The number sounds certain
A probability can make uncertainty sound remarkably authoritative. Say that an event is likely and someone may ask why. Attach a precise percentage and the same hesitation starts looking scientific. I find that transformation suspicious. Precision in the notation does not guarantee precision in the knowledge behind it.
What attracts me to probability is the opposite possibility. It can force uncertainty into the open. What information do I have? What assumptions connect that information to the event? Which possibilities am I ignoring because they are inconvenient to describe? The number becomes useful when its conditions become visible.
A small imaginary test
Imagine a fictional screening system used on ten thousand items. One percent of the items are defective. The system flags ninety percent of the defective items and also flags five percent of the sound items. It catches ninety defective items, but it also flags four hundred and ninety five sound ones. Among the flagged items, only about fifteen percent are actually defective.
Nothing mysterious happened. A small error rate applied to a large group produced more flags than a large detection rate applied to a small group. The base rate mattered. The example is deliberately simplified, but it shows how easily the phrase accurate system can hide the question a person actually needs answered.
I like this example because it corrects an intuition without humiliating intuition in general. The arithmetic is not telling me to stop thinking. It is telling me which part of the situation my first impression left out. A good model can become a disciplined way of noticing.
Update without pretending
The harder lesson is emotional. Revising a belief can feel like admitting that the earlier version of myself was foolish. Probability offers another interpretation. A belief held with limited information should change when the information changes. Refusing to update is not loyalty to reason. It is loyalty to an old snapshot of what I knew.
That does not mean every surprising observation deserves a dramatic reversal. Evidence has quality, selection matters, and the process that produced the observation may be part of the explanation. Humility includes being cautious about a correction that feels too satisfying.
I want to use probability as a language for responsible uncertainty. State the assumptions. Check the arithmetic. Ask what evidence would move the estimate. Leave room for the possibility that the model itself needs revision. A percentage should make it easier to discuss what I do not know, not give me a more impressive way to conceal it.