Fact, Inference, and Opinion

The three-way sort

Take any sentence in a public argument and it will usually fall into one of four categories.

Observation. Something recorded or directly witnessed. “The building’s fire alarm sounded at 9:14 pm.” Observations can be wrong, because instruments fail and witnesses misremember, but they are the kind of thing that could in principle be checked against a record.

Inference. A conclusion drawn from observations. “The alarm sounded because someone pulled it deliberately.” Nothing was observed about intent. Intent was inferred from the alarm plus something else, and that something else is where the argument lives.

Assumption. Something treated as true without being stated or checked, usually because it seems obvious. In the example, the inference may rest on an unstated assumption that this alarm type cannot be triggered by steam.

Value judgment. A statement about what is good, right, or important. “Pulling a fire alarm as a prank is unacceptable.” No amount of additional data settles it, because it is not a claim about the world.

The sort is useful because these four categories fail in different ways and are repaired in different ways.

Why mixing them causes trouble

When observation and inference are presented in one breath, the confidence attached to the observation transfers to the inference. “The alarm sounded at 9:14 and someone pulled it” reads as a single fact, and a listener who accepts the first half tends to accept the second.

When a value judgment is presented as a fact, disagreement becomes unresolvable. Two people who agree completely about what happened can still disagree about whether it was acceptable, and they will keep exchanging evidence that cannot settle the question.

When an assumption is left unstated, it cannot be examined. Most bad inferences are not bad reasoning from stated premises. They are fine reasoning from a premise nobody noticed.

Working through a paragraph

Consider this passage:

Attendance at the community centre dropped by a third after the new parking fee started. The fee is clearly driving people away. Charging for parking at a public facility is wrong.

Sorting it:

• Observation: attendance dropped by roughly a third after the fee began, assuming the attendance count is accurate.
• Inference: the fee caused the drop. This is a causal claim built on sequence. Something else changed in the same period in most real cases.
• Assumption: that the attendance measure is comparable across the two periods, and that nothing else relevant changed, such as a season, a schedule, or a competing venue.
• Value judgment: that charging for parking at a public facility is wrong.

Now the disagreement becomes tractable. Someone who disputes the causal inference is not denying the attendance drop. Someone who disputes the value judgment is not denying either. Naming the level lets each be argued on its own terms.

Degrees within inference

Not all inferences are equally speculative, and it is worth having language for the difference.

• Near certain. Many independent observations point the same way and no plausible alternative explains them. “The alarm was triggered on the third floor” when three separate systems logged the same panel.
• Reasonable. The inference is the best available explanation, but at least one alternative has not been ruled out.
• Speculative. The inference is consistent with the evidence but so are several others.

Saying which of these you mean is more informative than saying “I think” or “obviously”. It also makes it possible for someone to change your mind by ruling out an alternative, which is the only way disagreements actually resolve.

A note on the word opinion

In everyday use, “opinion” covers two very different things. It covers value judgments, which are not the kind of claim evidence settles. It also covers uncertain factual estimates, as in “in my opinion the fee caused it”, which is a claim about the world that evidence absolutely does bear on.

Collapsing the two lets a contestable factual claim hide behind the protection we normally give to taste. When you hear “that is just my opinion”, it is worth asking which kind is meant.

Practical exercise

Take a single paragraph from an editorial, a company announcement, or a social post. Copy it out with four labels in the margin, and mark every sentence as observation, inference, assumption, or value judgment. Sentences that carry more than one label should be split.

Then answer one question: if you disagree with the passage, which label is your disagreement attached to?

Most people find that their disagreement lives one level away from where the argument is being conducted.

Summary checklist

• I separated what was observed from what was concluded.
• I stated the unstated assumption the inference depends on.
• I identified any value judgments and did not treat them as factual disputes.
• I described my inference as near certain, reasonable, or speculative.
• I know which level my disagreement with the passage sits at.

No external references were required for this page. The distinctions described here are standard in analytic tradecraft and are illustrated with constructed examples rather than reported events.

Next step

Sort real statements in the Think Lab

Topics covered: Critical Thinking