Vision
From History to Possibility.
Investigating whether observable work data can support meaningful, calibrated predictions about how humans, AI agents and hybrid teams perform.
The question
Organizations record an enormous amount about how work happens: tasks, dependencies, reviews, approvals, incidents and delivery dates. Almost all of it is used to explain what already happened.
PsychoHAistory asks a narrower and more testable question: can observable historical work data support meaningful, calibrated predictions about how a particular team — of people, AI agents, or both — will perform on a particular piece of work?
Why now
AI agents are joining teams as participants that plan, execute, verify and hand off work. Adding them changes coordination costs, review load and failure modes in ways that intuition does not reliably anticipate. More agents may not mean faster delivery.
Hybrid workforces make the cost of a wrong team configuration visible and expensive. That makes prediction worth studying — and makes honest measurement of prediction quality essential.
The name
The name nods to the fictional “psychohistory” of Isaac Asimov’s Foundation novels, a science that predicts the behaviour of very large populations. PsychoHAistory borrows the spirit and none of the claims. The integrated HA marks the meeting of Human and AI work, the hinge between observation and prediction.
Our scope is deliberately small and practical: projects, workflows, agents and teams — not the fate of civilizations.
What we will not do
We will not profile personalities, infer psychology, or rank individual employees. The variables we study are explicit and work-related: capacity, availability, approvals, dependencies, verification and coordination.
We will not present demonstration numbers as findings. Everything on this site that looks like a measurement is labelled DEMO / SYNTHETIC DATA until real, validated evidence exists.
Principles
Falsifiable
Every prediction states what would prove it wrong.
Calibrated
A 70% forecast should come true about 70% of the time — and we will publish whether it does.
Explainable
Every output traces back to named assumptions and data windows.
Consented
No hidden data collection; integrations require separate review.