Doctrine · The Operational Framework

Learnability Score

Paper PAR-010
Author Parthenius Air Intelligence Unit
Published 2026
Category Doctrine
Read time 5 minutes
Abstract
This paper defines the second variable in the Adaptive Deterrence Intelligence framework — the Learnability Score, a measure of how reliably an operation’s own behaviour can be inferred from its past. Learnability is a property of the operation’s own signature — how inferable its rhythm is. It is the mechanism connecting operational input to outcome, and it functions as a leading indicator rather than a lagging one: rising learnability warns of a maturing adversary model before any incident occurs. This paper establishes Learnability Score as the state variable and primary diagnostic measure of the discipline.
System Variation Rate
What is the operation intentionally changing?
Learnability Score
What can the adversary reliably learn?
Opportunity Denied Rate
What operational outcome is achieved?
Reduced Operational Risk

1. Position in the framework

PAR-009 defined System Variation Rate — the variable the operation controls. This paper defines the variable that variation acts upon: the Learnability Score, a measure of how predictable the operation's own signature has become.

Learnability Score is the mechanism that connects operational input to operational outcome. System Variation Rate influences it; it in turn influences the Opportunity Denied Rate. It occupies the centre of the causal chain — which is precisely why it is the most useful diagnostic. A change in learnability explains why an outcome moved, and points to whether the cause was operational or adversarial.

2. Definition

Learnability Score is a measure of how reliably an operation's future behaviour can be inferred from its past — the predictability of its own operational signature.

It is computed from the operation's own activity, not from assumptions about any specific adversary. This is what makes it measurable: like System Variation Rate, it is read from the operation itself. Where SVR measures the variation an operation introduces, Learnability Score measures the predictability that remains — how inferable the underlying rhythm still is despite that variation. The two are distinct. An operation can vary its surface while holding a deeper pattern constant, keeping SVR high yet learnability high with it.

A higher Learnability Score indicates the operational signature has become increasingly predictable — its future is inferable from its past. A lower score indicates that meaningful adaptation is disrupting any stable model of it.

How much learnability an operation can safely carry depends on the adversary it faces. A patient, sophisticated adversary can act on a lower score than an opportunist. The adversary does not define the metric — it sets the threshold at which a given score becomes dangerous.

This definition originates in The Learnability Problem (PAR-001), which established learnability as a property of security operations rather than of adversaries. PAR-010 operationalises that definition as a measurable score.

3. A leading indicator, not a lagging one

Conventional security metrics are lagging indicators. Thefts, alarms, arrests and losses all move only after damage has occurred. By the time they change, the opportunity has already been exploited.

If learnability starts creeping up, you do not wait for thefts to increase.
The rising score is the warning.

Learnability Score is a leading indicator. If learnability begins climbing, the operation does not have to wait for an incident to confirm that something is wrong. Rising learnability is itself the warning — the adversary is building a viable model. Operators can intervene immediately by changing operational behaviour, before the model matures into an executed attack.

Field observation · Active deployment · South Africa · Month 9
Probe frequency reached 52 confirmed approaches in a single month — each adding data to a model under construction. Learnability was assessed as approaching threshold well before any extraction was attempted. The response was a variance directive, issued on the leading indicator, not an incident report issued after a loss.

4. Reading the middle of the chain

Because Learnability Score sits between input and outcome, it separates two very different failure modes. When Opportunity Denied Rate falls, the operation must know why. Learnability Score answers the question.

Learnability rising while variation is low means operations have become routine. The fix is operational: increase meaningful variation. Learnability rising despite high variation means the adversary has adapted faster than expected; the fix is analytical — the variation is no longer meaningful to this adversary and must change in kind, not just degree. Learnability stable while outcomes fall indicates the cause is likely external, not a learning failure, and should be investigated elsewhere.

Without the middle variable, a declining outcome is just a declining outcome. With it, operators know where to intervene. This is why Learnability Score is the primary diagnostic measure of the discipline.

Conclusion

Learnability Score is the interpretive core of the framework. System Variation Rate is what the operation does; Opportunity Denied Rate is the result. Learnability Score is what connects them — the variable that tells operators not just that performance changed, but why, and what to do about it. The final paper in this framework defines the outcome it produces: the Opportunity Denied Rate.

Iterative research
Parthenius Air Research is published iteratively. Papers are revised as field evidence accumulates across deployments. Claims reflect current operational data, not final conclusions.
How to cite this paper
Parthenius Air, Learnability Score, Parthenius Air Research, PAR-010, 2026.
parthenius-air.com/learnability-score