Research clusters

Five lenses used to design DevPlane instrumentation. Each links to the canonical write-up on peopleanalyst.com. This site lists how product features map to those lenses; it does not replace the papers.

Clusters

Coordination

Stigmergic queues, handoffs, and collision mechanics in multi-agent + multi-tool work. Measures where information loss and false completion accumulate when nobody messages anyone directly.

https://peopleanalyst.com/research/coordination — canonical cluster page

Working alliance

Brief fidelity, repair after rupture, and whether the operational contract between operator and tooling holds across rewrites, rollbacks, and rework. Borrowed measurement language from psychotherapy process research; applied to dispatch and merge events.

https://peopleanalyst.com/research/working-alliance — canonical cluster page

Trust calibration

Whether approval behavior tracks ground truth: reviewer axes, minority objections, silent acceptance, and claim-level checks against repo state. Separates headline scores from the dissent the headline hid.

https://peopleanalyst.com/research/trust-calibration — canonical cluster page

Governance

Access, cost of legibility, refusal and halt semantics, and which failures the board renders visible vs. invisible. The interface arm asks what the joint system optimizes when automation improves.

https://peopleanalyst.com/research/governance — canonical cluster page

Identity

Provenance, attribution, and continuity of voice across human and model-generated artifacts. Minimum metadata so longitudinal panels know which signals are operator-written vs. model-written.

https://peopleanalyst.com/research/identity — canonical cluster page

Three-arm program

Coordination cost in heterogeneous AI tool ecosystems — pre-registered designs, open methods where production data permits, and DevPlane as measurement apparatus.

Full protocol draft: docs/research/PROGRAM.md in the devplane repo.

Origin: Agent-drafted index; operator may edit before treating as canonical.