How Judgment Grows

The sheet has no obvious mistakes. Every cell is filled. The totals add up. The dates are in order.

The new employee sends it to her supervisor, who asks why a delivery has been promised for Friday.

Because that is the date in the customer request.

Has the supplier confirmed it?

The sheet does not have a column for that.

Here is a lesson that a tidy result can hide. The beginner has answered the visible question and missed a condition that makes the answer safe to use. A better spreadsheet might help. So might a better brief. But she also needs to learn that a requested date and a confirmed date are different kinds of information.

Someone can tell her the rule. To make it part of her own judgment, she needs a chance to use it when the next request arrives in a slightly different form.

That is the kind of learning a first job should make possible.

Give the decision a name

Judgment sounds grand until we look at an ordinary decision. Should I send this? Is the information good enough? Which part needs another check? Who should hear about the delay? May I change the plan myself?

A beginner may know how to carry out an instruction and still need help with these questions. The difficulty is often deciding what the task really requires. A finished document is visible. The doubts considered and rejected are usually harder to see.

If you are teaching, bring some of that hidden work into view. Before doing a task together, name the decision. We need to choose a delivery date we can stand behind. We need to tell the customer what is known without making a promise we cannot keep. We need to find whether the error is in the source file or in the way we processed it.

These sentences give the learner something more useful than watch me work. They identify what the effort is for.

Then let the learner make a first proposal where it is safe to do so. They might say, I would call the supplier before sending the date. Ask what they expect the call to settle. A vague instruction to double-check everything becomes a specific test: is the shipment available, and has the supplier actually committed to the date?

You are trying to reveal the connection between evidence and action. The learner should see why a fact matters to the choice, rather than collect facts as a sign of diligence.

The same approach works for a small repair, a support message or a budget draft, with the safety limits appropriate to each. Some actions must remain with a qualified person. The learner can still inspect the reasoning, prepare questions or work through a protected example. Participation does not have to begin with permission to cause real harm.

Make the first attempt small enough

A badly designed first assignment offers too little responsibility or far too much.

At one extreme, the newcomer copies data for weeks without seeing where it goes. They may learn speed and consistency, but they cannot connect the task to a customer, a decision or an error that matters. At the other, they are told to take ownership of a live problem whose limits no one explains. When something goes wrong, the organization discovers an urgent interest in how junior they are.

Both arrangements are avoidable.

A first assignment can have a real purpose and a narrow boundary. Prepare a draft for these three orders. Identify information that remains unconfirmed. Do not contact the customer or change the live schedule. We will review it together before anything is sent.

This gives the beginner room to think. It gives the supervisor a clear place to check. It gives everyone else protection from an unsupervised decision.

The next assignment can change one condition. Perhaps the learner handles a request with two missing details. Later, they make a routine confirmation themselves. Eventually, they may manage a familiar set of orders and bring only the unusual cases for review. The progression should depend on observed work, not merely on the passing of a calendar date.

This is a proposal for organizing responsibility, not a promise that every skill follows neat steps. Some work is messy. Some learners move quickly in one part and slowly in another. A person may be excellent at spotting an inconsistency while still needing help to communicate it. A good supervisor should be able to describe that unevenness without turning it into a verdict about the whole person.

The purpose of a smaller task is to make learning possible. It should not become a permanent excuse to deny someone harder work.

Let an error teach something specific

Imagine that the delivery draft still contains the wrong date. The supervisor could fix it silently. The team would get the right sheet, but the learner might never discover what happened.

The supervisor could also write careless at the top. That would communicate displeasure while leaving the actual problem largely unexplained.

John Hattie and Helen Timperley’s review of educational research distinguishes feedback about a task or process from praise directed at the person. It also makes a useful qualification: when basic understanding is missing, the learner may need instruction before feedback can help. This is an educational review, not proof of a universal workplace method.¹

So first ask whether the newcomer understands what a confirmed date means. If that concept is missing, teach it. A useful review can then locate the error. The requested date was copied into the confirmed-date field. That matters because another team uses the field to promise a slot. The source email says the supplier is still checking availability. The repair is to mark the date as pending and ask for confirmation before release.

Notice the movement. The learner can now connect the original information, the choice, its possible consequence and the correction. They can examine their own method. Did they miss the word pending? Did they assume every date in the message had the same status? Did the template encourage the mistake?

If the template is poor, fix the template. Teaching should not require newcomers to overcome preventable traps so that the organization can test their character.

Ask the learner to make the correction, if time and safety allow. Then ask them to explain what they would look for in the next case. This creates an opportunity to see whether the explanation made sense. A nod may mean understanding. It may mean the person wants the uncomfortable conversation to end.

The review need not become a lecture. A short, precise exchange can be enough for a small error. A serious mistake may need a fuller investigation, especially when the system allowed it to reach a customer. In that case, separating the learner’s action from the conditions around it is part of being fair.

Blame that stops at the least powerful person can make it harder to bring problems into view. A first rung should make it possible to bring a problem early, while it is still small enough to fix.

Keep some work visible

AI can make the first result appear before a learner has had to decide much. So can an overhelpful supervisor. The issue is not unique to a machine: someone else can do the thinking and leave you with the appearance of completion.

A teaching task should therefore leave a few decisions visible. Before accepting a suggested answer, ask what information it relies on. Before sending a summary, identify one important point that might have been left out. Before approving a calculation, check a case you understand by hand or by another suitable method.

