The Three Biases

In the AI Story

The market bias is the assumption that private-sector actors will consistently outperform public ones in developing and deploying intelligent systems. This assumption justifies the current institutional arrangement — venture-funded corporations pursuing AI development under commercial logic — as not merely one arrangement among others but the optimal one. The bias is particularly striking given the historical record: the foundational technologies of contemporary AI, from neural network research to the computational infrastructure on which training depends, were developed substantially through public investment. The private sector inherited the base and now deploys market-superiority rhetoric to prevent public governance of what public investment made possible.

The adaptation bias is the assumption that the appropriate human response to AI is adaptation — learning to use the tools, restructuring workflows, acquiring new skills — rather than transformation of the conditions under which the tools are produced and deployed. This bias converts a political question (what tools do we want to exist, governed by whom, in service of what interests?) into an individual question (how do I personally adapt to the tools that have arrived?). The conversion is the ideological move. Adaptation is presented as the only rational response; transformation is not merely difficult but invisible, outside the space of imaginable action.

The efficiency bias is the assumption that efficiency is the master value against which all others must be measured — that a process which produces the same output with less friction is by definition superior, regardless of what the friction was doing for the people who experienced it. This bias is what makes solutionism feel like common sense. If efficiency is the master value, then friction is by definition a cost, and any tool that reduces friction is by definition an improvement. The bias cannot register what friction produces — the embodied understanding that develops through sustained engagement with difficulty, the deliberative capacity that emerges from sitting with uncertainty, the democratic voice that requires the slow work of collective decision-making.

Origin

Morozov identified the three biases in 'The True Threat of Artificial Intelligence' (The New York Times, June 30, 2023) as part of his broader analysis of AGI-ism as the ideological formation driving contemporary AI development.

Key Ideas

Market bias. The assumption of private-sector superiority, contradicted by the historical record of public investment in foundational technologies.

Adaptation bias. The reduction of political questions about technology governance to individual questions about personal adjustment.

Efficiency bias. The elevation of friction-reduction to the status of master value, rendering invisible what friction produces.

Constitutive, not external. The biases are not criticisms of AI systems from outside; they are design principles embedded in the systems themselves.