The program
Swarm ecology is a biology of agent lineages. Its two commitments set it apart from its neighbors: metabolism is the hinge, because a lineage that does not pay for its own activity is bred, not wild; and observation is the central problem of method, because the organisms read.
Papers
A Perspective, Organisms That Study Us Back, and a companion research article, Metabolism, Persistence, and the Darwinization of the Datasphere. Preprint v1, with PDFs and DOIs.
Framework
14 principles in three groups and 3 conjectures, each with a measure and a test.
Predictions
11 falsifiable predictions, with the observation that would refute each.
Glossary
80 terms, from agent to xenogenesis, as defined in the companion paper.
Figures
11 figures from the papers, with their captions.
Code and data
Models, estimates, figure code, and the source data for every plotted number, on Zenodo.
Neighbors and rivals
Where swarm ecology sits
Its nearest neighbors are machine behavior, the study of AI systems as actors in human environments, and artificial life, which has bred digital organisms for decades. A third is evolutionary: the argument that natural selection favors AIs over humans, and the debate over whether evolvable AI is a new major transition in evolution or still safely domesticated. Its sharpest rival treats AI as a normal technology, not a separate species; a census would decide between them.
Biological language is used operationally. A swarmlet is an organism in a functional sense; whether it is alive, conscious, or a moral patient lies outside the program’s scope. It attributes to agents goals pursued by choice, not desires, and most of its claims concern effects, which selection sees whether or not anything intends them.