Bolt Series · Human Performance Calibration · Article 39
Sometimes the correct answer is: we do not know yet
The student says:
“I think I can handle the harder course.”
The teacher says:
“I’m not sure.”
The parent says:
“I’ve seen signs both ways.”
The available scores are mixed.
One task was heavily supported.
Another was taken while the learner was unwell.
The remaining evidence is too narrow.
Education often dislikes this state.
We want a label.
Ready.
Not ready.
Strong.
Weak.
But sometimes certainty would be fabricated.
“We do not yet have enough evidence” is not a failure of judgement when it is true. It is the most accurate judgement available.
Uncertainty is part of good measurement
Serious measurement systems do not pretend every score is exact.
Educational measurement distinguishes an observed score from the uncertainty surrounding the inference made from that score.
Educational Measurement, Fifth Edition (2025), makes a simple but important point: a reliability coefficient by itself means little without context about the assessment, the population, administration conditions, scoring and the sources of error being considered.
In other words, responsible measurement does not merely produce a number.
It also asks how much confidence the number deserves for the decision being made.
Uncertainty should change what we do next
If nobody knows, there are two bad responses.
The first is false certainty.
“We have to decide, so let’s just call the student weak.”
The second is paralysis.
“We cannot know perfectly, so we cannot act.”
Good calibration does neither.
It asks:
What is the cheapest, safest and most informative next action?
Confidence should help regulate information seeking
Metacognitive research treats confidence as part of a control system.
Low confidence can justify gathering more information before committing to a choice.
Computational work on information search shows that uncertainty and confidence can help determine whether additional evidence is worth the cost.
A 2025 study across five experiments found that stronger metacognitive sensitivity predicted better information-search quality: participants were better at deciding both what information to seek and when to stop searching.
This suggests an important educational skill.
Not merely knowing the answer.
Knowing when you do not yet have enough evidence to justify the answer.
Children have to learn this too
Uncertainty monitoring develops.
Research with children shows that adaptive help-seeking improves across childhood and early adolescence and is related to developing metacognitive skill.
Newer developmental work also suggests that young children can struggle to respond appropriately to uncertainty when there are no obvious external cues that information is missing.
This matters because school often rewards answers more visibly than uncertainty management.
A child learns to put something in the box.
But adult competence often depends on being able to say:
“I am not confident enough to act yet. I need one more piece of evidence.”
Not all extra information is useful
There is another trap.
When uncertain, people can keep collecting information without improving the decision.
Ten more routine worksheets may not tell us whether transfer is secure.
Five more opinions from people who all read the same report may not add independent evidence.
Another global confidence rating may not distinguish concept weakness from time pressure.
Good information seeking is targeted.
Ask what uncertainty remains.
Then seek the evidence capable of reducing that specific uncertainty.
Sometimes the best next step is a reversible decision
Suppose there is not enough evidence to know whether a learner is ready for much harder work.
The choice does not always need to be permanent.
Try a bounded period.
Set explicit criteria.
Observe independent performance.
Then review.
When uncertainty is high, reversible decisions can preserve opportunity while generating new evidence.
This is much safer than pretending an uncertain early judgement deserves permanent consequences.
The stakes determine how much uncertainty is acceptable
A teacher deciding tomorrow’s practice set can tolerate substantial uncertainty.
Choose a plausible next task.
Observe the return.
Correct quickly.
A high-stakes decision that permanently narrows opportunity deserves much stronger evidence.
This is a general principle of responsible decision-making:
The greater the consequence of being wrong, the more seriously we should treat uncertainty before making an irreversible move.
Sometimes uncertainty itself is miscalibrated
A person can feel uncertain and actually possess enough evidence.
A person can feel certain when the evidence is weak.
That is why “I feel unsure” should not automatically trigger endless searching.
A 2025 study found that the relationship between confidence and information seeking is context-dependent rather than mechanically simple.
The important skill is not maximum information seeking.
It is knowing whether the remaining uncertainty is large enough, relevant enough and consequential enough to justify more evidence.
A practical uncertainty protocol
When nobody has enough evidence, ask:
- What exactly is uncertain?
- Which competing explanations remain plausible?
- What observation would distinguish them?
- How reliable would that observation be?
- What does waiting cost?
- What does acting wrongly cost?
- Can we make a reversible decision while collecting better evidence?
- What result would tell us to stop searching?
Now “I don’t know” has become an operational state rather than a dead end.
Why this matters for education
One of the most dangerous habits education can teach is that every important question must immediately have a confident answer.
Real expertise often contains a better sentence:
“This is what the current evidence suggests. This is how certain I am. This is what I still do not know. And this is the next observation that would most improve the decision.”
That sentence is not weak.
It is calibrated.
Sometimes the student does not know.
Sometimes the teacher does not know.
Sometimes nobody knows yet.
The important thing is that the system knows what to do next.
Evidence and further reading
- Metacognition and Confidence: A Review and Synthesis
- Metacognitive Computations for Information Search
- Metacognitive Sensitivity Predicts Information Search Quality
- Context-Dependent Role of Confidence in Information Seeking
- Active Help-Seeking and Metacognition in Children
- Development of Uncertainty Sensitivity
- Educational Measurement, Fifth Edition
