AI and education: Embracing the disruption

AI’s capacity to transform the education industry will be multi-faceted (encompassing teaching, learning, and administrative functions) and applied across schools, universities, academia, and businesses through professional education (see figure 3). The technology could also contribute to improved accessibility, for example, by providing high-quality education to lower-income or disadvantaged students, offering new possibilities to reduce inequality and improve living standards among educationally-underserved sections of society.

The use of machine learning and deep learning techniques in education isn’t new. Adaptive exams (e.g., tests that adjust to a student's capabilities, with correct answers prompting harder questions) and models that automate the reviewing and grading of students’ work have existed for years. However, the rapid evolution of natural language processing in the form of large language models (LLMs), and, more recently, in multi-modal generative AI models (which process information in different formats, such as audio, image, text, and video) brings new opportunities. 

One opportunity that is already in use is teacher augmentation, a human-centered innovation that provides teachers with expert guidance and consultation from AI tools (see the Appendix for an example of a learning path created with AI and using revised Bloom's Taxonomy, a learning framework developed in the 1950s by educational psychologist Benjamin Bloom). A recent study by Stanford University found that AI assistance for less-skilled tutors could improve students' assessment outcomes by nine percentage points. In that example, tutors received AI-generated questions and explanations to address student mistakes.

Stepping into the future, a multi-modal AI tutor/copilot might incorporate information on a student’s academic results, behavior, and learning needs, with the information coming from videos, Internet of Things (IoT) devices that measure quality of sleep and health metrics, and even brain-computer interfaces. Further autonomy could be offered by systems that utilize agentic AI (which offers the capacity to perceive an environment, act autonomously within defined parameters, and learn). With that technology, an AI tutor could act only when a need is identified and develop a personalized learning path based on experience with the student. Whether that sort of AI autonomy is ethical, appropriate, and safe will be a key consideration and challenge for regulators, policymakers, and parents. 

While governments are studying the potential rewards of using AI, they are also grappling with the potential risks of its applications in teaching and learning, including ethical and legal issues.

In Europe, the EU AI Act expresses the need to classify certain education-related AI models as high risk because they may determine the professional course of a person’s life or their ability to secure a livelihood. Such AI models are thus deemed to require closer supervision and greater transparency. Examples include AI systems that determine access to educational or training institutions, models that evaluate individual learning outcomes, and those that assess the appropriate level of education for a professional role.

In the U.S., the regulatory approach is less prescriptive and not dictated by comprehensive federal AI regulation. However, the White House Office of Science and Technology Policy's "Blueprint for an AI Bill of Rights" notes that enhanced protections and restrictions for sensitive domains such as education and data pertaining to youth are a priority, and that continuous surveillance and monitoring should not be used in education. A U.S. Department of Education report also highlighted concerns about algorithmic bias and resultant discrimination with regard to surveillance and monitoring, admissions processes, and disciplinary interventions.

In China, the regulatory approach is proactive and seeks to ensure alignment with government policy and oversight, set out in the "Interim Measures for the Management of Generative Artificial Intelligence Services." Existing regulations provide guardrails for AI education algorithms to be aligned with "socialist values," ensure students’ data privacy, and maintain transparency and fairness, particularly when relating to recommendation algorithms. Additionally, the Ministry of Education actively promotes AI's integration into education systems.

More broadly, we observe a trend toward governments incorporating guidance from transnational organizations and standards-setting institutions (some examples are highlighted in table 1). While these frameworks provide valuable guidance, we believe there is a growing need for more specific and practical recommendations tailored to the unique needs of educational institutions, which are subject to significant public scrutiny and may be hesitant to adopt or experiment with AI technologies without clear and shared guidelines.