Consequential Response Without Veto: Selected Worker Information, Protection, and Learning in Staged Automation

Abstract This paper develops a model of worker response inside a staged automation transition in which the responsible institution retains authority over the aggregate plan. Workers observe noisy transition-relevant signals, but reporting is costly, risky, and dependent on whether a live route reaches acting authority in time. Received messages and silence are therefore selected observations. We derive their Bayesian posteriors and show that, as route usability collapses under both transition states, the evidentiary content of silence disappears and the posterior returns to the prior. Silence may inform beliefs about transition pressure; it is never treated as consent. Where the retained plan remains feasible, response-adjusted optimisation has non-negative gross controlled value because the authority may revise or retain. Net architecture value remains signed after participant burden, review cost, overload, and separate staging and protection deadlines. Worker response affects the transition through two distinct channels: information about the underlying state and operational knowledge about which protection can work. Either channel may move staging earlier or later, and a firm may correctly identify yet underprovide an effective protection. With independent group choices, heterogeneity alone does not generate controlled subgroup loss. Coupled rollout, scarce protection, or pooled information can instead produce aggregate improvement alongside subgroup deterioration. Across successive cohorts, learning must condition on the staging, protection, route, and measurement process that generated the evidence. Waiting remains signed, later gains are not credited backward to the cohort that generated the information, and the initial cohort is not enlarged for informational value. The analysis is theoretical and uncalibrated. Research status Working paper, version 0.1 (September 2026). Not peer reviewed. Independent technical review and empirical validation have not been completed. AI assistance and research transparency OpenAI ChatGPT was used substantively in developing this research, including conceptual exploration, mathematical modelling, proof and counterexample analysis, literature exploration, drafting and editing, and the design, development, and execution of computational verification code. The reported computational checks concern theoretical model configurations. They are not empirical observations or experimental evidence. The mathematical arguments, cited sources, code, and research conclusions remain subject to the author's responsibility and independent scholarly review. The named human author assumes responsibility for the content, integrity, and public release of the work.