What kind of logic guesses the best explanation?

You step outside and see the grass is wet. It probably rained overnight, you think. But it could just as easily be dew, or a sprinkler that ran at dawn. That everyday leap, picking the single most plausible explanation for something you've observed, without being able to prove it, is exactly what logicians call abductive reasoning.

It sits alongside two more familiar kinds of logic. Deduction guarantees its conclusion: if Socrates is a man, and all men are mortal, then Socrates is mortal, no room for doubt. Induction generalises from repeated observation, though it can still be wrong later, the way seeing only white swans once supported 'all swans are white,' until black swans turned up. Abduction is different from both: it doesn't generalise and it doesn't guarantee anything, it just proposes the best available explanation for one specific, puzzling observation, like wet grass.

The American philosopher Charles Sanders Peirce developed this idea from the second half of the 19th century, and he was remarkably honest about its limits. In his own words, abduction is really just guessing, 'very little hampered' by strict logical rules, and even a well-trained mind gets its individual guesses wrong more often than right. What makes abduction useful anyway, Peirce argued, is that educated guessing still beats random luck by a wide margin, and it's the only method humans have for generating genuinely new ideas in the first place. In 1910 he even admitted that in almost everything he had printed before 1900 he had more or less mixed up hypothesis and induction.

The idea found a second life a century later. As computing power grew in the 1990s, researchers in law, computer science and artificial intelligence rediscovered abduction, and today diagnostic systems, the kind that guess what's wrong with a patient or a machine from a list of symptoms, routinely run on exactly this kind of reasoning.