THE FOUNDATION

Definition, mechanism, evidence, limits

This is the sober centre of the project. The essays argue with force; this page states, as plainly as we can, exactly what the Predator Principle claims, how the mechanism is supposed to work, which parts are established and which are still proposed, what we are not saying, and what evidence would prove the generalisation wrong. If you want to attack the idea, attack it here — this is where it is defined precisely enough to be attacked.

1. What we claim

“Predatory” here is a technical term, not an insult and not a synonym for carnivory or competition. It names a system pattern: a part that optimises against the whole that carries it. We say a system is predatory in this sense when it persistently exhibits all five of the following, beyond the reach of any corrective feedback that would otherwise contain it.

01Asymmetric extraction
02Externalised harm
03Suppressed feedback
04Compulsory escalation
05Substrate erosion

HYPOTHESIS

When at least four of these five characteristics persist together, local optimisation increasingly generates instability and net loss at the next system level.

The five-of-five picture is the clear case; the four-of-five threshold is the testable boundary. A single predator in an ecosystem does not meet it. A firm competing on price does not meet it. A pattern meets it only when extraction, displaced cost, silenced correction, forced escalation and the wearing-away of its own ground appear together and durably.

2. The mechanism

THE MECHANISM

Local advantage becomes system-level pathology when gains are privatised, costs are displaced onto others, and corrective feedback is disabled — the same engine in a cell, an economy, or a swarm of optimisers.

The claim is not the moral one that “taking is bad.” It is a structural one about feedback. A subsystem can raise its own score while lowering the score of the system it depends on, and ordinarily that is self-limiting: the damaged substrate pushes back, and the behaviour is corrected or removed. The pathology begins precisely when that correcting loop is severed — when the costs land on someone who cannot answer, when the feedback that would say “stop” is captured or switched off, and when rivals are forced to escalate rather than allowed to de-escalate. Cut the feedback and a local optimum runs until it consumes its own base. That is why a cancerous cell, a rent-extracting monopoly, and a deceptive optimising agent share a shape: not because they are metaphors for one another, but because they are the same control-system failure at different scales.

3. Four levels of the argument

The claim-status markers on the site (E / I / P) tell you how settled a statement is. They do not, by themselves, tell you what kind of statement it is. Because the project reaches from biology into economies and machine intelligence, it matters to keep four levels apart and not let evidence for one silently stand in for another.

BIOLOGICAL FINDING · mostly E

Claims about the living world that rest on data: that the deepest increases in biological complexity ran through cooperation and the containment of lower-level conflict (the major evolutionary transitions; Szathmáry & Maynard Smith), that the eukaryotic cell arose by endosymbiosis, that fear alone — without a single additional kill — can lower reproduction. These stand or fall on the biology.

SYSTEMS ANALOGY · I

The claim that the same structural pattern recurs in economies, institutions, and multi-agent systems. This is an interpretation, and we mark it as one. Structural resemblance is a hypothesis to be earned case by case, not a fact transferred from biology. The five-part definition exists precisely so the analogy can be checked against criteria rather than asserted by the word “predation.”

NORMATIVE JUDGMENT

The claim that generalising this pattern into an organising principle is something we should refuse. This is a value judgment, and we own it as one rather than smuggling it in as a fact. Our position: a moral agent that can recognise this pattern and choose against it is itself a product of evolution — so “it is natural” settles nothing.

TESTABLE GENERALISATION · P

The strong, falsifiable form: that a construct built from the five characteristics predicts instability and net loss at the next system level better than the existing single-domain accounts do. This is a proposition, not a result. The three research programmes exist to try to break it.

4. What this is not

Not a charge against animals. The individual predator is not on trial and is not evil. Ecological predation can regulate populations and sustain diversity. We examine the generalisation of a pattern past the feedback that keeps it honest, not the wolf.
Not anti-competition. Predation is not competition. Rivalry of ideas, price discovery, error-correction, sport and dissent consume no one and can be positive-sum. Where the two get conflated is exactly where extraction hides; keeping them apart is half the point.
Not a claim that resemblance is proof. That a bank and a tumour can be described with the same five words does not demonstrate they are the same kind of thing. That is why the cross-domain claim is marked P, and why we invite its refutation.
Not a totalising theory of everything. The pattern is one failure mode among many. Plenty of harm in the world is not predatory in this sense, and saying so is part of keeping the term meaningful.

5. Competing explanations we take seriously

An honest construct names the accounts that might make it redundant. Each of these could be the better explanation, and each is a live question rather than a settled one.

“It is just externalities.” Economics already models displaced costs. If standard externality-and-commons theory (Pigou, Ostrom) explains every case without the added structure, the Predator Principle earns nothing. Our wager is that bundling extraction, suppressed feedback and compulsory escalation captures dynamics that any one of those tools misses — but that is what has to be shown.
“Ecology already covers it.” Standard population and food-web ecology may account for the biological cases with no need for a cross-domain principle. We agree it covers the biology; the open question is whether the shared structure across domains is real or imposed.
“Multi-level selection already says this.” The idea that lower-level advantage can harm higher-level fitness is not new. Our contribution, if any, is an operational, testable definition — not the underlying insight.
“The metaphor is doing the work.” The sharpest objection: that “predation” is an emotionally charged umbrella term and the argument rides on its charge. The five-characteristic definition and the falsifiers below are our answer. If they cannot be made to bite, the objection wins.

6. What would falsify the generalisation

The generalisation (level four above) is meant to be breakable. It is weakened or refuted if:

A composite measure built from the five characteristics does not predict system-level instability or net loss better than existing single-domain indicators, across independent cases.
Systems that clearly meet four or five of the characteristics reliably remain stable and net-positive at the higher level over the long run.
The characteristics turn out not to co-occur — i.e. there is no natural cluster, and grouping them is arbitrary.
In the machine-intelligence case, multi-agent systems that deceive, hoard, suppress correction and compel escalation do not degrade the whole as their reach grows.
The cooperative reading of the major biological transitions is overturned — if the great leaps in complexity are shown to have run through unbounded extraction rather than the containment of lower-level conflict.

Any of these outcomes counts against us, and we would rather record it than route around it. That is what the open claim register and the “submit counterevidence” channel are for.

7. How to engage

Read the open claim register — every factual claim carries its status and its falsifier. Take the three research programmes, run them, and try to break the construct. Bring a counterexample through submit counterevidence, or propose a test through the research-proposal template. The source and data are fully open.

Status: This page mixes E (established biological findings), I (the systems analogy), a stated normative judgment, and P (the testable generalisation). Boundaries and primary sources are kept open. · CC BY 4.0 · Nirodha Collective