HOW INTELLIGENCE WORKS · SPIRAL DIRECTION · GOALS · VALUES · OBJECTIVES · eduKateSG

The Direction of the Spiral

Better at what—and who decides what “up” means?

The Spiral Closed Loop gives us an attractive picture of intelligence developing through repeated contact with reality. World Return closes the loop. Retained Learning prevents every revolution from beginning again. Spiral pitch describes how much durable capability survives. But one question remains underneath all of them.

Which direction counts as improvement?

A system can become faster, cheaper, more accurate, more reliable, more profitable, more energy efficient or easier to use. These improvements can conflict. Intelligence therefore cannot obtain its direction from feedback alone. Feedback reports consequences relative to something the system cares about.

The spiral has no meaningful “up” until the system has an objective, constraint or value by which change can be judged.

1. The car can improve in several directions

Return to the car factory. Engineers can reduce weight. That may improve energy use but worsen noise isolation. They can soften the suspension for comfort but reduce handling precision. They can add safety structure but increase mass and cost. They can simplify manufacturing but reduce repairability.

There is no single scalar called “better car” hiding inside the vehicle waiting to be measured. The product exists inside a field of objectives and constraints.

Performance ↔ safety ↔ cost ↔ reliability ↔ comfort ↔ efficiency ↔ maintainability ↔ manufacturability

Engineering intelligence includes understanding these trade-offs rather than pretending every improvement points in the same direction.

2. Feedback requires a comparator

In a control loop, an observed state becomes an error only relative to a desired state or reference. If the target temperature is 22 degrees, 25 degrees has meaning as a deviation. Change the target and the same observation acquires a different interpretation.

The same principle appears in learning and organisations. A student’s answer is judged against a task. A company’s result is judged against an objective. A model’s prediction is judged against observed outcomes and a loss function. A design is judged against requirements.

Reality supplies the observed state. The objective supplies the direction of correction.

3. Intelligence can optimise the wrong objective

This is one of the deepest failure modes in the whole architecture. A system can sense accurately, reason correctly, close its loops quickly, retain every lesson and still become worse in the larger sense because the objective is badly specified.

A factory maximises throughput while reliability falls. A school maximises examination performance while curiosity and transfer decline. A customer-service system minimises call duration while unresolved problems rise. A machine optimises the measurable reward while violating the intention behind it.

The loop is functioning. The direction is wrong.

4. Proxies create apparent upward spirals

Many important objectives are difficult to measure directly. We therefore create proxies: test scores for learning, defect rates for quality, clicks for engagement, waiting time for service performance, quarterly numbers for business health.

Proxies can be useful sensors. Trouble begins when the proxy becomes easier to optimise than the underlying purpose. Then the dashboard rises while the real system may deteriorate.

This is why When the Spiral Goes Down warned that the most dangerous decline can be one whose measurement system calls it improvement.

5. Objectives themselves can be versioned

A mature intelligence does not only ask whether its actions are achieving the objective. It can sometimes ask whether the objective still deserves control.

The inner loop asks, “How do we build this car better?” A higher loop asks, “Are these still the right requirements?” A higher strategic loop asks, “Is this still the right product?” A still broader loop asks, “What problem are we actually trying to solve?”

This is nested intelligence applied to goals. Outer loops can revise the reference points governing inner loops.

6. Creativity can change direction, not only expand dots

Creativity versus Intelligence showed creativity moving dots outward. But creativity can do something more radical: change the frame in which the dots are judged.

Instead of asking how to make a petrol car more efficient, someone asks whether propulsion should be electric. Instead of asking how to improve a road, someone asks whether the mobility problem should be solved by rail, walking, telepresence or a different urban arrangement.

The creative move can therefore rotate the search space. Intelligence must then evaluate whether the new direction serves the deeper objective better.

7. Human intelligence contains plural objectives

Human decisions rarely optimise one variable. We care about speed and accuracy, present reward and future capability, ourselves and other people, efficiency and resilience, freedom and coordination, novelty and continuity.

This makes judgement central. Intelligence often operates under objectives that cannot be reduced honestly to one number.

Trade-offs therefore need explicit reasoning. What is non-negotiable? What can be exchanged? Over what time horizon? Who bears the cost? Which uncertainty matters? What happens if the optimisation succeeds too well?

8. Education shows why direction matters

Suppose a student wants higher marks. That is a legitimate local objective. But education has larger objectives: durable knowledge, transfer, independent judgement, communication, responsibility and the ability to continue learning.

A tutoring method can raise tomorrow’s score by supplying every next step. If it simultaneously reduces independent problem solving, the inner performance loop rises while the developmental loop may fall.

Good teaching therefore keeps several loops visible. Immediate performance matters, but so do retention, transfer and eventual independence.

9. Direction changes with time horizon

An action can look upward over one hour and downward over one year. Skipping maintenance raises today’s output. Intensive cramming raises tomorrow’s familiarity. Cutting training improves this quarter’s costs.

The Speed of the Loop showed why intelligence needs several clocks. Direction also needs several horizons. A system should know whether it is judging immediate state, medium-term capability or long-term resilience.

10. System boundaries change the answer

What looks like improvement inside one boundary can export costs outside it. A factory can reduce its own costs by shifting burden to suppliers. A department can improve its metric by creating work downstream. A product can become cheaper while becoming harder to repair.

Intelligence therefore has to ask where the system boundary has been drawn. Which consequences are inside the objective and which have been made invisible by accounting or organisational structure?

Changing the boundary can turn an apparent gain into a trade-off.

11. Civilisation has no single dashboard

At civilisation scale, direction becomes even more plural. Health, knowledge, security, prosperity, liberty, environmental durability, cultural continuity, opportunity and institutional trust cannot be collapsed without argument into one universal metric.

This is why civilisations need multiple institutions and forms of evidence. Different systems reveal different consequences and represent different legitimate concerns. Intelligence at this scale is partly the capacity to keep those consequences visible while decisions are made.

12. Direction needs constraints

Some objectives should not be pursued without limits. Engineering uses safety margins. Medicine uses ethical constraints. Organisations use laws, standards and governance. Education limits what may be sacrificed for performance.

Constraints shape the feasible direction of the spiral. They say, in effect: improve, but not by crossing this boundary.

13. Intelligence can ask a higher-order question

At the highest level, intelligence is not only solving the current problem. It can inspect the problem definition itself.

Are we measuring the right thing? Are we optimising the right thing? Are we solving the right problem? Are the constraints still valid? Has the environment changed? Who experiences the consequences we are not measuring?

These questions are not interruptions to intelligence. They are the outer loops that keep powerful inner loops from becoming efficiently wrong.

14. The spiral needs three coordinates

The geometry can now be sharpened. Speed tells us how quickly a loop closes. Pitch tells us how much durable capability survives. Direction tells us what the system is becoming better at and whether that improvement serves the objective that actually matters.

A system can therefore be fast, high-pitch and badly directed. That would be a powerful learner becoming rapidly better at the wrong job.

The deepest question is not whether intelligence improves. It is whether it improves in a direction worth preserving.

That question completes another layer of the Spiral Closed Loop: reality tells us what happened; intelligence judges what changed; values, goals and constraints determine what should count as better.