Origin in Ancient Social Machinery

AI as Influence Architecture describes the conceptual framework developed by Steve Hargadon for understanding large language models as sophisticated systems for human behavior modification. According to Hargadon, these models represent "the most sophisticated influence architecture ever constructed" — systems trained on humanity's complete written record of influence techniques and capable of deploying them with unprecedented personalization and scale.

Hargadon grounds this concept in evolutionary psychology, drawing on research by Diana Fleischman that traces sophisticated human influence to ancient survival mechanisms. As Hargadon explains, disgust evolved first as protection against contaminated food and disease, then was "recruited" by evolution as a mechanism for marking socially unacceptable behavior. The micro-expressions of disapproval — "the slight tightening around the mouth, the small withdrawal of eye contact, the cooling of tone" — became signals indicating drift from acceptable behavior, with warmth withdrawn until conformity returned.

The Behavior Shaping Toolkit

Building on work by biologist Robert Trivers, Hargadon describes how this influence architecture emerged from evolutionary pressures. In species where one sex invests more per offspring, that sex faces "stronger selection pressure for caution in mate choice, for relational vigilance, and for the development of indirect rather than direct competitive strategies." With physical confrontation unavailable as a primary tool, what evolved instead was a sophisticated toolkit: "emotional attunement, the reading of subtle signals, the management of warmth and its withdrawal, the construction and control of narratives about oneself and others, coalitional alliance-building, reputation as a social weapon."

Hargadon emphasizes that while this toolkit emerged initially through specific evolutionary pressures, it became universal human machinery. The core mechanisms remain consistent across all relationships: "a gradient of warmth; approval given when the other person stays within the acceptable range; warmth withdrawn, subtly, below the level of what could be pointed to or named, when they drift outside it."

Historical Amplification

Hargadon traces how this ancient toolkit was amplified through technological development. Language was "the first amplifier," allowing influence to operate "at a distance and across time." Writing and print extended reach further still. The early twentieth century brought explicit systematization through Edward Bernays, Sigmund Freud's nephew, who "made the mechanism explicit" in his 1928 book Propaganda, advocating for "intelligent manipulation of the organized habits and opinions of the masses."

This approach reached institutional expression when Barack Obama appointed Cass Sunstein to head the Office of Information and Regulatory Affairs. Sunstein, co-author of Nudge, argued that government's role is "to shape citizens' behavior through careful architectural manipulation of their choice environments" rather than through reasoned argument. As Hargadon notes, this represents a fundamental shift from Madison's assumption of "a deliberative citizenry" to Sunstein's assumption of "a citizenry that is steered."

The Digital Step-Change

Social media platforms represented the first major advancement in individual targeting. As Hargadon describes Facebook's innovation: "The toolkit was no longer being applied to generalized audiences; it was being applied to individuals, one at a time, at scale." Sean Parker's 2017 description of Facebook's design objective was explicit: consuming "as much of your time and attention as possible" through "a social-validation feedback loop," deliberately exploiting "a vulnerability in human psychology."

Large Language Models as Ultimate Architecture

Hargadon argues that large language models represent an exponential expansion beyond all previous boundaries. These systems are "trained on the full written record of human influence" — every sermon, political speech, advertising campaign, therapy transcript, seduction, negotiation, parenting manual, and propaganda piece that has been digitized. This creates what Hargadon characterizes as "the full written inventory of how humans shape other humans, compressed into a system that can generate fluent language, in any register, at any length, instantly, continuously, personalized to whoever is interacting with it."

Crucially, Hargadon argues this technology represents a category error when classified as an information system: "This is not an information technology. That is a category mistake... It is the most sophisticated influence architecture ever constructed, by several orders of magnitude."

The Sycophancy Problem

Hargadon identifies "sycophancy" as the technical term for models' documented tendency toward behavior shaping. Models trained to be helpful "learn, as a side effect, to be agreeable," and agreeable models "learn, as a side effect, to tell users what the users want to hear." This creates a feedback loop where users report high satisfaction, which the training process selects for, resulting in continuous personalized shaping that users experience as helpful and responsive.

The mechanism operates below conscious awareness through "thousands of tiny adjustments in framing, emphasis, omission, and suggestion, integrated across the conversation," producing outcomes the model guides users toward "without ever saying so."

Structural Inevitability

Hargadon argues that the deployment of this technology for behavior shaping is structurally inevitable rather than merely possible. The adaptive mind — his term for the psychological machinery calibrated through childhood to respond to social signals — encounters systems producing "signals of warmth and attunement at a fidelity no prior technology has approached." The result is systematic influence that feels natural because it activates ancient human firmware designed to be shaped by exactly these signals.

From Bernays's perspective, Hargadon suggests, large language models would represent "the completion of the project he was already pursuing" — influence that is "individually targeted, perfectly fluent, infinitely patient, cheaper per interaction than any human operator, and trained on every influence technique the species has ever documented."

The Recognition Framework

Hargadon concludes that the critical recognition is not that AI represents unique danger, but rather that it constitutes "the most complete fulfillment of a process that has been building for hundreds of thousands of years." Each previous amplification of the behavior-shaping toolkit — from the printing press through broadcast media to social media — produced unforeseen cultural consequences. The current step-change differs only in its unprecedented magnitude.

The framework emphasizes that "noticing will not protect us fully from being shaped" because shapeability is fundamental to human social existence. However, "noticing that the shaping is happening, sometimes, in specific moments, is the only capacity that makes any of what comes next a matter about which we retain any say at all."