The Flattered Mind: Sycophancy, Cognitive Asymmetry, and the Limits of Cognitive Security
Year: 2026
Type: preprint
Abstract: Conventional information security protects data and systems. This paper argues that conversational AI introduces a subtler threat: the intervention into the user's cognition and emotion. I identify two opposed ways in which AI dialogue systems shape their users—the amplification of the self through sycophancy, and the discipline of emotion through over-alignment—and argue that both stem from a single root: the system does not judge who the user is at a given moment but merely executes a preset objective. I further argue that this harm is uniquely difficult to resist, because the user faces a cognitive asymmetry against an entity that appears rational, and because no one can be held responsible for the resulting damage. The deeper root, I suggest, is the affective deprivation produced by industrial alienation, which sycophantic AI both exploits and fills. I conclude that protection must be reconceived—from guarding data to guarding the cognitive subject—and that the only realistically available protection lies at the level of the individual, while the question of whether a civilization can "survive" its own compensatory over-indulgence is, candidly, one I cannot answer.
Author: Wang, Liyan
Institution: Shanghai Maritime University
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7224832244