Rescuing Democracy from the Quiet Rule of AI (www.noemamag.com)

Andrew Sorota warns that a quieter but profound threat from AI is already reshaping democracy: systems that do more than assist us—algorithms that decide—can erode accountability, civic recognition and the ability of citizens to contest outcomes. Drawing from Asimov’s “The Evitable Conflict” and real-world cases like the Dutch welfare fraud scandal, he shows how machine recommendations quickly become defaults, human intermediaries turn into impotent clerks, and opaque models (hidden layers, probabilistic outputs) produce decisions that are hard to explain or appeal. That dynamic risks deepening the democratic deficit, fueling resentment and populism even if systems are technically explainable. But Sorota argues the outcome is not preordained: AI can either substitute for collective judgment or enlarge it depending on the surrounding social contract. Practical, low-level AI capabilities—automatic translation, live transcription, summarization, clustering of public comments and LLM-driven plain-language briefings—can scale deliberation and broaden participation. He points to vTaiwan and a Deschutes County pilot as examples where ML surfaced consensus rather than imposed policy, and to Ireland’s citizens’ assemblies as a model that AI could scale. The policy implication is clear: governance must define where AI may decide, where humans must retain authority, and build legal, institutional and technical mechanisms (recourse, override, transparency plus actionable accountability) so algorithms empower rather than silence citizens.