Generative AI technologies are set to radically reshape the way we teach and learn over the coming years.

The year 2023 was a landmark in the use of generative AI (genAI) – a rapidly developing form of AI that is distinguished by its ability to create entirely new content.  In particular, the launch of major text-based genAI chatbots received a phenomenal reception.  In November, OpenAI’s ChatGPT reached 100 million users a week[1].  The ability of ChatGPT to quickly produce cogent written material on a wide range of topics, with a high level of accuracy – and to converse naturally and flexibly with humans – quickly captured the public’s imagination.  And it led to huge interest in how such technology could change the way we live and work.

ChatGPT is believed to have been the fastest-growing consumer internet app of all time, reaching 100 million monthly users within just two months of its launch.  And while it now faces competition from other offers such as Google’s Gemini (previously known as Bard), it remains the most popular genAI app.

There are also popular apps (such as Dall-E and Midjourney) that use genAI to create images from text, while others generate music and video.  GenAI technology is expected to have a major impact on economies and societies.  An analysis by Bloomberg Intelligence predicted the genAI market would grow from US$ 40 billion in 2022 to reach US$ 1.3 trillion over the next 10 years[2].  Meanwhile, a report by McKinsey found that 2023 was genAI’s “breakout year” in the business world.  In a survey in April 2023, almost a quarter of respondents (22%), from all levels of organizations, said they were already regularly using it in their work[3].

Education is one area where genAI is likely to make a significant difference.  While there have been some concerns about the possibility for it to enable plagiarism and spread misinformation, there is also exciting potential for it to create new value for both learners and educators.

Educational opportunities

When genAI first emerged, it prompted fears that learners might use it to cheat, such as by producing plagiarized work.  Some schools in the US even banned ChatGTP, citing “concerns about negative impacts on student learning, and concerns regarding the safety and accuracy of contents”[4].  There were similar concerns raised about the technology’s impact on higher education.  In March 2023, an academic paper discussed how AI tools raise concerns relating to academic honesty and plagiarism.  Only after the research was published was it revealed that the paper itself had been written by ChatGPT.[5]

The attitude of educators has softened in the intervening period, however, and those hasty bans were soon lifted.[6]  It’s now widely accepted that the technology is here to stay – and rather than being a threat to education as we know it, it could actually deliver huge benefits.

Many experts and practitioners believe genAI offers the possibility of new approaches to teaching and learning that can support engagement and productivity across a broad spectrum of educational levels.  Cornell University’s Center for Teaching Innovation, for example, notes that genAI offers instant access to vast amounts of information, and could support ways to help learners with different abilities, backgrounds or needs.  It could also boost exploration and creativity, spark curiosity, and suggest new ideas and ways of thinking.[7]

GenAI technology also offers considerable benefits for educators.  It could help faculty staff produce content and course materials, such as lesson plans, quiz questions, sample problems, or writing scenarios.  And it could be used by lecturers for producing material such as learning objectives, course descriptions, syllabus statements, or course policies.

A report last year from the US Department of Education said educators are exploring how AI might be used for writing or improving lessons, as well as their process for finding, choosing, and adapting material for use in their lessons.  The report notes that despite its risks, genAI may enable achieving educational priorities in better ways, at scale, and with lower costs.  It adds that “it is imperative to address AI in education now to realize key opportunities, prevent and mitigate emergent risks, and tackle unintended consequences”.[8]

Growing usage

This note of urgency is well-placed.  Evidence from the UK indicates there is already significant use of genAI among both educators and students.  A survey in 2024 by Internet Matters found that 44% of children are actively engaging with genAI tools – with over half of this group (54%) using them for homework or school work[9].  Meanwhile, an earlier survey by the UK government’s Department for Education (DfE), carried out in April 2023, found 23% of college educators, 14% of secondary school teachers and 9% of teachers in primary schools had used genAI.

