To get around such philosophical tangles, the researchers turned to Mechanical Turk, a website run by Amazon, a big tech firm, that allows people to post odd jobs. They recruited 2,046 human participants and asked them to rate 50 statements from a corpus of 4,407 claims that might plausibly be seen as commonsensical.

As common sense might have predicted, the researchers found that plainly worded claims concerning facts about the real world were the most likely to be rated as demonstrating common sense (“triangles have three sides”, for example, which is true by definition, or “avoid close contact with people who are ill”). The more abstract the claims, the less likely participants were to agree that they were common sense (“all human beings are created equal”; “perception is the only source of knowledge”).

When they split the claims by subject, the researchers found that those concerning technology and science were the most likely to be rated as commonsensical, while matters of history and philosophy were the least likely. A respondent’s age, sex, income and personal politics had little effect on what they thought counted as common sense, although psychological measures of social perceptiveness and the ability to reflect on one’s opinions did.

Having investigated individual opinions, the researchers looked at how common sense works across big groups. Here, they found much less agreement than might have been expected. Only around 44% of claims in the corpus were rated as commonsensical by at least 75% of respondents. A stricter definition of common sense, in which everyone has to agree with a claim for it to count, cut that number to just 6.6%. Where exactly a sensible cut-off lies is a matter for debate. But truly “common” sense, it seems, is an elusive thing.