Do not turn this into a requirement to redo every automated step forever. That would make the tool rather pointless. Choose checks that tell you something about the result and about the learner’s grasp of it. The right check changes with the work and its possible harm.

For the delivery example, the learner might identify which promises have a source and which remain uncertain. That is more relevant than asking them to recreate the font and spacing without assistance. If the learning goal is a particular calculation, however, they may need to work through that calculation themselves before relying on automation.

Ask what must be understood for this person to use the output responsibly. Then design the practice around that need.

A beginner can do this for themselves too. When a tool gives you an answer, pause before you admire the finish. Find the part that would be hardest to defend to a skeptical colleague. Try to explain it in ordinary language. If you cannot, you have found something worth learning next.

That moment is useful information. There is no need to treat it as a moral failure. The problem would be hiding the gap and making a promise that depends on knowledge you do not yet have.

Change the case before you raise the stakes

A learner may repeat a procedure well because the new example looks almost exactly like the old one. To understand the limit, change a detail while keeping the exercise safe.

Give the delivery planner a customer who accepts a partial order. Give them a supplier whose date is confirmed but whose quantity is uncertain. Ask what changes if a delay affects an event that cannot be moved. A person who learned only to fill the sheet may repeat the same steps. A person who sees the purpose can begin to explain why the choice differs.

Again, the exercise is a way to inspect understanding, not a proven score for ranking human potential. It should reveal where more teaching is needed.

A small experiment by Nina Keith and Michael Frese gives this distinction some support. Fifty-five university students practiced presentation software under different training conditions. Error-management training helped with a later task requiring changed procedures; performance on more familiar transfer tasks did not differ. This narrow result supports testing adaptation separately. It does not authorize risky errors in live work or establish a complete training model.²

The distinction matters in assessment. If an employer uses a changed case as a trick, a nervous applicant may spend the interview guessing what the interviewer wants. If the employer explains the purpose, the applicant can show how they respond to new information. The interviewer learns more from the reasoning than from watching someone discover a hidden rule too late.

The changed case can also show a learner that there is no single perfect answer independent of circumstances. A cheap delivery may be sensible for an ordinary restock and a poor choice for a time-critical event. A short email may be clear for one recipient and incomplete for another. The aim is to connect the answer to the purpose.

That is why judgment needs contact with consequences. Not every consequence has to be suffered in real life. A well-chosen example can make it visible. But a person should eventually see how work is used, what happened after a decision and which concerns turned out to matter.

If the beginner never hears what happened next, the task ends too early for learning.

Responsibility includes a stopping point

Some workplaces praise independence in a way that makes asking for help feel like failure. This is especially unhelpful for someone who has little basis for knowing how serious a problem is.

A supervisor can make the stopping point explicit. You may confirm routine dates within these limits. Bring any change in price to me. Stop if the customer’s instructions conflict with the approved order. Do not put personal data into an outside tool. Tell us immediately if a record may have gone to the wrong person.

The exact boundaries will differ. They should be clear before an incident makes everyone wish they had been clear.

As confidence grows, revise the limits. Otherwise the newcomer can become dependent on approval for decisions they are ready to make. A supervisor who insists on checking every harmless detail may feel secure while keeping the learner from progressing.

There is a balance to work out, and it needs conversation. What can you handle? What have I seen you handle? What would make me comfortable giving you more room? Which risks still require another pair of eyes?

These questions make trust more concrete. Trust can grow from evidence without requiring a newcomer to pretend certainty. It can also survive a mistake when the person reports it, helps repair it and changes the method that led there.

Amy Edmondson’s study of fifty-one teams in one manufacturer found an association between psychological safety and learning behavior. Its cross-sectional design could not establish cause and effect. The finding is a reason to take the interpersonal cost of asking questions seriously, while examining the actual conditions in a particular workplace.³

A workplace should make those actions possible. If asking a question leads to ridicule, the official instruction to escalate concerns will have little practical force. The learner will follow the rule that affects their day, not the one printed on the wall.

The lesson the workplace is teaching

Every entry role teaches something about power.

A beginner notices whose work gets checked and whose mistakes are quietly corrected. They notice whether the person who teaches them is given time or merely thanked for being generous. They notice whether speed always wins when speed and care come into conflict.

These are not small background details. They tell the newcomer how to survive in that place. An organization can teach careful methods in a training session and reward haste for the rest of the week. It should not be surprised when the second lesson takes hold.

A useful first rung brings the daily incentives closer to the stated aim. Give review time a place in the schedule. Make it acceptable to report uncertainty. Recognize the colleague who prevents an error by asking a good question. Let the learner see the downstream result, including when a cautious choice protected someone else’s work.

This takes resources. It also takes restraint from people who could finish a task more quickly themselves. Sometimes they should finish it themselves; a real emergency is a poor classroom. But if every ordinary deadline becomes an emergency, the promise to teach will always move to next week.

The employee with the delivery sheet does not need a speech about becoming a leader. She needs another chance to prepare it, this time with the distinction between requested and confirmed made clear. She needs someone to check the parts that still carry risk. Then she needs room to handle the next familiar case.

The growth may look modest from outside. A field is corrected. A question arrives before the promise is sent. One routine decision no longer needs the supervisor.

That is how the first rung starts to hold a person’s weight.

Sources and limits