According to a report in January 2024 again from the DfE[10], the most common ways that educators are currently using genAI is to create content for lessons and administrative support.  Among the leaders and teachers who were using the technology, 62% were deploying it to create lessons or curriculum resources.  Some 42% were using genAI to plan lessons or curriculum content, and 17% were using it to engage with parents or carers.  Other uses included delivering lessons, marking or giving feedback, and drafting policy documents.  There was evidence of genAI being used across a wide range of subjects.  In science and computing, it was being applied to produce experiment ideas, revision tools and code.  In mathematics, educators had created self-marking quizzes, while in language-based subjects the technology was utilized to generate writing examples and comprehension questions.

AI as a tool for personalized learning

In 1984, the educational psychologist Benjamin Bloom identified the so-called ‘two sigma problem’.  Bloom’s research highlighted the considerably better performance that learners could achieve through one-on-one tutoring compared to being taught in the classroom: students with a personal tutor performed better by two standard deviations.  For reasons of practicality and resources, access to this kind of support has so far remained limited.  But some initiatives are now exploring the potential of genAI to make personalized education mainstream.

The Singapore-based company Noodle Factory, for example, provides a gen AI-driven teaching assistant platform called ‘Walter’ that can personalize and enhance students’ experience while relieving the administrative burden on educators.  Walter can provide instant feedback to students on completed exercises, including open-ended questions, while a Q&A function can answer students’ specific queries.

This prioritizes institutional content, such as lecture notes and slides – but educators can also choose to allow access to large language model genAI systems such as ChatGPT to help develop students’ knowledge.  Educators can view analytics to understand how students interact with the chatbot and identify where extra support or intervention may be needed.

Institutions using Noodle Factory’s platform include Ngee Ann Polytechnic’s School of Health Sciences in Singapore, where it has been used by more than 1,000 students.  According to Noodle Factory, students rated the tutoring experience 4.26 out of 5, while the tool’s automated grading saved each lecturer 100 hours a year.[11]  The platform is also being used at the country’s Institute of Technical Education (ITE).  In a nine-month study involving ITE students and lecturers, more than 80% said the platform was effective and were willing to continue using it.  Some 84% of students said the AI tutor was user-friendly and easy to navigate, while about two thirds (64%) of lecturers said they had saved between 50% and 75% of time during the marking process.[12]

The Walter platform has also been used in the US, Mexico, Portugal and the UK – where the University of London is currently piloting the chatbot’s use in its online law programs.[13] Jonathon Thomas, the university’s associate director of learning design and production, says he is impressed by the software’s ability to offer accurate and meaningful real-time feedback.  “Importantly, we don’t envisage that the software would, or could, replace a tutor,” comments Thomas.  “Instead, I believe AI can help bridge that gap by answering common academic questions, removing student frustration and freeing up tutors to focus on more complex problems.”

Thanks to safety features that improve data security and help prevent students from accessing inappropriate content, bespoke genAI tools can offer an advantage over publicly available options such as ChatGPT.  Similar functionality exists for Khanmigo, a genAI-powered ‘personal tutor and teaching assistant’ developed last year by Khan Academy, an educational non-profit organization in the US.

Khanmigo is based on ChatGPT technology – but rather than simply giving answers, it has been adapted to guide learners towards solving problems themselves.  And the app has been additionally trained on Khan Academy’s own content library, which includes courses on subjects including math, science, economics and history.

Khanmigo’s conversational support for learners includes asking challenging questions and encouraging them to think critically – and it can also act as historical characters.  For teachers, the app offers assistance in activities such as lesson planning, creating rubrics, and providing feedback.[14]

The chatbot has been positively received, with a Washington Post journalist recently describing it as the “best model we have for how to develop and implement AI for the public good” and “the first AI software I’m excited for my kids to use”[15].  Khanmigo is currently being piloted by about 20,000 students as part of a US$ 2 million program in the US state of Indiana.  The technology can help students chat with literary characters, craft stories, navigate the college admissions process, and receive coaching to achieve their academic and career goals, with one student commenting that it was “more fun than TikTok”.[16]

