Quick Insight: Tom Chatfield on What Skills We Need in an AI Age
Philosopher Tom Chatfield, the author of widely read guide to critical thinking, discusses the role that artificial intelligence can play in helping human beings learn in this Quick Insight video. His message is that AI need not — and in fact must never — supplant human thinking but should enhance it. These are similar to ideas he fleshed out in a recent white paper. Among his most recent accomplishments, he designed Critical Thinking and Understanding AI business courses for The Economist and is co-creating an AI critical thinking “cognitive co-pilot” for universities.
A transcript of his Quick Insight appears below the video.
I’m Dr. Tom Chatfield, and I’m a writer and philosopher with a special interest in critical thinking, AI, and education, among other things. I write books for Sage exploring critical thinking, and I’m really interested in what skills we need as educators and citizens in an age of AI.
One of the places I like to start is with the incredible power of large language models and the large amounts of uncertainty that lies ahead. And it seems to me that we have tools which increasingly can answer any question we ask fluently and rapidly on the basis of vast amounts of data.
This is powerful but also seductive. And for me, it puts an enormous onus on two things. First of all, knowing which questions are worth asking in the first place. Secondly, knowing what it means to evaluate answers, what a good, a better or a worse answer looks like.
So how do we think about education in an age when it’s possible to simulate understanding of a topic without possessing it? And at the heart of this is the problem that learning cannot be outsourced. To learn something is inherently to engage in a transformative act combining skill and memory. It’s effortful. It is high friction. It’s about integration.
We know a great deal about the cognitive fundamentals of education that we want to scaffold concerns, that we want people to have knowledge and understanding, but then develop skill through the application and integration of these things, that we want people to use things like spaced repetition to consolidate knowledge. And fairly clearly, if someone is simply outsourcing their thinking to an algorithm, even if it’s excellent, they’re not learning in a meaningful sense.
And for me, there are, again, sort of two complementary aspects of education that I think we’re only going to need a greater and greater interest in when it comes to formulating the right questions. One thing I observe is that no matter how powerful an AI is, it will never be able to answer a whole host of questions about, for example, what you love or value or enjoy, what you did yesterday, why you get up in the morning, what someone wants to do with their life, how they feel about themselves as a citizen, as a colleague, as a student. These questions are powerful not in proportion to the speed with which we can answer them, but in proportion to the quality of the time we spend holding them and attending to them individually and together.
And when I’m working with students — all the way from kids in school, teenagers, up to post docs — I put a huge emphasis on both collective and individual reflection, Reframing the critical, creative and empathetic interrogation of questions. What’s worth asking? Why? What assumptions am I bringing?
And these human skills are, in a way, how we decide where we’re going to point or direct these hyper powerful tools at our disposal. In some ways, I think everybody is now becoming the head of a miniature research lab. They’ve got incredible inhuman colleagues and subordinates, so they can set off to do things and find things.
But as every researcher knows, dangers come with this terrain. If you set out to confirm a favored hypothesis, you will end up confirming something you want to be true, rather than something that is true. If you sift through vast amounts of data looking for things that seem interesting or significant, you’re hacking the research process rather than actually finding out what’s going on.
And so I think this interpersonal process of determining where you want to point your machines is incredibly important. And then similarly, I think the research process, the methods for which Sage among others, is known as a kind of leading advocate, is incredibly important for differentiating between an iterative, thoughtful process that makes the most of the tools, the individuals involved, or a confirmatory process that potentially embeds biases and shallow thinking at its center.
A research question ought to be something that we can, as it were, disprove rather than prove, that allows us to test an idea rather than merely confirm it. And similarly, the use of a powerful tool should involve a deepening and interrogation of data, of assumptions, rather than just a one-stop shop.
And so I guess to wrap all this up in something that’s practical, I hope, I would love to see academics and institutions and indeed researchers asking how they can, so to speak, double down on the bits you do before and around the questions you want to interrogate.
And this includes the metacognitive skills that people only build by exercising their minds individually and together, reflecting, sharing, listening, re-explaining, questioning and surfacing human assumptions and debating values and purposes. These processes are intrinsically important. They’re bound up with citizenship as well as ethical undertakings as a researcher, as a thinker, as a writer.
And so we do perhaps want AI to back off from these or to facilitate them, but to ensure that they are happening inside the human mind.
But then, while being agnostic to the particular tools we use, I think it’s incredibly interesting to ask how we can then encourage people to use all the tools at their disposal, but within a larger, rigorous, iterative investigative process, and to surface the evidence of that process. Whether it’s in journals, whether it’s by logging and reflecting, whether it’s by sharing prompts, whether it’s by publishing complete data sets or being more ambitious in outputs. Because this, of course, is the prize. It’s citizens, it’s researchers, it’s learners who go out into the world and have what we might call cognitive sovereignty. The ability to reflect upon their own thoughts, to change their minds, to collaborate with others, to define the purposes that are precious to them and why. And then to use tools not as magical boxes that deliver answers, but as engines of investigation, testing, action, and iteration.
The output of an LLM is not just a final answer; it’s evidence of an algorithmic process that is inhuman and perhaps inscrutable, but amenable to the tools of thought, investigation and analysis. And for all the panic as well as the hype, I feel hopeful because we’re challenged by this moment to ask what it is we really value about education, what it is we are trying to assess. And if old proxies for understanding and achievement no longer work, that challenges us to find the signals, the values that remain important.
And most of all, I think the transformative processes that take place inside someone when they learn and reflect alongside others about what they want to do in the world and how they can try and bring it about.
