A university’s raw material is intelligence. But what is that? What is this putative thing that a professor has more of than a gardener, an ape, or a computer?
We think of students that get A grades as having more of it than those that get C grades. This is not an unreasonable idea, considering that academic achievement correlates with scores on standard intelligence tests. There are two ways to think of intelligence: first, we can see it as the ability to interact with the world, to solve problems by trial and error. Alternatively, intelligence can be thought of as verbal logical reasoning inside our mind, if-then paths cascading to the solution. For Alfred Binet, intelligence was mainly thinking - inside our heads. Binet, indeed, developed his intelligence test with the purpose of assessing students’ educational achievements and detecting students who would benefit from remedial teaching classes.
Differences in concepts
Throughout the ages, continental scholars (famous figures being René Descartes, Gottfried Leibnitz, Baruch Spinoza) have tended to think about human thinking as a top down, internal, mental activity of idea generation. Intelligence is the ‘faculty of judgment’ and ‘good sense’, as Binet put it. On the other side (of the channel), British philosophers (David Hume and Francis Bacon, for example) favored a more bottom up conception of cognition, incorporating empirical knowledge gathered from ‘evidence’ from the outside world. It is in this tradition that, almost a century ago, the English psychologist John Raven developed an intelligence test that was used by the army to select military personnel during the Second World War. Raven’s test requires the test-taker to identify the missing element that completes a visual pattern. It requires no verbal responses, though it can be solved by an internal verbal reasoning process.
Rationalism vs Empiricism
The contrast between rationalism and empiricism is one of those philosophical issues without any need for solution, like a pair of incompatible mindsets the western world inherited from the Celts and the Romans. It occasionally pops up in European politics or in everyday life. The United Kingdom Hydrographic Office, for example, in publishing tide tables indicating the water levels in every English sea port at every hour of every day, claims to provide the most reliable values possible because they are gathered by officers of the Service by actual observation. The same claim to reliability, however, is made by the French Service Hydrographique et Oceanographique de la Marine, which produces the equivalent tables for the French ports, precisely on the argument that the values are NOT based on actual observations but calculated by a model. It is in the smoking room of this philosophical stalemate that our cognitive experimental study tries to blow a breath of fresh air.
Assessing intellectual performance
Cognitive scientist Bruno Bocanegra (first author, Erasmus University Rotterdam) teamed up with authors at Leiden University and Edinburgh University and the mental health institution Rivierduinen in Leiden, respectively, to explore whether we could identify which type of intelligence – the internal rational or the external empirical – best corresponds to general intellectual performance.
We used the Raven Progressive Matrices task and designed a computerized click-and-drag version of it that allowed participants (students from Leiden) to dynamically manipulate and reconfigure the visual pattern continuously during the course of problem-solving. They could, for example, create a tentative solution, observe and interpret the result they saw, make adaptations if needed, and generate a new tentative solution. Another group of participants carried out the traditional static version of the test. They could only rely on their internal reasoning.
The process to success
We found that the more students went through the cycle of conjecture and observation, the more successful they were at solving these complex puzzles. Moreover, the click-and-drag test (which enabled students to use the empirical try-and-observe cycle) was a better predictor of their academic achievement (grades) than the static Raven test. Hence, ‘empiricist’ IQ tests based on an empirical view of intelligence, which allow people to offload information and put their thinking to the test of the environment (something which conventional IQ tests are not designed to take into account), seem to tell us more about general cognitive functioning than ‘rationalist’ IQ tests. Not just reasoning, but continuously checking our thinking against what is out there, makes us smart thinkers. Our experiments make a point in favor of the empiricists.
Back to big questions
Speculating about our results slightly further than might be acceptable to an editor of Nature Human Behaviour, we might see a philosophical point ensuing from our study. Indeed, besides directions for improving intelligence measurements, our results suggest a contribution to a longstanding smoking lounge controversy similar to the Continental-Island one, albeit in its 21st-century, updated version. That update involves the comparison between human and artificial intelligence. Some modern scholars – like Turing, Newell, and Simon – see little difference between artificial and human intelligence, which raises the question whether the human mind will soon be caught up by artificial thinkers. The question is no less fascinating than the contemporary search by physicists for quarks and other types of subatomic particles, and has led to equally ambitious and expensive scientific projects like the construction of an artificial brain. But what if the perfect mimicry of human intelligence by robots is not a question of time, but rather an essential impossibility? Can we specify the thing that makes human intelligence unique?
No artificial experimenters
Our successful participants reasoned by creating configurations with the purpose of challenging the outside world to come up with cues to direct their thinking. Crucially, these cues could not fully be predicted in advance: they were observed post hoc, and interpreted for use in the search for a solution. A crucial step in ‘external’ thinking was to elicit information from the outside world to help. The internal thinkers however, automatically harvested their solution from a stepwise internal argumentation. Undoubtedly, nowadays computers are quite good at the latter. But will they ever experiment creatively, forcing new information relevant to their problem to show up in their environment and using that information to solve the problem at hand? Will they ever solve the Raven task by tentatively moving figures, looking what happens, eliminating options, shifting around another figure, looking again what comes out of it, etcetera.
Rather than the ability to perform complex computations, it is perhaps the crucial ability to co-opt the environment that is a key feature of intelligent living brains, helping us to obtain high grades in school and advance technology and civilization. Our work gives no reason to hope – or to worry – that we might be able to outsource these achievements to artificial thinkers.
Blog written by Fenna Poletiek as a result of Bocanegra, B.R., Poletiek, F.H., Ftitache, B., & Clark, A. (2019). Intelligent Problem Solvers Externalize Cognitive Operations. Nature Human Behaviour (3), 136–142