Abduction: How AI Thinks: Like We Do!

By Shlomo Maital  

       How does AI think?   Until now, we haven’t really understood this;  AI is often referred to as a ‘black box’   even by its inventors and developers.

       Now, in a New York Times article (Dec. 16),  Peter Coy explains some research, in which researchers did a brilliant and rather obvious thing – They asked AI to tell us how it is thinking!   And here is the result:  

        AI thinks as we do.  By abductive reasoning!

       What is that?   So, naturally, I asked ChatGPT to explain it:

      “Abductive reasoning  is a logical process in which an explanation is inferred from the available evidence, even though it may not be the only possible explanation. It’s often described as “inference to the best explanation” because it seeks to identify the most likely cause or reason for a set of observations, even if all the facts are not known.

       “Unlike deductive reasoning, which moves from general principles to specific conclusions (certain), and inductive reasoning, which involves drawing broad generalizations from specific instances (probable), abductive reasoning  aims to find the most plausible explanation given incomplete information.

        In other words:  Hey? Don’t know the answer?!  Well – take a shot at it.  Take a guess.

        ChatGPT’s example of abductive reasoning:

       You walk into a room and see a puddle of water on the floor. Possible explanations (hypotheses): 1. Someone spilled water  2. The roof is leaking. 3. The window was left open and rain came in.  Using abductive reasoning, you consider the available evidence (the puddle of water) and evaluate the most plausible cause. If you know that there was a heavy rainstorm earlier, and the window is slightly open, you might conclude that the rain came in through the open window, even though you can’t see the exact cause. Based on the best explanation, you might infer that the open window is the most likely reason for the puddle of water, even though there could be other potential causes.”

      AI is often lambasted, criticized, because it fantasizes – makes wild guesses when it doesn’t know the answer.  Well, so do we humans!   Abductive reasoning, connecting the dots that seem unconnected, often generates powerful innovative ideas.  AI makes inferences, sometimes wild ones.  Programmers are striving to ‘cleanse’ AI from this – when it may be one of its most powerful advantages!

     So —  the one thing we thought AI lacked, creativity, is actually something it does have.  Abduction, linking X with Y. 

      Holy smoke!