I have argued for a while that the promotion of understanding is as important for the information sciences as the communication of information and the sharing of knowledge; see an earlier post on this idea.
One of the difficulties in discussing this topic is the lack of clarity as to what exactly it means to understand something. It is often suggested that to understand is to be able to do something well, or to explain something to someone else. It seems sensible, and is certainly usual, from the perspective of the information sciences, to equate understanding to a conscious dealing with knowledge. But does understanding necessarily require human consciousness? We might well say that we ‘understand’ how to walk, or to process inputs from the visual system? We do these things reliably and effectively, though we are not consciously aware of how we do them. Other living creatures, including plants and bacteria, may also be said to have an unconscious understanding of how do to perform very diverse and sophisticated actions. The issue of unconscious understanding has been brought into sharp focus by advances in artificial intelligence. Take, for example, the AI chess playing programs, which can beat the best human players. Do we believe that they ‘understand’ chess, since they can play it so well? If they do, then their understanding seems to be different from that of a human; their style of play has been described as “alien … chess from another dimension”. Such systems are certainly very good at what they do, but their ‘understanding’ of their domain, rooted in very complex networks of statistical data, cannot possibly be explained to a person.
An interesting take on this issue is given by Nobel-laureate physicist Frank Wilczek in his popular science book Fundamentals (Penguin, 2021), which examines ten foundational principles of physics. Wilczek suggests (pages 220-221) that machine learning yields a form of unconscious understanding which is complementary to human comprehensibility. ‘Complementary’ is here used in the very specific sense first set out by Niels Bohr, the Danish physicist, and pioneer of quantum mechanics. Bohr noted that certain feature of the world, such as the position and momentum of a body, could not be determined simultaneously in the quantum framework. He described these as complementary, meaning that that they were equally valid, but alternative, ways of describing the physical world. He developed this beyond physics, into the idea that aspects of the world, including human experience, may be captured in distinct, unrelatable, complementary ways. Wilczek suggests that the capability of chess-playing AIs and human comprehension of the principles of chess, which we would usually regard as understanding the game, are complementary. This seems, as Wilczek argues, a fruitful way to view the topic; that rather than trying to agree on what understanding ‘really means’, we can agree that there may be different, and radically different, meanings; complementary in Bohr’s sense.
Wilczek makes another interesting point (page 223). Understanding is relative. Modern science raises the bar for what we mean by true understanding so high that what we have achieved always seems inadequate. Hence John R. Pierce’s ironic statement that “we will never again understand nature as well as Greek philosophers did”. Clearly ‘understanding’ is n not a simple concept, rather a dynamic set of complementary ideas.