Organisations invest in the right training. Governments design evidence-based policy. Clinicians follow best-practice protocols. And yet — the system doesn't change.
The assumption behind almost every intervention framework is that failure means the wrong tool was chosen. So we build better tools. More nuanced models. Richer inventories of what's missing.
But what if the problem isn't the tool? What if the problem is when and in what order the tool is applied — and that applying the right tool at the wrong moment in the wrong sequence actively makes things worse?
Two well-designed components that don't mesh. The problem is not the design. It is the configuration.
COM-B tells you what is missing — Capability, Opportunity, Motivation. The Behaviour Change Wheel tells you which intervention category addresses each gap. BJ Fogg's B=MAP tells you the conditions for behaviour to occur.
All three are inventories. They describe a system's components at a point in time. None ask: what direction is the system moving? Which constraint is binding right now? What happens if I add capacity before the system has an accurate model?
The missing layer is diagnosis before prescription — understanding the system's internal configuration, not just its resource levels.
The accuracy of the system's internal model of its situation. Not how much information it receives — but how well it understands what is happening and what comes next.
A person's accurate self-model of their behaviour.
An organisation's honest reading of its market position.
A policy system's grounded understanding of ground conditions.
The degrees of freedom available to act coherently on that model. Not raw resources — but whether the execution environment is open, directed, and aligned.
A person's real environmental opportunity to change.
An organisation's structural capacity to execute strategy.
A policy system's institutional delivery infrastructure.
Most frameworks treat capacity as additive. More capability plus more opportunity equals more behaviour. The assumption is compensatory — high motivation can substitute for low ability.
This is wrong. When I and A are misaligned, the system pays an exponential penalty on its effective drive — regardless of how high the absolute levels are.
A highly resourced organisation executing against a broken strategic model is not just suboptimal. It is running fast in the wrong direction.
These are not points on a continuum. They are qualitatively distinct regimes. A system in one state does not just perform differently — it fails differently, recovers differently, and responds to different interventions.
In the Mahabharata, the warrior Abhimanyu learned the secret of entering the Chakravyuha — an unbreakable battle formation — while still in his mother's womb. He heard his father Arjuna explain how to enter. But before Arjuna could describe how to exit, his mother fell asleep.
Abhimanyu enters the formation with extraordinary skill. He can see exactly what is happening. His understanding is complete. But he has no way out. He is destroyed not by ignorance but by structural incapacity.
The Abhimanyu Conundrum is this exact configuration. High context awareness, blocked response availability. The system sees the problem with perfect clarity — and has no mechanism to act on what it sees.
In the Ramayana, Hanuman was cursed as a child to forget his extraordinary abilities. He lived as an ordinary being, unaware of what he was capable of. Only when the sage Jambavan reminded him of his true nature did he remember — and then he leapt across the ocean in a single bound.
The problem was never capacity. Hanuman always had the power. The problem was the absence of an accurate model of what he was and what was possible.
The Hanuman Complex is the dangerous inversion. High response availability, degraded context awareness. The system has capacity — sometimes enormous capacity — but no accurate model to direct it. It acts blind.
Organisations and practitioners systematically prefer high-capacity, high-activity states — and are suspicious of the quiet, aware-but-blocked configuration.
But Abhimanyu preserves optionality. The model is intact. When capacity becomes available — through intervention, structural change, or resource provision — the system can act correctly immediately, because it already knows what to do.
Hanuman does not have this property. Restoring capacity to a system without a valid model just gives it a faster machine pointed in the wrong direction.
The central result of CARA is a sequencing law. Before selecting an intervention, you must know which of the two functions is the binding constraint — and address it first. Order is not a preference. It is structurally determined.
State classification gives you position. Trajectory gives you momentum. Six recurring dynamic patterns emerge from how I and A are moving — each with a distinct risk profile and intervention logic.
| Pattern | Signature | What it means | Intervention |
|---|---|---|---|
Bubble Precursor |
I stable, A falling from positive | System in false alignment. A quietly eroding while I holds. The warning arrives before the state changes. | Watch. Do not prop A artificially. |
Late Cycle |
I high, A just crossed zero | The internal break has happened. I still elevated. System at highest near-term risk. | Urgent. Force model update first. |
Trapped Frustration |
I high, A negative and worsening | Model intact, execution collapsing. Highest forward risk of all patterns. | Emergency A-restoration only. |
Freefall |
Both I and A collapsing | Active breakdown. No leading variable. Triage mode. | Minimum viable A to stop bleeding. Then I. |
Blind Acceleration |
I negative, A rising | Capacity growing without a valid model. Committed errors accumulating with each action. | Hard stop on A. Rebuild I completely first. |
Recovery Signal |
I turning up, A still low | I-led recovery underway. Correct sequence. Fragile — do not rush. | Protect the I recovery. Let A follow naturally. |
A mid-size professional services firm launches a strategic transformation. New operating model, new incentive structures, new technology platform. Board-level support, adequate budget, experienced consultants. By every resource measure, the conditions are right.
The most common error in behaviour change practice. A person enters support in Trooper state — low accurate self-model of what drives their behaviour, constrained environment. The practitioner correctly identifies low capability and low opportunity.
The standard response is an action plan and resource provision. This is A-restoration for a system without adequate I.
The action plan gets used to execute the person's existing — inaccurate — model of their situation. It produces activity that looks like engagement and produces no change.
A community has an accurate model of their situation. They understand the health risk, the economic constraint, the structural injustice. They are in Abhimanyu state.
The intervention is an awareness campaign, a psychoeducation programme. More I-building for a system where I is not the constraint.
The community knows about the food desert, the underfunded school, the inaccessible healthcare. Telling them about it again — however elegantly — changes nothing.
CARA does not replace COM-B, Fogg, or any domain framework. It runs before them. It tells you which section of the tool library to open — and which sections will make things worse.
CARA is not a replacement for domain expertise. It is the question that must come before domain expertise is applied. What state is the system in? Which function is the binding constraint? What direction is each moving?
The framework has been validated across macroeconomic systems, political institutions, immune response, governance, education, and individual behaviour — consistently showing that state and trajectory determine intervention responsiveness, and that misidentifying state converts well-resourced interventions into committed errors.