PSLE Science Reality Lab Vol No.035 | “We Predicted It” — Was the Prediction Written Before the Results Were Known?

PSLE-SCI-REALITY-0035

Wait, What? A perfect prediction can become much less impressive if it was written after the result appeared.

A science video shows ten trials. After the final result, the presenter says, “Exactly as predicted, the material with more holes cooled faster.” The graph does match the explanation. The words sound scientific. But one important fact is missing: when was the prediction written?

If the presenter stated the prediction before the trials began, the result can test that prediction. If the presenter first looked at the completed graph and then invented an explanation that fits it, the explanation may still be useful — but it is not the same kind of evidence. Seeing the answer first makes it easier to tell a story that matches the answer.

This is why scientific research sometimes records plans in advance. The Center for Open Science describes preregistration as specifying a research plan before the study and submitting it to a registry. One purpose is to distinguish planned work from decisions made after the results are known. Planned and exploratory work can both be valuable, but they do different evidential jobs.

For a Primary 5/6 learner, the idea can be much simpler: a prediction is strongest as a test when it exists before the result it is supposed to predict.

Quick Answer

When someone says “we predicted this result”, check for a record made before the result was known: a dated plan, lab note, worksheet, registered protocol, timestamped post or another traceable statement. If the explanation was created only after seeing the result, call it an after-the-result explanation or an exploratory idea, not an independent successful prediction. Then ask what new test could challenge it.

Reality Lab habit: before-result predictions test ideas; after-result explanations generate ideas for the next test.

Owned Learner Job — and the Boundary

This article owns one real-world evidence-transfer job: evaluating a claim that a scientific result was predicted by checking whether the prediction was recorded before the result was known.

It does not replace the PSLE Science owners for writing predictions, comparing predictions with results, building hypotheses, or keeping investigation records. Those pages teach the underlying inquiry skills. Reality Lab applies them to a public claim such as “our study predicted this”, “the model knew this would happen” or “we expected exactly this result”.

The Original Reality Lab Case: Four Boxes and a Surprise Pattern

A fictional student group tests four identical closed boxes under a lamp. Each box has a different covering material. They measure the air temperature inside each box after 20 minutes.

Box	Cover	Final temperature
A	White paper	31°C
B	Black paper	36°C
C	Silver foil	29°C
D	Clear plastic	34°C

After seeing the table, a student says, “I predicted silver foil would keep the box coolest because it reflects more light.”

Maybe the student really did predict that. Maybe not. The completed table alone cannot tell us. To evaluate the claim, we need provenance for the prediction: was it written in the group’s lab sheet before heating began? Was it stated in a timestamped message? Was the scientific reason recorded before the numbers appeared?

If yes, the observation can be compared with the original prediction. If no, the statement is better treated as an explanation suggested by the result.

Why Timing Changes the Evidence

Before an investigation, many outcomes may be possible. A useful prediction commits to one expected pattern and a reason. The result then has a chance to disagree.

After the investigation, the actual pattern is already visible. The mind can search for an explanation that matches what happened. This can be excellent scientific exploration — many important scientific ideas begin by noticing an unexpected pattern — but the same data have now helped create the explanation. They are no longer a fully independent test of that newly invented explanation.

A strong scientist therefore asks for a new opportunity to be wrong: a changed condition, new sample, later observation, or follow-up investigation that the new explanation predicts in advance.

Prediction, Explanation and Retrodiction Are Different Jobs

Reasoning job	When it happens	What it does
Prediction	Before the result	States what is expected to happen under stated conditions
Explanation	Before or after	Proposes why a result or pattern happens
After-the-result fit	After the result	Builds or selects an explanation that matches the observed pattern
New prediction	After forming the explanation, but before a new result	Creates a fresh test of the explanation

The problem is not that explanations are written after results. That is normal. The problem is calling an after-the-result explanation a successful prediction if no before-result record exists.

The Prediction Provenance Check

What exactly was predicted? A vague claim such as “something will change” is easier to match later than a precise expected pattern.
When was it recorded? Before data collection? Before analysis? After a first look at the results?
Where is the original record? Lab book, worksheet, timestamped document, preregistration or another traceable source.
Were the conditions stated? A prediction can depend on temperature, time, materials, sample type or measurement method.
Was the prediction changed after results appeared? Transparent revision is acceptable; pretending the revision was original is not.
What result would have counted against the prediction? A claim that can be stretched to fit every possible outcome is difficult to test.

Worked Case 1: The Exact Number That Was Not Actually Predicted

A public post says, “Our model predicted a 12% increase.” The archived plan made before the study actually says only, “We expect an increase.” After the observed increase was 12%, the public post uses the exact result as though the exact number had been predicted.

The evidence supports a weaker statement: the direction of change matched the recorded prediction. It does not support the stronger historical claim that 12% itself was predicted in advance.

Worked Case 2: An Unexpected Result Becomes a New Hypothesis

Students expect a covered cup to cool more slowly than an uncovered cup. Instead, one covered cup cools unusually fast. They notice condensation around the lid and propose a new explanation involving the lid not being sealed properly.

That explanation was not predicted beforehand. This does not make it useless. It creates a new question: if an imperfect seal caused the result, what should happen in a new comparison between a tightly sealed lid and a deliberately loose lid while other conditions are controlled?

