Human Judgement

Why human judgement is becoming the defining capability of the AI age.

For much of the modern knowledge economy, professional advantage came partly from knowing things other people didn't.

Expertise was scarce. Research took time. Access to information really mattered.

Artificial Intelligence is changing that equation.

It enables us to retrieve knowledge, generate ideas, summarise research and produce plausible answers in seconds.

The answers are becoming cheaper.
Good judgement isn't.

Information has never been more abundant.
Yet organisations aren't becoming proportionally wiser.

Leaders still make poor decisions.
Teams still optimise the wrong systems for the wrong things.
Institutions still struggle to adapt.

The problem is no longer a lack of information.
It's knowing what to do with it.

• The ability to distinguish signal from noise.
• To recognise trade-offs.
• To decide what is important.
• To act responsibly when certainty isn't available.

As technology reduces the cost of producing information and options, the value of deciding what deserves attention and action increases.

The defining challenge of the AI age isn't learning to use increasingly capable tools.
It's developing the judgement to know when, why and whether we should use them at all.

Judgement is the disciplined movement from interpretation to action.

Information tells us what we know.
Interpretation asks what it means.
Judgement determines what matters enough to act on.
Commitment turns that judgement into a choice.
Action exposes the choice to reality.
Reflection closes the loop.

Most people think judgement is something we possess.
I think it's something we practise.

What Is Human Judgement?

Human judgement is often treated as a mysterious quality people either have or don't have.

We talk about someone having "good judgement" as though it were an innate personality trait, or a soft skill that sits somewhere alongside communication and teamwork.

I think that's a mistake.

Judgement is a capability. And capabilities can be developed, practised, and conversely degraded.

It's also a deeply practical capability, and one whose value is increasing as machines become better at producing information and options.
Human judgement is the capability to interpret imperfect information, weigh competing priorities and consequences, and commit to responsible action when no answer is certain.

A good decision can still produce a bad outcome.
A bad decision can occasionally produce a good one.

Judgement is therefore not the same as getting the answer right. It concerns the quality of the reasoning available at the time a decision is made.

Knowledge contributes to judgement.
Experience strengthens it.
Reflection refines it.
Character guides it.

But none of these is judgement on its own.

Knowledge tells us what we know.
Experience shows us patterns.
Reflection lets us update those patterns.
Character determines what we're prepared to do when values and interests collide.

Judgement emerges from the interaction of these.

Judgement Is Not Intelligence

Intelligence helps us understand complexity.
Knowledge gives us information.
Critical thinking helps us question claims and reasoning.
Decision-making turns evaluation into choice.

Judgement sits across all four of these capabilities.

Decision-making asks what we will do. Judgement asks what deserves to matter when we decide.

It determines what deserves weight when information conflicts, priorities compete and certainty is unavailable.

An intelligent person can still be overconfident.
An experienced person can still miss an important signal.
A highly informed organisation can still make a terrible decision.

More information doesn't automatically produce better judgement.

How Human Judgement Erodes

Poor judgement rarely begins with bad intentions.

More often, it begins with small compromises that gradually disconnect people from reality.

We become overwhelmed by information and mistake volume for understanding.
Speed gradually replaces reflection.
Incentives reward what is immediately measurable rather than what is ultimately valuable.
Increasingly capable tools subtly tempt us to stop exercising the very capabilities they were meant to support.

Cognitive offloading
We stop performing parts of a task because a tool can perform them for us.

Automation bias
We give machine-generated recommendations more confidence than they deserve.

Capability atrophy
Capabilities weaken when they're no longer exercised.

Responsibility diffusion
We become less certain about who owns the decision when humans and machines both contribute to it.

Judgement rarely disappears all at once.
It erodes.

That's why developing judgement is not simply a matter of learning more.

It's also a matter of protecting the conditions that allow good judgement to emerge.

Reflection.
Attention.
Intellectual humility.
Constructive disagreement.
Responsibility.

These are not obstacles to better decisions.
They are often the very things that make better decisions possible.

Good judgement is rarely lost through one dramatic mistake. More often, it's slowly optimised away.

What Should We Delegate?

The question isn't whether AI can perform a task.
Increasingly, it can.

The better question is whether delegating the task changes our relationship with the judgement required to do it well.

Some work is harmless to automate.
Some work can be automated while keeping a human meaningfully involved.
Some decisions become dangerous when the human becomes detached from the reasoning behind them.

The more consequential, ambiguous or difficult to reverse a decision becomes, the stronger the case for retaining human contact with the reasoning that supports it.

The Contact Test

So how do we decide what to delegate?

I use what I call the Contact Test. It asks whether automating a task removes contact with something we still need in order to exercise good judgement.

Before removing a human step, ask:

• Does this build judgement?
• Does this expose me to useful signals from the real world?
• What reality do I stop touching if I remove it?

These questions don't assume that human involvement is inherently better. Much routine work can and should be automated.
The test is whether the efficiency gain also removes something we still need to notice, understand or practise.

Required Friction

The Contact Test identifies what automation might remove. Required Friction asks whether some of it is worth preserving.

Not every form of effort is waste.

Some effort is just friction. It slows us down without improving the outcome.
Other effort is developmental, forcing us to interpret, decide, notice consequences and ultimately learn.

Required Friction is the effort worth preserving because removing it would also remove an opportunity to develop or exercise judgement.

AI should remove unnecessary effort. It shouldn't automatically remove the effort through which capability is built.

How Do You Develop Better Judgement?

Good judgement develops through contact with consequences.

This is one reason complete delegation can be developmentally dangerous. If we never make the decision, we don't experience the consequence, and without the consequence there is less to learn from.

1. You encounter a problem.
2. You make a decision.
3. Reality responds.
4. You reflect on what happened.
5. You revise the model you were using.
6. Then you make the next decision with slightly better information.

This is why experience alone isn't enough.

Ten years of repeating the same assumptions isn't ten years of better judgement.

Experience becomes developmental and beneficial when it's properly examined.

Judgement within Stoic Futurism

Within Stoic Futurism, judgement is the capability that connects knowledge to agency.

It sits between understanding what is happening and deciding how to respond.

This is why judgement connects the other Stoic Futurist disciplines: information literacy helps us establish what deserves belief; decision-making turns judgement into choice; resilience preserves our ability to exercise judgement under pressure; leadership determines where judgement lives within a system.

Judgement and Responsibility

Judgement sits at the intersection of knowledge, experience, values and responsibility.

Judgement can be supported by tools.
Parts of it can even be delegated.

Responsibility is different, though.

Someone still has to decide what deserves to be done, what risks are acceptable and what consequences they're prepared to own.

Artificial intelligence will continue making information cheaper.

It will continue accelerating analysis.
It will continue expanding what is technically possible.

None of that removes the need for judgement.
If anything, it increases it.

Technology can generate options.
It can accelerate analysis.
It can recommend a course of action.

But someone still has to decide what matters.
And someone still has to answer for what happens next.