This Is What Human Expert Planning Looks Like (and It Isn’t Like AI)
A new study reveals how London cabbies plan routes.
Posted March 11, 2025 | Reviewed by Michelle Quirk
Now Kublai Khan no longer had to send Marco Polo on distant expeditions: he kept him playing endless games of chess. Knowledge of the empire was hidden in the pattern drawn by the angular shifts of the knight, by the diagonal passages opened by the bishop’s incursions, by the lumbering, cautious tread of the king and the humble pawn, by the inexorable ups and downs of every game. –Italo Calvino, Invisible Cities
Invisible Cities is one of my all-time-favourite books. In it, Marco Polo travels across Kublai Khan’s empire to bring back tales of various fantastical cities. The emperor decides that just by playing games of chess with the explorer, everything that could be learned about the land would be revealed. I find this passage to be closely connected to a problem that has plagued empirical psychology since its inception, that of the ecological validity of experimental settings.
Psychologists often have to resort to controllable tasks and settings as a stand-in for complex environments in the real world. To study human planning, for instance, researchers often resort to games such as Go, the Tower of Hanoi, or chess. This makes sense because studying planning “in the wild” is incredibly difficult. But how far can this analogy be stretched? How much is thinking about routes in a labyrinthine city like thinking about moves on the checkered board?
In a recent experiment, now out in PNAS , we found a way to go beyond the chess set and test planning in the real world. To do this, we relied on the incredible skills of London taxi drivers, celebrated for having mastered "the Knowledge of London." This deep knowledge of routes across 26,000 streets is unique in the world. Their very brain changes as a result of their training, and cabbies have famously been found to have a larger posterior hippocampus—the seat of spatial cognition —than the average person.
Hugo Spiers and Eva Griesbauer (from UCL) figured out that a key part of the cab driver training could be adapted to study their cognitive processes: Cab drivers have to be able to “call out” the streets and the turns on a route between any origin and destination in London. That cabbies are able to do it is by itself pretty incredible. But the reason it was interesting for us is that it opens a window into their minds. They pause before starting to call out streets, and they pause in between streets. How long their pauses are is a reliable indicator of the cognitive load involved in calling the next street on the route.
Dan McNamee, a computational neuroscientist at Champalinaud Foundation, was able to use the length of the pauses in the cabbies’ planning to model the cognitive processes that underlie how cabbies plan routes. What we found is that what they do is radically different from classical artificial intelligence (AI) algorithms about planning. The classic AI approach is a tree search algorithm, which involves representing trajectories and branching decision points. Each sampled trajectory is assigned a probability to lead to the goal, and the chosen route is the one maximising that probability. That is roughly what your satellite navigation will do when you ask it to find a route.
To make that work, the AI needs immense computational resources. In environments as big as a city, this type of approach clashes against what’s known as the curse of dimensionality: There are so many options that the computations become intractable. What taxi drivers do turns out to be much smarter, in that it is vastly more efficient. This is how Dan McNamee describes it: “They look at the entire network of streets, prioritising the most important junctions on the route first, using theoretical metrics to determine what is important. This is a highly efficient way of planning, and it is the first time that we are able to study it in action.”
There is a big discussion in philosophy about whether technologies can become extensions of our minds . Something that our study shows is that human minds work quite differently from AI. So, if we want to offload our route-planning to Google Maps, there is quite a lot that might be lost in the transfer. Instead, we can learn about the smart ways in which humans operate to make algorithms that are less resource-intensive, and we can focus on the ways in which AI can complement, rather than substitute, human cognition.
A. Newell, H. A. Simon, Human Problem Solving (Prentice-Hall, 1972).
Calvino, Invisible Cities (Houghton Mifflin Harcourt, 1978).
Fernandez Velasco, P., Griesbauer, E. M., Brunec, I. K., Morley, J., Manley, E., McNamee, D. C., & Spiers, H. J. (2025). Expert navigators deploy rational complexity–based decision precaching for large-scale real-world planning. Proceedings of the National Academy of Sciences, 122(4), e2407814122.
K. R. Allen et al. Using games to understand the mind. PsyArXiv Preprints. https://doi.org/10.31234/osf.io/hbsvj .
N. D. Daw, Y. Niv, P. Dayan, Uncertainty-based competition between prefrontal and dorsolateral striatal systems for behavioral control. Nat. Neurosci. 8, 1704–1711 (2005).
Q. J. M. Huys et al., Bonsai trees in your head: How the Pavlovian system sculpts goal-directed choices by pruning decision trees. PLOS Comput. Biol. 8, e1002410 (2012).
Jody Rosen. The Knowledge, London’s Legendary Taxi-Driver Test, Puts Up a Fight in the Age of GPS. New York Times. November 10, 2014.
Brain changes seen in cabbies who take 'The Knowledge.' BBC News. December 8, 2011.
Share this post Facebook Bluesky Linkedin Email
There was a problem adding your email address. Please try again.
By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy
Pablo Fernandez Velasco, Ph.D., is a British Academy postdoctoral fellow at the Department of Philosophy of the University of York, and an affiliated member of the Spiers Lab at the Institute of Behavioural Neuroscience of University College London. He works on philosophy of mind and cognitive science.
Get the help you need from a therapist near you–a FREE service from Psychology Today.
This article is part of the Bringwise Psychology Journal — daily insights on human behavior, mental health, and personal growth.