Infrastructure Otherwise

In the AI Story

The project's central move is refusing the assumption that infrastructure must be extractive to function. Cowen's case studies demonstrate that infrastructures optimized for community sustenance can be technically sophisticated, geographically extensive, and logistically complex — they are not less capable than extractive systems, they are differently capable. The question is not whether an infrastructure of care is feasible but whether the political conditions for building it exist.

Applied to AI, the framework generates specific design proposals. Valves: session-length indicators, graduated response modulation, architectural mechanisms that modulate the tool's behavior in response to signals of human depletion. Buffers: institutionally required reflection periods between AI-assisted production and release, with different reliability standards assigned to first-draft and reviewed output. Audits: distributional analysis of who captures productivity gains and who absorbs the costs, extended through the entire cognitive supply chain. Voice: independent research institutions, community forums, mechanisms through which workers and their families can articulate the costs in contexts where the articulation leads to action.

These proposals slow the pipeline. Cowen treats this as a feature rather than a bug. Every sustainable logistical system in history has included mechanisms that reduce peak throughput in exchange for the sustainability of the human and ecological infrastructure on which the throughput depends. The resistance to such mechanisms — the argument that any reduction in response time is a competitive disadvantage — reveals the system's priorities more clearly than any critique could.

The project is explicit that redesign is not primarily technical but political. Every infrastructure encodes priorities. The current AI infrastructure's priorities are legible in its architecture. An alternative priority — sustainability, distribution, care — would produce a different architecture. The choice between them is not technological; it is a determination about whose needs the system will serve.

Origin

Cowen initiated the Infrastructure Otherwise research collective at the University of Toronto in 2019, collaborating with indigenous sovereignty scholars, urban planners, and community organizers. The project draws heavily on Lee Maracle's work on Indigenous water governance, the Red Nation's writing on Indigenous infrastructure, and the feminist infrastructure scholarship of Lauren Berlant and AbdouMaliq Simone.

Key Ideas

Infrastructure is never neutral. It encodes priorities, and the priorities are legible to anyone who knows how to read them.

Extractive design is a choice, not a necessity. Alternative logistical architectures already operate at scale in communities the dominant framework renders invisible.

Redesign is political, not technical. The tools and principles exist; what is missing is the organized will to rebuild.

Care is an engineering specification. It can be operationalized through valves, buffers, audits, and voice — not as sentiment but as architecture.

Debates & Critiques

The project has been criticized for romanticizing pre-industrial or indigenous infrastructural practices without engaging the scale questions that modern logistics must address. Cowen responds that the scale critique itself reflects the extractive framework — that many of the infrastructures she documents operate at substantial scale, and that the assumption that only extraction scales is an ideological artifact of the system the framework critiques.