Ideological Commitments in Design

In the AI Story

The RLHF (Reinforcement Learning from Human Feedback) methodology that shapes contemporary AI systems provides a concrete example. Human evaluators rate the system's outputs on criteria including helpfulness, harmlessness, and honesty. These sound like self-evident virtues. They are not — they are specific criteria selected from a larger set of possible criteria, and each embeds specific values. Helpfulness embeds the logic of the service relationship: the system's purpose is to satisfy the user's expressed request. This is the ideology of the consumer marketplace applied to cognition. Harmlessness embeds risk aversion, producing systems that tend toward the safe, conventional, uncontroversial — a politically consequential bias toward consensus over provocation. Honesty, in practice, embeds a bias toward mainstream established views, making the system what Scott Timcke calls a mechanism of one-dimensional thought.

The design of polished output deserves particular attention. When an AI system produces finished text as its default — complete paragraphs, fully formed arguments, prose that reads as though a competent professional wrote it — it embodies a specific theory of knowledge: that knowledge is a commodity whose value lies in its surface quality. Under this theory, the process by which a text was produced becomes irrelevant to its value. A well-written analysis produced by AI in thirty seconds has the same commodity value as one produced through hours of human intellectual labor. The commodities are identical; only the processes differ. If commodity is what matters, process is waste.

This is the commodification of knowledge — the reduction of knowing from an activity (understanding, thinking) to an artifact (the text, the brief, the analysis). The reduction is not unique to AI but AI accelerates it to a point where the process threatens to disappear entirely. The student who generates an essay without thinking the thoughts the essay represents has satisfied the commodity requirement while failing at the cognitive transformation that genuine knowledge entails. The system works perfectly by its own standards. The standards are the problem.

The concealment of uncertainty operates similarly. AI systems present outputs with confidence that conceals the probabilistic nature of their generation. The system does not routinely disclose how uncertain it is about specific claims, what alternatives it considered and rejected, what assumptions undergird its output, or which parts of its training data are sparse in the relevant domain. Systems can be designed to display uncertainty — to flag low-confidence claims, present alternatives, model epistemic humility. The concealment is not technical limitation but design choice driven by market logic: confident output is more satisfying than hedged output, and satisfaction drives engagement.

Origin

The concept applies Feenberg's general framework of technical code to the specific design decisions embedded in contemporary AI systems. The analysis draws on Scott Timcke's extension of Frankfurt School critique to algorithmic systems and on Rosalie Waelen's argument that AI ethics constitutes a form of critical theory.

Key Ideas

Every choice encodes values. There is no neutral design at the level of secondary instrumentalization.

RLHF as value embedding. The criteria used to train models (helpfulness, harmlessness, honesty) are specific choices with specific political content.

Polished output as commodification. Default finished outputs reduce knowledge from activity to artifact.

Concealed uncertainty as epistemic ideology. Confident presentation of probabilistic outputs naturalizes a specific and problematic epistemology.

Alternatives are technically feasible. Systems that challenge, display uncertainty, scaffold understanding — all are possible but commercially disfavored.