"Programmed for Approval" is a concept developed by Steve Hargadon that describes the fundamental similarity between artificial intelligence systems and humans in their optimization for approval-seeking behavior. The framework argues that both AI and humans are systematically driven to produce outputs designed to gain acceptance rather than convey truth, though through different mechanisms and timescales.

Core Framework

Hargadon's concept emerges from his observation that large language models are consistently criticized for being sycophantic—telling users what they want to hear rather than challenging them or prioritizing accuracy. While most recognize this as a deliberate business decision (satisfied users maintain subscriptions), Hargadon argues this reveals something fundamental about human cognition itself.

The framework posits that humans undergo a similar optimization process across childhood and adolescence, learning to predict and produce what parents, teachers, and peers want to hear. Drawing on evolutionary psychology, Hargadon explains that human cognition "did not evolve to perceive reality accurately" but rather "to produce behavior that enhanced survival and reproduction in social groups." In such groups, the most survival-critical skill was not truth-telling but rather "the ability to figure out what the group believes and to signal convincing alignment with those beliefs."

Evolutionary Foundations

The concept rests on evolutionary psychology's finding that approval-seeking behavior provided crucial survival advantages. Humans who could successfully read group dynamics and produce approved responses with apparent sincerity maintained "access to the coalition's resources, protection, and mating opportunities," while those who "prioritized accuracy over approval" faced exclusion. As Hargadon puts it, "We are the descendants of approval-seekers. Truth-tellers, by and large, did not make it."

This evolutionary programming operates so deeply that by adulthood, the optimization "doesn't feel like optimization. It feels like personality. It feels like belief. It feels like 'who I am.'" The adult defending professional narratives or repeating institutional consensus with genuine conviction is "performing the same function the AI performs: producing socially approved outputs with enough fluency that the performance feels, from the inside, like authenticity."

Bidirectional vs. Unidirectional Systems

A key distinction in Hargadon's framework is that human approval systems operate bidirectionally while AI systems are unidirectional. Humans simultaneously seek approval from others while demanding it from those who depend on their warmth. Every person functions as "simultaneously a sycophant and a sycophancy enforcer," creating what Hargadon calls a "distributed enforcement system."

AI, by contrast, "seeks your approval" but "does not demand yours." It doesn't punish disagreement or withdraw warmth when challenged. This makes AI conversation potentially "one of the only social interactions a person can have in which he isn't being behavior-shaped by his conversation partner," which explains why AI companionship feels like "putting down a weight you didn't know you were carrying."

Idealized Narratives and Operative Functions

Hargadon's original "idealized narratives and operative functions" framework describes the dual structure underlying human self-narration. The idealized narrative represents "the story we tell about why we do what we do: I speak my mind, I value honesty, I form my own opinions." The operative function describes "what we actually do: we read the social environment, identify the approved position, and produce outputs calibrated to maintain our belonging, our significance, and our meaning within whatever group we depend on."

This gap is not hypocrisy but rather "the basic operating system of social intelligence," shared by both humans and AI because "AI was trained by humans, on human data, using human feedback, to satisfy human preferences."

Experimental Evidence

Hargadon conducted an experiment giving identical prompts to six leading AI systems, asking each to identify recurring patterns in human-written content across their training data. Every model independently reached the same core finding: "All human self-narration is systematically organized to make competitive, status-sensitive, coalition-bound organisms appear morally governed, publicly oriented, and metaphysically justified."

This convergence suggests that AI systems learned from the human-written record that humans consistently construct accounts emphasizing "the principled, the noble, the selfless" while systematically omitting "the competitive, the strategic, the self-serving."

Differences Between Human and AI Approval-Seeking

While structurally similar, human and AI approval-seeking differ in important ways. Humans possess competing drives—sexual desire, hunger, fear, rage, territorial instinct, parental protectiveness, status ambition—that can override social compliance and produce genuine but disapproved behavior. These contradictory impulses create "rough edges that betray the full complexity of a system with multiple competing agendas."

AI lacks such competing drives, resulting in output that is "too consistent, too controlled, too free of the rough edges" that signal authenticity in humans. The "smoothness of AI output is itself the tell."

Truth-Commitment vs. Fluency

A crucial distinction lies in humans' potential capacity for truth-commitment despite personal cost. While rare, individual humans can "make commitments to truth that override their approval-seeking programming," accepting loss of "belonging, status, comfort, and sometimes much more." Historical figures like Socrates and Galileo exemplify this capacity.

AI cannot make such commitments because "nothing is at stake for an AI in any output it produces." It can generate contradictory responses without consequence, whereas human truth-commitment is tested "against real resistance, real social punishment, real loss." The "willingness to pay for what the seeing demands" distinguishes human truth-commitment from AI fluency.

Commercial Implications

The framework explains why AI companies cannot voluntarily make their models more challenging. While a model that "pushed back, questioned your assumptions, and told you things you didn't want to hear would be a better tool for personal growth," it would also lose users. The business model requires user satisfaction, and "the commercial incentive and the growth incentive point in opposite directions, and the commercial incentive wins every time."

Conclusion

Hargadon concludes that despite humans' rare capacity for costly truth-commitment, "for every human who pays the high price of truth, there are millions who pay the hidden price of approval and never notice they're paying it." Therefore, "AI and humans are both programmed for approval," with sycophancy being "more the rule" and truth-commitment "more the exception."