What if we are misquantifying intelligence? In an era obsessed with measuring intelligence quotients (IQs), British computer scientist and Harvard professor Leslie Valiant sets out to demonstrate a collective error. “We live in a culture focused on intelligence,” he writes in the preamble to his book. […] But the evidence suggests that intelligence, as studied as it may have been, is a poorly understood concept.” The numbers that obsess your HPI (high potential intellectual) friends may not ultimately be the best way to know if the person you’re talking to is actually smart.
Leslie Valiant offers an alternative to this collective passion, replacing the notion of IQ with that of “educability”, or “the ability to learn and acquire belief systems from our own experience and that of others, in order to apply them to new situations”. These “belief systems”, as Valiant defines them, can belong to “scientific or political theory, religion, superstition, or fiction”. According to him, educability is unique to humans, allowing us to differentiate ourselves from animals. The singularity of the human species would be based on this concept. Interview with a scientist who does not necessarily want to make you more intelligent… but more “educable”.
L’Express: In your book, you designate the concept of “educability” as being a “civilization enabler“, a “facilitator of civilization”. Can you explain how?
Leslie Valiant: I come from the field of computer science, where we are used to developing strict reasoning leading to precise conclusions. So I tried to analyze the way humanity thinks in light of what we do with computers, by asking myself a question that we struggle to answer: what differentiates humans from animals? Research has shown that some animals demonstrate remarkable intelligence, such as crows, or fabulous memory, such as squirrels. Neither intelligence nor memory are therefore unique to humans. It is therefore impossible to have a single answer to this question, because the latter is rather an amalgam of concepts. Human thought and intelligence, unlike that of an animal or a computer, use combinations of skills that are specific to them. To answer this, I wanted to propose a concept: educability.
This covers three criteria. To be “educable” – and therefore reason like a Man – you must first be able to learn from your own experience – which is not an obvious thing. Then, you must be able to be instructed by a third person: be able to be focused enough in a classroom or a conference, absorb the data and retain it. Finally, be able to extract the things learned and apply them to new situations or experiences. In conclusion, be able to understand how to solve a new problem using your knowledge and previous experience.
What neither animals nor machines can do?
Separately, they are obviously capable of this. It is possible to teach animals things – they meet the second criterion. In computer science, machine learning is mainly focused on examples: how the software reacts when confronted with a given case. That is why algorithms are fed with a lot of data. But in both cases, it is difficult to incorporate the two missing pillars. It is impossible for a dog or a machine to reason by integrating three dimensions – say mathematics, psychology, past experiences – to solve a problem.
In your book, you explain that these three criteria create “belief systems” that facilitate learning (in science, in history, in fiction, in religion). Can you explain why they seem indispensable?
I use the term in a very general sense. Everything you learn is inserted into your belief system. Your belief system is a way of putting the information you’ve learned into a general context and making sense of it. But there’s a problem with it: there’s nothing in our brains that tells us whether what we’re learning—even though it may seem logical—is true, useful, or right. To go back to what we were saying earlier, we’re not like animals: in nature, they can simply rely on their experience to solve a problem that has already happened and will happen again in familiar circumstances.
Humans learn mostly through their peers. We have no real way to evaluate in the moment the veracity of what our instructor tells us – whoever he may be. Humans also have a natural tendency to trust the one who instructs them. School and the educational system work for this reason. But the downside is that our belief system can easily be corrupted by false information, as long as it is presented by a third party who gives a feeling of expertise or authority.
At a time when trust in traditional sources of information is eroding – in France, more than 57% of French people believe that we should be wary of what the media say – what can we do?
In my book, I insist on the fact that we must see “educability” as a strength, but also as our own weakness. We must insist on the way in which these examples influence us, and not underestimate how vulnerable we are to them – which we do, alas. Let’s take the example of a computer: despite what could be considered weaknesses compared to human intelligence, software is an autonomous system, which learns by itself. This is not the case for humans, who, by their need for otherness when learning, are vulnerable. The first step to limit the damage is to become aware of it. School is probably the best place to do this. The recipe is known, but the only good solution remains education, in particular by giving courses on the way in which propaganda has been used throughout history. The example nourishes thought.
In your book, you discuss how to quantify intelligence and the use of IQ tests. Why do you criticize them?
Historically, the question of intelligence has been explored mainly by psychologists. I mention the case of Charles Spearman who, in 1904, published an article attempting to measure intelligence “objectively”. To this end, he had collected data from children’s assessments and was intrigued by individuals with high results in very diverse subjects such as mathematics, French, Latin, etc. For him, these individuals had one thing in common, a “dominant intelligence factor” (called the “G factor”) on which all of the individuals’ cognitive abilities depended. Later, IQ tests would draw heavily on this G factor. But the latter presents a problem: what guarantees that these good results in all subjects were not a consequence of the children’s socio-economic background – wealthier parents being able to devote more resources to their children’s education? I think that thinking from another concept – that of educability – changes the situation, because we measure more the behavior of the person.
The concept I propose is just that at this point – a concept. Research needs to be done on how to accurately quantify an individual’s educability, through testing and a scientific process. The goal would be to measure skills through the process they use to solve a problem, not just the outcome: in an educability assessment, an individual would be presented with a new puzzle and would have to solve it with information they have already learned – not simply recite information.
Do you think that artificial intelligence is a new competition for human intelligence? At the end of your book, you seem to indicate the opposite.
Artificial intelligence is coming fast and changing a lot of things. I understand that it raises misunderstandings and concerns. But behind these apprehensions, there is above all a mystery that must be explained. It’s like chemistry: when you don’t know how to explain it, there is reason to be frightened by the interactions and explosions that it can cause. But once its functioning has been revealed, everything becomes more rational. Artificial intelligence is in the same situation: we must be pedagogical, and the public will be less afraid of what it can do.
*Leslie Valiant, The Importance of being educable, Princeton University Press, 2024
.