Increasing accessibility

So while the use of genAI in education is in its early stages, we’re already seeing some exciting applications.  And it’s clear that whether through direct interactions with learners, or the new capacities it offers for educators, there is great potential for the technology to support more personalized learning paths.  Indeed, Khan Academy founder Sal Khan believes that “we’re at the cusp of using AI for probably the biggest positive transformation that education has ever seen”.[17]

As well as enhancing mainstream education, genAI can make it easier for students and educators to adapt or create content in a wide range of ways to suit different learning needs.  Researchers with conditions such as brain injury, dyslexia and aphantasia (the inability to visualize), for example, recently told the Nature journal how converting academic content into different forms – such as transcribing and summarizing lectures or generating images from descriptions – could support their work[18].

And the UK educational non-profit Jisc notes that genAI can be particularly effective at reworking text, which has the potential to benefit neurodivergent students.  This could include translating, simplifying, and rephrasing material, or providing examples[19].  One popular accessibility resource is the goblin.tools platform, which uses genAI to offer a range of tools designed to support neurodivergent learners in different tasks. These include changing the tone of writing and estimating how long a task might take to complete.  One student with attention deficit hyperactivity disorder (ADHD) at the University of Technology Sydney recently discussed how goblin.tools, as well as the text-to-speech app Speechify and Yoodli, a speech coach program, helped her maintain focus and manage her time effectively[20].

Researchers are investigating further possibilities for using advanced AI technology to support students with special educational needs.  Amid a shortfall of speech and language pathologists in the US, one project led by the University at Buffalo aims to develop AI-based tools to screen children for speech and language challenges and then work with teachers to provide individualized interventions[21].  And the University of Maine is leading a project investigating innovative practices in using AI in education, funded by the Council for Exceptional Children.[22]

Future possibilities

These are just a few examples of the huge ambition around the possible future applications of genAI in education more broadly.  In a speech last year at University College London, for example, professor Mike Sharples of the Open University highlighted the potential for the technology to be used as “a personalized AI guide for learning and life”.  As genAI becomes more sophisticated, he explained, chatbot tools could be used not just to help with particular tasks but as “lifelong mentors” with persistent memory.[23]

Another fascinating development is Google’s experimental note-taking app, NotebookLM.  Previously known as Project Tailwind, the app synthesizes and analyzes information in uploaded documents.  Because the tool is grounded in these particular sources, Google says it “creates a personalized AI that’s versed in the information relevant to you”.  The company describes it as a “virtual research assistant that can summarize facts, explain complex ideas, and brainstorm new connections”.  It gives the example of a medical student uploading a scientific article about neuroscience and asking the app to create a glossary of key terms related to dopamine.[24]

Stanford University’s seed grant scheme supporting early-stage research into educational uses of genAI[25] gives an idea of the breadth of the technology’s potential uses.  The initiatives it is backing include a project researching the creation of 3D virtual teaching assistants that can simulate real-world medical training, and the generation of high-quality descriptions of data visualizations – to improve accessibility for blind and low-vision learners. Other projects have been awarded funding to research the detection of AI-generated text, and how to help teachers write effective feedback for students.

Frameworks for the future

A recent analysis by Morgan Stanley argued that genAI’s potential to increase efficiency, while improving quality and access, could bring US$ 200 billion in value to the global education sector by 2025[26].  But while governments and educational institutions increasingly appreciate such opportunities, they remain keenly aware of the technology’s potential hazards. Unesco’s director general Audrey Azoulay has warned that genAI “cannot be integrated into education without public engagement, and the necessary safeguards and regulations from governments”.[27]

To get the best out of genAI in education, we will need effective policy frameworks to underpin its use. Unesco’s ‘Global Guidance on Generative AI in Education and Research’[28], for example, suggests an age limit of 13 for using AI tools in the classroom and calls for teacher training on the subject.  Noting that only a handful of countries have adopted specific policies or plans for the use of AI in education, it also recommends eight key measures that governments can take to develop policy frameworks for regulating the use of genAI.  These include promoting inclusion, equity and diversity, protecting human agency and monitoring and validating systems.

