IELTS Academic · Reading
Passage · Suggested time:
The useful limits of expertise
Public discussion often treats confidence as evidence of expertise. A speaker who answers without hesitation can seem more knowledgeable than one who carefully distinguishes what is known from what remains uncertain. Yet confidence is a property of a judgement, not a guarantee that it is well founded. The useful question is whether the strength of a claim bears a defensible relation to the evidence supporting it. This is difficult to assess from a single performance. Good judgement becomes visible through a pattern of predictions, explanations and revisions, especially when the person making them has something to lose by acknowledging uncertainty.
Forecasters use the term calibration for one aspect of that pattern. If, across a sufficiently large set of comparable events, predictions assigned a probability of seventy per cent occur about seven times in ten, the assessments are well calibrated at that level. Calibration is not the whole of forecasting skill. Imagine a fictional contest in which half the events occur and a participant assigns every event a probability of fifty per cent. The participant is calibrated across the set but offers little help in distinguishing likely events from unlikely ones. A useful evaluation must therefore ask both whether confidence is warranted and whether the forecasts discriminate.
The fictional forecasting tutor Sila Corren insists that revisions should be evaluated by their reasons, not merely by their direction. A forecaster who reduces a probability after receiving relevant contrary evidence may be reasoning well, even if the event later occurs. Someone who retains the original estimate solely to appear consistent may be reasoning poorly, even if the outcome happens to vindicate it. Corren does not praise every change of mind. A revision made to follow a prestigious colleague, without any new reason, is different from an update prompted by evidence. The visible action is the same; the intellectual process is not.
Group judgement introduces another distinction. Combining estimates can be useful when contributors make different errors, but a diverse list of occupations does not guarantee independent reasoning. Members may all rely on the same report or adopt the view of the first confident speaker. Corren therefore asks participants to make an initial estimate privately before discussion and to record which information they used. Later conversation can expose misunderstandings and improve those estimates, but independence must not be assumed simply because several people contributed. Ten repetitions of one analysis are not ten independent reasons to trust its conclusion.
The researcher Len Foyle focuses on the conditions under which expertise was acquired. An experienced planner may recognise familiar patterns quickly because the environment has provided frequent, interpretable feedback. If the rules governing that environment change, the same habits can become unreliable. Foyle's example is a transport planner whose assumptions were learnt when all services followed fixed timetables, then carried into a system with demand-responsive routes. Experience remains useful, but its relevance has to be examined. The choice is not between respecting expertise and discarding it; it is between using experience with an account of its limits and treating it as universally transferable.
Even an excellent probability estimate does not determine an action. Two organisers who agree about the likelihood of rain may rationally make different plans for outdoor events. One can hire inexpensive cover; the other must decide whether to cancel a performance whose equipment cannot be protected. Their decisions reflect different consequences as well as the same forecast. Disagreement over action need not indicate disagreement over evidence, and agreement over a probability need not eliminate a dispute about priorities. A decision process should therefore distinguish the assessment of what may happen from the judgement about which risks are worth accepting.
Feedback is most informative when the original reasoning remains inspectable. Keeping a dated record of assumptions and revisions makes it harder to reinterpret a forecast after the result is known. It also prevents a rare event from automatically being treated as proof of incompetence: a low probability is not a prediction of impossibility. One outcome can reveal a neglected mechanism, but calibration requires a pattern rather than a verdict based on a memorable case. The record should show what information was available at each stage, so that a good result does not excuse a poor process and a bad result does not automatically condemn a reasonable one.
These distinctions suggest a less theatrical account of authority. I would place more trust in an expert who can state the conditions under which a conclusion would change than in one who offers certainty regardless of circumstance. This is not a preference for perpetual hesitation. Decisions still have to be made, often before all useful information arrives. The aim is to make a judgement whose confidence, reasoning and consequences can be examined separately. Expertise earns trust by making its limits usable, not by performing an absence of limits. A credible expert can commit to a decision while remaining answerable for the assumptions behind it.