I’m not talking about robot teachers

The recent achievements of artificial intelligence (AI) have been frequently amazing but all too often somewhat disturbing. On the one hand, some AI technologies are now capable of analysing medical images to predict early-onset breast cancer, kidney disease and diabetes better than human experts – raising exciting possibilities for the future of human health. On the other hand, some AI technologies have been responsible for automatically generating fake news and images, systematically discriminating against particular ethnic groups and women, or compromising democratic elections and referendums – raising serious concerns for the future of human society.

Of particular interest is that artificial intelligence has also quietly entered the classroom. Whether we welcome it or not, multimillion-dollar funded companies are increasingly selling their education-focused AI technologies into schools worldwide – despite limited evidence of effectiveness, and with little consultation or debate.

In fact, artificial intelligence impacts on education in three distinct but complementary ways, in each of which educators have a unique and crucial role to play: learning for AI (helping all members of society to live effectively in a world increasingly affected by AI), learning about AI (helping students of all ages learn the mathematics, statistics and coding that underpin AI technologies), and learning with AI (using AI technologies to support teaching and learning).

To date, the focus (driven mainly by technology rather than pedagogy) has been on learning with AI technologies that have been designed to automatically personalise instruction. With these so-called “intelligent tutoring systems”, the student engages with a screen-based learning activity. How the student responds (the choices they make, the things they click, the answers they give) determines the next learning activity. In other words, the system adapts the student’s pathway through the material to be learned according to their individual capabilities, with every student following their own personalised efficient route.

Like so much in AI, on the surface this is impressive. However, the underlying assumptions need to be scrutinised. A flawed metaphor that I still find useful involves school buses and Uber taxis. The idea is that the school bus represents classroom-based learning. Every student is on the same bus together, travelling along the same route, to the same destination. On the other hand, intelligent tutoring systems are the Uber taxis, taking each student along their own personalised route.

However, we don’t ride Ubers for personalised routes. We use them because they take us to exactly where we, as individuals, want to go. Similarly, for learning to be genuinely personalised, it should lead to personalised learning outcomes. It should give students agency over their own learning, and empower them to fulfil their personal ambitions and potential – to do what psychologists call “self-actualise”. Intelligent tutoring systems, mainly due to the knowledge-transfer instructional approach that they embody, achieve little of this.

Inevitably, contrary to much marketing spin, all use of AI in education brings significant challenges that need to be addressed. For example, AI-powered continuous assessment might usefully replace stop-and-test examinations but will also require constant student surveillance; the use of AI to detect and enhance student emotions might boost learning but also represents an unprecedented invasion of student privacy; and automated writing-evaluation might reduce teacher workloads but will also eliminate key opportunities for teachers to learn about their students’ capabilities. There are also the ethical issues of fairness, accountability, transparency and bias that need to be confronted.

Nonetheless, the argument is not that AI has no role to play in classrooms. Indeed, from my perspective, AI tools that are designed to support teachers, rather than to replace them, are to be welcomed. However, the widespread introduction of AI in education does need to be subjected to critical scrutiny, and educators need to engage in early discussions with the AI technologists and entrepreneurs – all to ensure that the use of AI in classrooms is genuinely beneficial, and that it always starts and ends with the learning.

Dr Wayne Holmes (@wayneholmes) is Principal Researcher (Education) at Nesta, the UK’s leading innovation foundation, and lead author of Artificial Intelligence in Education: Promise and Implications for Teaching and Learning (2019).