Analogy and the Roots of Creative Intelligence
Keith Holyoak has spent half a century studying how we recognize and use analogy — the ability to see relational similarities between things that, on the surface, seem unalike. His work has spanned cognitive psychology, neuroscience, linguistics, artificial intelligence, and even poetry, all in pursuit of understanding how analogical reasoning underpins human intelligence and creativity.
Holyoak’s new book, “The Human Edge,” takes a fresh look at analogy as a defining feature of human thought. In it, he argues that analogy doesn’t exist in isolation but is deeply intertwined with two other uniquely human capacities — language and the ability to understand other minds. Together, these abilities rely on specialized neural circuits that set us apart from other species.
In the conversation that follows, Holyoak offers insights into how analogical reasoning develops in children, its role in creative thinking, how our brains organize and access knowledge, and whether AI can ever truly master this powerful cognitive tool.
Let’s start at the beginning: What exactly is analogy, and why is it important?
When trying to make decisions or solve problems, a basic principle is that what works in one situation is likely to work in similar situations. But what makes two situations similar? Often, it’s simply that the objects involved look alike — after tasting a first mango and finding it to be delicious, it’s a good bet that the next mango, of similar shape and color, will also taste good. But sometimes the similarity is more subtle.
For example, there’s an old proverb, “While two dogs fight over a bone, a third runs away with it.” A similar situation might arise in the business world. Suppose two corporations aggressively compete for market share, so that both are weakened (perhaps by taking on too much debt), eventually enabling other rivals to prevail over both. Recognizing the similarity to the proverb is an example of reasoning by analogy: The two situations involve similar relations between things, even though the specific things involved (dogs and companies) aren’t at all alike. Understanding analogies widens our capacity to apply prior knowledge to novel situations and come up with ideas for effective action.
How does analogy relate to intelligence?
It turns out that analogical reasoning is a core component of what is being measured in traditional tests of intelligence. Decades ago, a research team led by the educational psychologist Richard Snow analyzed how people performed on a wide range of ability tests and found that performance is generally correlated across tests: If a person does well on one ability test, they tend to do well on others. Some of the very strongest correlations turn out to involve analogy tests. Even though analogy tests may focus on different areas (like verbal, mathematical, or spatial reasoning), they remain strongly correlated. This makes them excellent measures of what is termed fluid intelligence — the ability to reason about novel situations.
How does analogy tie into creativity?
While analogy problems on intelligence tests directly compare two situations, analogies can also guide more independent creative thinking. However, we often struggle to retrieve relevant analogies on our own.
There’s a classic experiment — the “radiation problem” — introduced in the early 20th century by the Gestalt psychologist Karl Duncker — that illustrates the problem well. A doctor needs to destroy a tumor using radiation, but high-intensity rays will damage surrounding healthy tissue. The crux of the problem is that it seems that the rays will have the same effect on the healthy tissue as on the tumor — high intensity will destroy both, low intensity neither. The key issue is to figure out how the rays can be made to selectively impact the tumor while sparing the surrounding tissue. Duncker recorded and analyzed people’s proposed solutions. He found (as did later investigators who used the same problem) that people often reach an impasse — basically become “stuck” without an idea that seems at all realistic.
The ability to see connections between problems drawn from very different domains is a hallmark of human intelligence.
A solution exists, of course — sending multiple low-intensity rays from different directions so they converge on the tumor — but most people don’t think of it immediately. However, as my colleague Mary Gick and I discovered, people were much more likely to arrive at this solution if they had previously read an unrelated story that got them thinking about the solution in a different way. In our experiment, we had participants read a story about a general who divides his army into small groups to attack a fortress from multiple directions, avoiding a trap set by the enemy. This story, though unrelated to medicine, provided a useful analogy that helped participants apply the same strategy to the radiation problem, allowing them to think outside the box and find the solution.
Of course, sometimes a person will spontaneously notice a connection between situations without someone else calling attention to the resemblance. For example, the earliest scientific theory still accepted today — the wave theory of sound — was proposed by a Roman engineer who noticed an analogy between the behavior of sound and water. Sound rebounds off barriers, creating an echo, much as water waves rebound off the shore. Thinking by analogy can foster creative thinking when something puzzling triggers the recognition of a similar situation that is easier to understand.
At what point do we acquire the ability to think by analogy?
We acquire the capacity to reason by analogy in the course of typical cognitive development. Before they are taught anything about analogy in school, young children already appreciate relational similarities, and can use analogies to solve problems. Experimental tests indicate that children begin to think analogically at around three years of age, and by around age five or six this capacity is well established. Analogical thinking emerges quite early and without explicit instruction.