Now the new explanation produces a before-result prediction for a new test. Scientific inquiry continues.

Worked Case 3: “We Always Knew” After the Headline

A viral science account posts an unexpected animal-behaviour result. Several commenters write that the result was “obvious” and “exactly what biology predicts”. But none links to a prediction, prior study, registered plan or earlier statement describing the specific pattern.

The result may fit existing biological ideas, but the comments alone do not prove that the exact result was predicted before observation. A strong reader separates consistent with a theory from successfully predicted in advance by a specific statement.

What Preregistration Adds — Without Turning It Into a Magic Badge

The Center for Open Science describes preregistration as writing a research plan in advance and submitting it to a registry. This helps make the timing of plans visible and helps distinguish planned confirmatory work from later exploratory work.

Preregistration is not a guarantee that the method is good, the measurements are accurate or the conclusion is correct. A poor plan can be registered in advance. New evidence can still reveal errors. Researchers can also transparently explain justified changes to a plan.

For a Primary learner, the transferable lesson is simpler than the research-system details: record important expectations before seeing the answer, then compare the answer with what was actually written.

PSLE-Style Transfer: Keep the Prediction From Moving

A student predicts that a plant kept in brighter light will grow taller over one week because it will receive more light for photosynthesis. After one week, the brighter-light plant is shorter. The student crosses out “taller” and writes “shorter” before completing the conclusion.

The change destroys the evidence trail. The original prediction should remain visible. The learner should compare the actual result with the original prediction, state that the result did not support it under those conditions, and investigate possible reasons. That exact inquiry skill is owned by eduKate’s guide on comparing a prediction with the actual result without rewriting the prediction afterward.

What Evidence Would Strengthen a “We Predicted It” Claim?

a timestamped prediction made before the result;
a clearly stated quantity, direction or pattern;
specified test conditions;
a reason grounded in prior scientific knowledge;
a clear statement of what outcome would disagree with the prediction;
the original prediction preserved even if later revisions are made;
a new independent test if the explanation was created after seeing the first result.

What Would Weaken It?

no record before the result was known;
a prediction so vague that almost any result fits;
changing the stated prediction after seeing data without disclosing the change;
claiming an exact number was predicted when only a direction was recorded;
using the same data to invent an explanation and then describing those same data as an independent test of it;
highlighting only the predictions that worked and hiding the ones that failed.

Tempting Reasoning That Fails

“The explanation fits perfectly, so it must have predicted the result.” Fit after observation is not proof of a before-result prediction.
“If an idea was created after the result, it is unscientific.” No. Exploration can generate excellent new hypotheses; the next step is a fresh test.
“Preregistered means proven.” No. Timing transparency does not guarantee method quality or truth.
“A prediction that fails is useless.” A failed prediction can reveal a missing condition, wrong assumption or better explanation.
“I remember thinking that would happen.” Memory after seeing a result is weaker evidence than a preserved before-result record.

Explained Practice

Practice A: The Direction Was Predicted

A dated lab sheet says, “We predict Group X will have a higher temperature than Group Y.” The result is 42°C for X and 38°C for Y. A later poster says, “We predicted exactly 42°C.” What is justified?

Answer: The direction — X higher than Y — was predicted. The exact value 42°C was not shown in the original prediction.

Practice B: The Explanation Came Later

A graph unexpectedly dips on Day 4. After seeing it, the group proposes that heavy rain on Day 4 reduced the measured temperature. Is that allowed?

Answer: Yes, as a possible explanation generated after the result. It should be labelled that way and tested against weather records or a new investigation rather than called a successful prior prediction.

Practice C: The Prediction Was Written First

A timestamped document says before testing that a rougher surface will produce greater friction in the specific set-up. The later result shows that pattern. What can you say?

Answer: The observed result is consistent with the recorded prediction under those tested conditions. Do not automatically generalise to every material and every situation.

Delayed Independent Return

Before your next science practice investigation, write one prediction and one reason on paper. Do not erase them. After the result, draw a line under the original prediction and write a separate evaluation: supported, contradicted, or not enough evidence. Then write one new prediction for a changed test. The physical separation between “before” and “after” makes the evidence history visible.

Routes to Existing PSLE Science Owners

Teaching Guide for Parents and Tutors

The useful habit here is not teaching young learners formal research-registration systems. It is teaching evidence timing. Ask the learner to maintain a two-column record: “before we knew” and “after we knew”. Predictions, planned variables and expected patterns go in the first column. Observed results, revised explanations and new questions go in the second.

If a learner changes the first column after seeing the answer, ask them to restore the original version. The point is not to embarrass them for being wrong. The point is that science learns from the difference between expectation and reality.

For stronger learners, give them an unexpected result and ask for two things: one after-the-result explanation and one new prediction that would test that explanation in a fresh situation. That makes the inquiry loop visible without requiring university-level statistics.

Authoritative Sources

The Quiet Return

Science does not become weaker when a prediction fails. It becomes more informative when the prediction was preserved clearly enough for the failure to be seen.

The next time somebody says, “We predicted exactly that,” ask one calm question before being impressed: where is the version that existed before the result?

Small Group Tutorials