Soft Determinism

In the AI Story

The intellectual genealogy of soft determinism runs through Thomas Hughes's concept of technological momentum—the observation that technologies acquire social, economic, and institutional investments that make them progressively harder to redirect, without ever becoming fully autonomous. Hughes distinguished this momentum from determinism: the system has direction and mass, but it can be redirected by sufficient institutional effort applied at the right leverage points. Smith adopted and refined this framework, emphasizing that the formative period of a technology—when momentum is still building—is the period when institutional choices carry disproportionate weight.

Soft determinism's practical value lies in its capacity to hold contradictory truths simultaneously: that AI capabilities constrain what is possible (no institutional arrangement can make large language models incapable of generating competent text) and that institutional arrangements determine deployment terms, distribution of benefits, and protections for affected populations. The natural language interface that emerged in 2025 was a genuine threshold—a capability expansion that foreclosed the pre-AI world. But whether that capability produces democratization or concentration, human development or cognitive atrophy, depends entirely on institutions.

Critics charge that soft determinism is conceptually unstable—that it tries to have things both ways by acknowledging constraints while insisting on choice. The charge misses the framework's empirical grounding. The historical record documents that identical technologies produce different outcomes in different institutional contexts (the comparative evidence), and that early institutional choices create path dependencies making later alternatives costly (the temporal evidence). Soft determinism is not philosophical compromise but empirical finding—the position that the documented evidence supports.

Origin

The position emerged from Smith's archival confrontation with the divergence between Springfield and Harpers Ferry. The puzzle demanded explanation: if the technology determined the outcome, why did identical technologies produce different results? If human choice alone determined the outcome, why were both armories eventually compelled to adopt precision manufacturing? The answer—that technology constrained while institutions determined—resolved the empirical puzzle and became the framework for analyzing every subsequent technological transition Smith studied.

Key Ideas

Technology constrains, institutions determine. The technology limits the range of possible futures; institutional arrangements determine which specific future, within that range, materializes—neither sovereignty nor surrender.

The formative period carries disproportionate weight. Early institutional choices, made when alternatives remain accessible, establish trajectories that later investments make prohibitively costly to alter—making the present moment decisively consequential.

Comparative evidence reveals contingency. The divergence between Springfield and Harpers Ferry, American and European telegraph systems, regulated and unregulated factories demonstrates that outcomes are not determined by technology alone.

The position enables agency. Hard determinism produces passivity (outcomes are inevitable); soft determinism produces engagement (outcomes depend on institutional quality, which human effort can improve).

Constraints are real but not total. Acknowledging that AI forecloses certain futures (the pre-AI world cannot be restored) does not require accepting that the specific post-AI future is predetermined—the range within constraints remains genuinely open.