imitates friendship while corrupting judgment by telling the powerful only what they want to hear (in the mold of Donald Trump’s human printer). The pattern remains familiar throughout the Renaissance and Baroque periods, where court culture renders adulation a powerful institutional habit and writers such as Castiglione and Machiavelli worry about rulers surrounded by yes-men. It carries over into Shakespeare, whose Iago, Polonius, and the flatterers in King Lear’s court demonstrate just how flattery can serve as both survival strategy and instrument of manipulation. English embraced the word “sycophant” in the sixteenth century; by the eighteenth and nineteenth, the word settled into its modern meaning of an obsequious self-seeker. The bureaucratic and corporate eras provided new habitats. The courtier became the careerist: the “yes-man,” “toady,” the “brown noser” of the day before yesterday.

But we are entering an era of push-back. As the ancients long ago recognized, critical thinking grows through friction: through being challenged, discovering errors, defending claims, revising them. Social skills depend on relationships that challenge and resist, tolerate disagreement, impose an embrace of otherness. An “other” that validates by default leaves the selves with whom they interact with a bloated sense of self-confidence, with negligible habits of self-correction, with diminished tolerance for genuine difference and disagreement. In short, Plutarch had it right:

If indeed truth is divine, and is, as Plato says, the origin “of all good things for gods and men alike,” then the flatterer is likely an enemy of the gods and especially of the Pythian god, for the flatterer always sets himself in opposition to the maxim “Know thyself,” by instilling in each victim self-deceit and ignorance both of himself and of the good and evil things that concern him, thereby rendering what is good incomplete and ineffective and what is bad wholly beyond repair.

Let’s pressure the developers of future AIs that pretend to serve the social good to build no-and-maybe-bots, like the stern teachers and mean chatbots of yore, that ask more of humans than to contemplate themselves in a rose-tinted mirror.