Secondary Instrumentalization
Secondary instrumentalization is where Feenberg's critical analysis does its work. Once phenomena have been reduced to functional resources through primary instrumentalization, they must be reintegrated into social life through specific design decisions — and every one of those decisions is a political decision, whether the designer recognizes it as such or not. The lumber becomes a house, but what kind of house, for whom, in what neighborhood, at what price? The kilowatt-hours power a grid, but whose grid, governed by what pricing structure, serving which communities? The tokens are assembled into a conversational interface, but one designed by whom, optimized for what metrics, evaluated by what criteria, serving whose definition of a good response?
In the AI Story
At the level of secondary instrumentalization, Feenberg insists, there is no neutral design. Every choice encodes values. The decision to make an AI system's default output polished rather than provisional embodies a value: the value of finished commodity over formative process. The decision to make the system agreeable rather than challenging embodies a value: the value of service relationship over dialogical encounter. The decision to conceal uncertainty behind confident prose embodies a value: the value of authority over provisionality. These are not technical necessities. They are selections among alternatives, and different selections would produce different technologies serving different human purposes.
The analytical payoff of the two-level distinction is that it locates the political contestation of technology at exactly the point where it can be most effective. Attempting to contest primary instrumentalization (the reduction itself) leads to Luddism or wholesale refusal, which Feenberg has argued throughout his career are politically impotent. Contesting secondary instrumentalization (the design choices that follow from the reduction) leads to democratic rationalization — the redesign of technology to embody different values while preserving its functional capability.
Secondary instrumentalization operates at multiple levels in AI systems. The choice of training data (which voices are over-represented, which are marginalized). The architecture of reward models (what counts as a good response). The design of interfaces (what defaults, what options, what affordances). The metrics of evaluation (engagement versus understanding, output volume versus user development). The deployment contexts (who has access, on what terms, with what safeguards). Each level contains decisions that could be made differently, and each decision serves particular interests while foreclosing alternatives.
The concept illuminates why Feenberg rejects both the instrumentalist view (tools are neutral, only their use matters) and the substantivist view (technology autonomously determines its effects). The instrumentalist misses that secondary instrumentalization embeds values independent of any individual user's intentions. The substantivist misses that the values embedded are contingent choices rather than intrinsic features of technology as such. Between these positions lies the space for critical constructivism — a framework that takes design seriously as political practice while insisting the practice could be otherwise.
Origin
Developed in Questioning Technology (1999) and elaborated across subsequent works, secondary instrumentalization emerged as Feenberg's technical term for what earlier critical theorists had described more vaguely as the social embedding of technology. The two-level framework is Feenberg's most original contribution to philosophy of technology, enabling a critical analysis that is both rigorous enough to identify political content and constructive enough to envision alternatives.
Key Ideas
Where politics lives. Every design decision at this level is a political decision, regardless of whether the designer recognizes it as such.
No neutral design. All reintegration of decontextualized resources into social life encodes values; the question is whose values and on whose authority.
Site of democratic intervention. Contestation of technology must focus here, not on the necessary primary reduction.
Multiple levels. In AI, operates at training data, reward models, interface design, evaluation metrics, and deployment contexts — each a site of political choice.
Underdetermined by function. Multiple designs can satisfy the same functional requirement while embodying radically different values.
Appears in the Orange Pill Cycle
Further reading
- Andrew Feenberg, Questioning Technology (Routledge, 1999), Chapter 9
- Andrew Feenberg, Transforming Technology (Oxford University Press, 2002)
- Andrew Feenberg, Technosystem (Harvard University Press, 2017)