Nicholas A. Caputo (Oxford Martin School) has posted Technology, Complexity, and Administration: AI and the Capability-Accountability Trap on SSRN. Here is the abstract:
This Article argues that AI offers a path out of the “capability-accountability trap” that has structured administrative law since 1887. Technological change forces government to become more sophisticated, but sophistication reduces accountability because it reduces scrutability, the cognitive comprehensibility of administration to courts, legislatures, and the public. Unable to evaluate the substance of agency actions, overseers rely on procedure, which accumulates over time. Recent Supreme Court decisions can be read as a scrutability intervention that restores comprehensibility by shrinking administrative capability, just as climate change, pandemics, and AI risks demand better governance.
AI offers a way to break the trap. Drawing on Herbert Simon’s cognitive model of administrative behavior, this Article shows how AI could serve as scrutability infrastructure that translates complexity into accessible terms, enables meaningful public participation, and surfaces the assumptions and tradeoffs in agency actions that matter for oversight. More fundamentally, advances in technical alignment and interpretability could restore the substantive accountability that procedural review was always a substitute for, reducing ossification rather than compounding it. The success of governmental transformation is not guaranteed, and AI alone is not a full solution. However, a new capability crisis is coming. The history of administrative law and technology shows that this crisis could be the occasion for building institutions that are both more capable and more scrutable than any that have come before.