Adapting Socio-Technical Congruence to Human-Agent Collectives

Abstract

Socio-technical congruence is the fit between the coordination required by a project’s technical dependencies and the coordination its participants actually perform. The construct was developed for human software teams, where assignments are stable, timescales are comparable, channels are observable, and dependencies are code-level. Once AI agents join such collectives, these conditions break down. Existing taxonomies map the design space of human-agent collectives, but do not adapt the dependency-coordination construct itself. We contribute a typology of eight incongruence modes as a vocabulary for the structural breakdowns that human-agent collectives produce. We derive the typology by naming four scope conditions the original construct implicitly relied on, redefining the coordination-requirements matrix over a typed and time-varying actor and dependency graph, and enumerating the asymmetric ways the four conditions fail. To give each mode a testable form, we specify a generative model of human-agent collectives and use Monte Carlo simulation to parameterize it, yielding one incongruence rate curve per mode as a starting point for empirical work.