AI可以批改作业，却教不会人类独立思考

Will artificial intelligence damage the ability of humans to think critically? This has become an urgent question for those involved in education, from early-years teachers to trainers of experienced working people. For universities, the question is existential. Teaching students how to think, and why that thinking is important, is core to a university’s mission and purpose. Many research leaders, potential students, parents and carers are asking if there is any point in working towards a qualification if someone can answer a complex question or write an essay using a few chatbot prompts. The same question arises when considering that teachers can use AI tools to mark assignments without engaging with content.

We tackle this and other questions in a News Feature, which reports on the literature about AI and critical thinking. With studies at an early stage, it is too soon to draw definitive conclusions. But AI use is widespread, and teachers are already reporting signs that chatbots are reducing opportunities for thinking and reasoning.

Universities around the world are deep in preparation for what they expect to be a revolution in teaching, learning and assessment, perhaps even greater than the introduction of personal computers, e-mail and the Internet.

As part of these preparations, last month, the Massachusetts Institute of Technology (MIT) in Cambridge announced the results of a review of AI use by its staff and students (see go.nature.com/4h4ajx6). “Not surprisingly, students told us the temptation was greatest when they feared they would miss a deadline,” its report says. It adds that there will need to be “substantial adaptation to survive an encounter with AI without serious disruption”.

Universities are starting to incorporate AI tools into learning and teaching, and using them for tasks that were previously done by humans. For instance, chatbots are answering questions that a student might ask university administrative staff in their first week. A few universities are experimenting with AI-generated lectures.

AI tools are also providing feedback on coursework submissions. According to researcher Thomas Corbin, who studies assessment and digital learning at Deakin University in Melbourne, Australia, one of the benefits that students report is that they don’t have to wait for feedback — it is instant and available around the clock.

What is essential while this transition happens, as the MIT report emphasizes, is to retain the human teacher’s guiding role in showing students how to learn to think as they are using AI. That means teachers still need to be up to speed with their students’ progress, even when feedback is outsourced to AI tools. There can be no shortcuts to this, as the Nobel-prizewinning economist Philippe Aghion said earlier this month at a Nature Conference in Paris on AI in health care: “We need schools in which we learn to learn.”

Although the term critical thinking is used in everyday language, researchers across disciplines conceive of it in different ways. Those who study cognition and its relationship to education describe the high-level processes that underpin critical thinking, such as memory, reasoning, decision-making and problem-solving. Each involves different but overlapping brain mechanisms. For social scientists who study the processes of teaching and learning, critical thinking includes the ability to weigh up evidence, to consider trade-offs when making decisions and the capacity to consider someone else’s view when engaging in an argument.

Learning how to apply these techniques well demands both deep subject knowledge and hours of practice at tasks and exercises. Crucially, this includes learning from mistakes — what teachers call teachable moments. It can also be fun. Such feedback in a student–teacher relationship — the ‘light-bulb moment’ when a student recognizes that they have achieved something after a lot of effort — is irreplaceable.

The history of education is littered with examples of fears of falling standards among students, such as when pocket electronic calculators replaced logarithmic tables and when word processors replaced essays written by hand or on typewriters. Educators were also concerned about the impact of word-processed assignments on thinking.

One argument was that computers’ ability to correct errors meant that students would not invest time in thinking about what to write. These technologies didn’t end critical thinking, although some might argue it was diminished. In the same way, AI might pose less of a risk to this skill if the technology is managed well.

But some differences must be accounted for. Thinking requires focus. Smartphones and social media have made students much more distracted compared with previous generations, and there is evidence that this is contributing to a dip in standards (see go.nature.com/4hxpytx). At the same time, a student who uses a calculator or spreadsheet can be confident that the calculation will be correct — there might be human error, but the device or software itself can be trusted. We are not yet there with AI. Its results need to be interrogated carefully to check the sources and evidence used. This requires knowledge and practise in thinking.

AI is the next classroom revolution. For both students and their teachers, it will be necessary to make time for thinking and for learning how to think. Equally important is teaching how to think in exciting and engaging ways. This is where teachers can be at their best, at the heart of that process, guiding students at every step.