Leslie Valiant, The Importance of Being Educable: A New Theory of Human Uniqueness (Princeton University Press, 2024).

Leslie Valiant begins his book with an extraordinary claim: “We live in an intelligence-driven culture.” (xiii) By this, he means that intelligence is everywhere in our vocabulary. We seek to measure it — as in an Intelligence Quotient—we seek to simulate it in machines—calling it Artificial Intelligence. We even look for evidence of it in space, as when we search for Intelligent Life.  “This is all even though we are as far as ever from a widely accepted definition of what intelligence is,’” he concludes. “The evidence suggests that intelligence, much investigated as it has been, is an ill-conceived notion.”

And so, Valiant offers an alternative notion—educability–which he defines as “the capability to learn and acquire belief systems from one’s own experience and from others, and to apply these to new situations.”  A belief system is any body of structured knowledge and educability includes the capacity to incorporate that system of beliefs. So when one learns chemistry from a textbook or accepts that the Earth is billions of years old based on scientific consensus, one is operating through a belief system. Valiant argues that it is educability that distinguishes humans from other animals, and that this capacity serves as a “civilization enabler,” referring to the foundational role educability has played in making complex human civilization possible in the first place.

I am drawn to two important implications that stem from this concept. The first is the idea of “state change” as a way of describing the quality of an education. “I regard both educability and education as phenomena of computation,” the Turing Award-winning computer scientist begins.

“In the course of education, information is presented, whether as the description of a specific situation or as an explicit description of a general belief. The result of the presentation will be to make a difference in the student’s subsequent behavior as compared with the past. The change in behavior will be attributable to a change of the state of the student, realized as some physical change in the brain that persists for some time. Computation is about changes of state that can be realized by step-by-step processes. Physical systems also change state–if you boil water, there is a change of state. In a computer or a brain, there is extreme flexibility in the realizable state changes and in their possible effects.” (10)

This might seem overly technical, but I think it nevertheless an intriguing description of what happens when we say someone is “educated” or has been educated.  That is, an educated person has been cognitively and intellectually changed in some fundamental way.  When we describe an educational experience as “transformative”—as opposed to merely transactional—I think this state change is what we might mean.  That the student is now a different person, as opposed to being the same person plus the addition of new information or new skills.  Education is a cognitive state change, which should be the ultimate learning outcome of any course in a curriculum.

One practical consequence of this idea of education as state change is that we might reimagine what a student grade is supposed to reflect. Perhaps grading becomes the assessment of the degree of “state change” that has occurred in the student.  An A for a class means a student has achieved “complete state change;” a C means the student has only partial state change.  I will leave it to others to determine how “state change” is to be measured—if that is even possible– but the metaphor is nevertheless a good one.

In foregrounding educability, Valiant wades into the debate between education versus training.  “In my view,” he begins,

“a critical feature of education is that it can impart knowledge that will be useful later in ways not foreseen at the time of the imparting. In this sense, education is different from training, which imparts the skill to perform a task that is foreseen at the time of training. Indeed, training sessions often consist of instances of actions that exemplify the task. Education will have elements of training. The question is, what else is there in education that goes beyond training?”

He answers his own question by saying “The educability notion offers some clarification on this. It emphasizes that a system of beliefs is being acquired, and not just a single belief…[further] being educable means that one can combine pieces of knowledge gained years apart, decades later.” (37)

Remember when the Lumina Foundation proposed a “Degree Qualifications Profile,” a framework or rubric that would identify and standardize “what a student should know and be able to do to earn the associate, bachelor’s or master’s degree.” What if we made “state change” the new measure of the quality of a degree?  Lumina wanted to quantify what a student learned in their four years at university. An assessment based on state change would determine the degree of transformation the student had undergone.

From this, we might develop a new rating system for universities, to supplant the US News rankings.  This rating system would assess the strength of state change induced in graduates of that university. A new Carnegie classification might distinguish between post-secondary institutions that promise education and state change versus those that emphasis training.

The other interesting question Valiant’s book raises is about the future of artificial intelligence.  The focus of AI research has been, obviously, on the simulation of intelligence.  But as Valiant argues, we should not be seeking to have machines mimic intelligence but rather become educable: “artificial educability” he calls it. This means building machines that can do what educated humans do: not just narrow task-learning, but the full package of absorbing, generalizing and applying knowledge across diverse contexts. He proposes “to amplify the reach of machines in performing cognitive tasks by adding aspects of educability into such systems.” (165)

“Artificial educability” is distinct from what he sees as the current state of AI. When people say “artificial intelligence” today, they largely mean large language models and things that manipulate language and images. An artificially educable system would need to do something far more ambitious.

To achieve artificial educability would mean paying particular attention to teaching materials: 

“One can expect that the creation of educable technology will require a constant effort to develop new data sets or new teaching materials to fit the evolving requirements. This would complement the enormous efforts constantly made to keep up the steady expansion of knowledge when teaching humans. The challenge of continuously creating new teaching materials for machines will persist, I expect, long after the current questions about learning and reasoning algorithms are largely settled.” (184)

I can imagine a future scenario where a teacher is defined as someone who educates–provokes a state change—in an artificially-educable system. The methods we use today to educate young humans might work on artificially-intelligent systems. Schools of education might welcome cybernetic students into their programs.  But it is just as likely that we will need to develop new teaching methods to induce state change in AI systems. Instead of pedagogy, we might call this practice synthopaedy.

David J. Staley is a higher education philosopher, strategist, futurist, theorist and designer, and is an Honorary Faculty Fellow in Innovation at CHELIP.  He is the author of The AI Symposium; Anticipatory Biographies: Personal Histories of the Future;Alternative Universities: Speculative Design for Innovation in Higher Education and the co-author of Knowledge Towns: Colleges and Universities as Talent Magnets.