Nonetheless, in the book I argue this capacity depends on what I call a “Late System” for general reasoning. This Late System is a recent evolutionary development that appears to be specific to humans, first appearing in children at about age three and continuing to develop until adulthood. In contrast, several Early Systems, which humans share with various species of nonhuman animals, are already available to infants. These Early Systems allow infants to approximate relational thinking in various specialized domains, such as understanding cause-effect relations, or what other people are thinking.
Is analogy a unique property of human cognition, or is it shared with the minds of some nonhuman animals?
Many nonhuman animals — chimpanzees, dogs and dolphins, and birds such as ravens and parrots — are capable of performing tasks for which we humans are likely to reason by analogy. However, the successes of nonhuman animals appear to depend on a variety of specialized Early Systems, each adapted to a limited domain over the course of evolution. Humans also make use of Early Systems, but what’s special about our thinking is the Late System that allows us to think about relations in any domain.
In a study modeled after our experiments with adults solving the radiation problem by analogy, my students and I showed children at various ages illustrated stories, and then gave them a problem to solve. In one version, the story described how a genie wanted to move his jewels from a near bottle to a distant one, which he accomplished by rolling his magic carpet into a tube, stretching it from one bottle to the other, and then rolling the jewels through it.
The children were then asked to use any of various physical objects that were provided (including scissors, string, tape, and a large sheet of paper) to move some balls from one bowl to another (where the bowls were too far apart for the child to reach from one to the other). In this case, the analogous solution was to roll the paper into a tube (the way the genie had rolled his carpet), and then send the balls through the tube. One precocious four-year-old came up with this “rolled paper” solution after hearing the genie story, without a hint. Older children (aged 10 to 12) almost invariably succeeded in finding the analogical solution, at least after a hint was given — the same qualitative pattern observed for adults solving the radiation problem. This ability to see connections between problems drawn from very different domains is a hallmark of human intelligence.
Unfortunately, the special intellectual abilities of humans do not guarantee either wisdom or morality. Uniquely among all the animals on earth, human have invented both cures for diseases and weapons of mass destruction. If one day the earth is left uninhabitable for humans, it won’t be the fault of the chimpanzees.
There’s a fascinating chapter in your last book on AI and poetry, in which you consider whether artificial intelligence will eventually produce authentic creative work. Do you think AI has the potential to reach or exceed human analogical ability?
Recent advances in AI, particularly large language models (LLMs), have already created systems that can solve novel analogy problems at roughly the same level as college students. As the cognitive scientist Tom Griffiths has pointed out, humans operate under fundamental limits of time, computation, and communication. We are mortal beings, granted only finite time to learn anything. The human brain is complex and powerful, but we each are allotted just one of them, so our computational capacity is bounded. And we have no way to directly transfer the contents of our brain to that of anyone else — our shared knowledge has to pass through the bottleneck of our limited communication ability.
In contrast, the “lifespan” of an AI program is indefinitely long, during which the program can be revised through learning. Meanwhile, computing power is continually increasing as engineers build faster machines with greater memory and processing capacity. Programs can easily be copied from one machine to another, and huge networks of computers can be put to work on a shared problem. It is certainly possible that AI will create new forms of superhuman intelligence.
But to answer your question, while AI programs can generate novel and useful products, contributing to creativity, so far, these systems lack full autonomy. Although they can solve analogies posed to them, the initial analogical spark — noticing spontaneously that a certain known analog might illuminate a new problem in a way that advances knowledge — has yet to be struck by an AI.
Particularly in the case of artistic creativity, AI faces what may well be an insurmountable limit. In most forms of art — perhaps most obviously music and lyric poetry — the essential point is for the creator to convey an emotional experience to their audience through the medium of their artistic creation. An AI has none of this — no emotions, no consciousness, and no individuality. In fact, the general complaint about AI-generated products, whether in the form of writing, music, or visual art, is that they have no soul — no expression of their own emotions or of an individual point of view.
You’ve been studying analogy for almost 50 years. What have been the biggest surprises in your research?
One big one, thanks to contributions from many labs over the last quarter century, is that separate brain systems support different aspects of high-level human cognition. In the book I talk about a trinity of capacities that set humans apart from other animals: thinking about relations, communicating by language, and understanding other minds. These capacities all support one another, yet each depends on distinct neural pathways.
I was also surprised to find that although human cognition is based on a Late System that enables complex relational reasoning, this system was built upon — and now cooperates with — the Early Systems that reflect our vast biological heritage. This integration of abstract thinking with our animal nature makes humans special.
Keith J. Holyoak is Distinguished Professor of Psychology at the University of California, Los Angeles. A recipient of a Guggenheim Fellowship and the Warren Medal from the Society of Experimental Psychologists, he is the author of several books in cognitive science, including “The Human Edge,” as well as four volumes of poetry and a book of translated classical Chinese poetry. “The Human Edge” is available in a print and freely downloadable open access edition.
