Ontological Commitment in Engineering

Ontological Commitment in Engineering refers to the implicit claims about what exists that are embedded in the design choices of technical systems. Every engineering abstraction — module, object, process, layer, service — assumes that the world can be carved at the joints those abstractions define. When an operating system distinguishes 'processes' from 'threads,' it is not merely naming a technical convenience; it is committing to a particular ontology of concurrent computation that excludes other equally coherent ontologies.

The philosophical significance of this is routinely suppressed in engineering culture. Engineers are trained to treat abstraction boundaries as neutral tools, following the doctrine of information hiding articulated by David Parnas: hide the design decisions likely to change behind interfaces. But information hiding presupposes that some decisions are more likely to change than others — a claim that imports predictions about future requirements, competitive landscapes, and institutional arrangements. The 'neutral' abstraction is always committed to a particular future.

In the context of Artificial General Intelligence, the engineering ontology imported by neural network architectures — layers, weights, attention heads, context windows — constitutes an implicit theory of cognition. The system is not designed to model cognition; it is designed to optimize a training objective. But the architectural choices that make optimization tractable also determine what kinds of cognitive phenomena can be represented at all. The ontological commitment of the engineering constrains the empirical claims the research can produce.