HOW INTELLIGENCE WORKS · PLAUSIBILITY JUDGEMENT · eduKateSG
How Intelligence Decides Which Explanations Are Coherent Enough to Keep Alive Before Testing
Plausibility judgement is the intelligence process that triages possible explanations before expensive testing by asking whether they are coherent with established knowledge, compatible with the known constraints, specific enough to make contact with evidence, and credible enough to deserve scarce investigative effort.
Candidate explanation → prior credibility → coherence → mechanism → constraint fit → generality → cost of testing → keep, revise, park or reject.
This article belongs to the How Intelligence Works series. Hypothesis Generation owns producing possible explanations. Hypothesis Testing owns separating them with evidence. Plausibility judgement owns the triage between those stages: which candidates deserve to remain active competitors before testing resources are spent?
Not Every Possible Explanation Deserves Equal Attention
A problem can support dozens of imaginable stories. Intelligence cannot test all of them with equal depth. It therefore needs a pre-test filter.
That filter must be disciplined enough to remove incoherent candidates, yet open enough not to eliminate novel explanations merely because they are unfamiliar.
Plausibility is not proof. It is a decision about whether an explanation deserves to stay in the game.
1. Plausibility Begins With Prior Credibility
Before new evidence arrives, some explanations fit established knowledge better than others. Prior credibility does not decide the case, but it affects how surprising it would be for the explanation to be true.
A very low-prior explanation may still deserve testing if the potential payoff is large or the evidence would be decisive. Plausibility judgement therefore ranks rather than simply bans.
2. Coherence Is Multi-Dimensional
| Plausibility dimension | Question |
|---|---|
| Internal consistency | Does the explanation contradict itself? |
| External consistency | Does it conflict with well-established knowledge? |
| Mechanistic credibility | Is there a possible route by which the cause could produce the effect? |
| Constraint fit | Does it respect known boundaries, timing and quantities? |
| Generality | Can it explain more than one narrow observation without ad hoc repair? |
| Testability | Can the explanation make contact with discriminating evidence? |
3. Plausibility and Truth Are Different
Many true discoveries initially looked implausible because they conflicted with prevailing models. Many false explanations look plausible because they fit familiar stories.
Plausibility therefore regulates attention before evidence, not final belief after evidence.
4. Unspecified Alternatives Matter
A familiar explanation can feel more plausible partly because it is easier to retrieve. Explanations not currently in mind may be underestimated simply because they are absent from the active comparison set.
Before calling one explanation “obviously plausible,” ask what credible alternatives have not yet been generated.
5. Plausibility Judgement in Mathematics
Mathematical conjectures can be screened by consistency with known theorems, boundary cases, dimensions, symmetry and small examples before full proof.
This does not replace proof. It helps allocate effort toward conjectures that are coherent enough to justify deeper investigation.
6. Plausibility Judgement in Learning
A teacher diagnosing a student’s error may consider several explanations. “Carelessness” is easy to say but often mechanistically weak. A more plausible diagnosis should fit the error pattern, prior performance, task demands and observed behaviour.
The aim is not to choose the final cause immediately. It is to maintain a small set of credible candidates that deserve diagnostic testing.
This is where plausibility triage protects teaching from both overreaction and random intervention.
7. Good Explanations Should Earn Their Complexity
An explanation that requires many special exceptions, hidden causes and one-off adjustments may be less plausible than a simpler mechanism with broader coverage.
Simplicity is not an absolute law, but unnecessary complexity creates more ways for an explanation to be wrong.
8. High-Risk, High-Reward Hypotheses May Deserve Testing Despite Low Prior Plausibility
Research triage should consider not only prior credibility but also test cost, expected information gain and the consequence of being wrong.
A low-probability explanation can deserve investigation when the test is cheap and the potential importance is large. Plausibility judgement therefore interacts with Value of Information.
9. Plausibility-Judgement Failure Atlas
| Failure | What happens | Repair |
|---|---|---|
| Familiarity capture | Known explanations feel more plausible because they are easy to retrieve | Generate alternatives deliberately |
| Novelty rejection | Unfamiliar explanation is dismissed without mechanism review | Separate unfamiliar from incoherent |
| Story coherence illusion | A smooth narrative substitutes for evidence | Ask what would discriminate it |
| Ad hoc rescue | Every contradiction produces another exception | Track complexity cost |
| Prior dominance | Low prior prevents all testing | Include information value and test cost |
| Mechanism blindness | Correlation is accepted without a plausible route | Specify causal or logical mechanism |
| Overconfidence | One active hypothesis crowds out unspecified alternatives | Reserve probability for unknown candidates |
10. Plausibility Judgement and Hypothesis Testing Are Different
Plausibility judgement decides which hypotheses deserve testing attention. Hypothesis testing determines what evidence can separate them and what happened when the test was run.
One allocates search. The other adjudicates with evidence.
11. Teams Need a Plausibility Ledger
When a team considers multiple explanations, it helps to record why each remains plausible, which assumptions it requires, and what would make it lose priority.
Do not allow the favoured story to become the only story merely because it was spoken first.
12. Institutions Need Space for Low-Prior but Important Questions
Institutions that fund only familiar hypotheses become efficient at confirming the current map. Institutions that test every speculative idea become wasteful.
A healthy portfolio reserves some capacity for plausible, testable and potentially consequential ideas outside the dominant model.
13. Artificial Intelligence and Plausibility Triage
Generative AI can produce many candidate explanations cheaply. That makes triage more important, not less.
A reliable system should distinguish a fluent explanation from a plausible one by checking constraints, mechanisms, prior knowledge, source quality and whether the explanation can make discriminating predictions.
Generation creates the hypothesis pool. Plausibility judgement prevents the pool from becoming an unranked flood.
14. The Plausibility Judgement Audit
- Candidate: What exactly is the explanation?
- Prior: How credible was it before new evidence?
- Consistency: Does it contradict itself?
- Knowledge fit: What established findings support or constrain it?
- Mechanism: How could the explanation produce the observed effect?
- Generality: Does it explain more than one narrow fact?
- Alternatives: Which plausible explanations are not yet active?
- Testability: What evidence could move it up or down?
- Cost: What does testing consume?
- Decision: Keep, revise, park, reject or prioritise?
15. CivDJ Reading: Decide Which Masters Deserve to Reach the Test Bench
In the CivDJ frame, the Warehouse may contain many candidate Masters. Plausibility judgement decides which ones are coherent enough with the Receiver, constraints and known structure to deserve live comparison.
The process should be selective without becoming closed.
Do not test every imaginable mix. Do not test only the familiar ones either.
16. Return to the Candidate Explanation
Plausibility judgement keeps intelligence from drowning in possibilities or hardening around the first familiar story.
It uses prior credibility, coherence, mechanism, constraints and testing value to maintain a disciplined shortlist.
The mature mind does not confuse plausibility with truth. It uses plausibility to decide which questions deserve the next unit of evidence.
Research reading: Bayesian hypothesis generation: evaluating novel hypotheses before data collection · Plausibility assessments in hypothesis generation.