As the document’s foreword notes, the capacities of genAI have potentially huge implications for education.  Fully capitalizing on these implications in a way that delivers measurable benefits for both educators and students will be no easy task, however.  While the early skepticism was understandable, it’s encouraging that many places are now implementing more balanced approaches.  Cornell University, for example, suggests educators take a flexible approach to the use of genAI, in which they can either prohibit, allow with attribution, or encourage its use among students[29].  Approaches such as this – recognizing the benefits of genAI while creating sensible structures to manage its risks – will be essential over the coming years, as the technology plays an increasingly large part in teaching and learning.

GenAI is coming to disrupt our classrooms and lecture halls whether we like it or not.  Our challenge is how to best harness its power as a positive tool to enhance education, not an excuse to undermine it.

 

[1] https://www.theverge.com/2023/11/6/23948386/chatgpt-active-user-count-openai-developer-conference

[2] https://www.bloomberg.com/professional/insights/data/generative-ai-races-toward-1-3-trillion-in-revenue-by-2032/

[3] https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2023-generative-ais-breakout-year

[4] https://www.theguardian.com/us-news/2023/jan/06/new-york-city-schools-ban-ai-chatbot-chatgpt

[5] https://www.theguardian.com/technology/2023/mar/19/ai-makes-plagiarism-harder-to-detect-argue-academics-in-paper-written-by-chatbot

[6] https://www.nytimes.com/2023/08/24/business/schools-chatgpt-chatbot-bans.html

[7] https://teaching.cornell.edu/generative-artificial-intelligence

[8] https://www2.ed.gov/documents/ai-report/ai-report.pdf

[9] https://www.internetmatters.org/hub/research/generative-ai-in-education-report/#full-report

[10] https://assets.publishing.service.gov.uk/media/65b8cd41b5cb6e000d8bb74e/DfE_GenAI_in_education_-_Educator_and_expert_views_report.pdf

[11] https://www.noodlefactory.ai/using-ai-to-improve-student-engagement-and-educator-productivity-ngee-ann-poly

[12] https://articles.noodlefactory.ai/ai-education-tackling-two-sigma-problem-walter

[13] https://onlinelearning.london.ac.uk/2023/03/29/ai-teaching-assistant-pilot/

[14] https://guides.libraries.uc.edu/ai-education/kh

[15] https://www.washingtonpost.com/opinions/2024/02/22/artificial-intelligence-sal-khan/

[16] https://content.govdelivery.com/accounts/INDOE/bulletins/372fd58

[17] https://blog.khanacademy.org/sal-khans-2023-ted-talk-ai-in-the-classroom-can-transform-education/

[18] https://www.nature.com/articles/d41586-024-01003-w

[19] https://nationalcentreforai.jiscinvolve.org/wp/2023/10/25/generative-ai-and-accessibility-in-education/

[20] https://lx.uts.edu.au/blog/2023/10/23/genai-student-perspectives-accessibility/

[21] https://www.buffalo.edu/ai4exceptionaled.html#title_301642566_copy

[22] https://umaine.edu/news/blog/2024/01/11/umaine-faculty-investigating-best-uses-for-ai-in-special-education/

[23] https://www.ucl.ac.uk/teaching-learning/case-studies/2023/aug/generative-ai-and-education-futures

[24] https://blog.google/technology/ai/notebooklm-google-ai/

[25] https://acceleratelearning.stanford.edu/funding/generative-ai/

[26] https://www.morganstanley.com/ideas/generative-ai-education-outlook

[27] https://www.unesco.org/en/articles/unesco-governments-must-quickly-regulate-generative-ai-schools

[28] https://unesdoc.unesco.org/ark:/48223/pf0000386693

[29] https://teaching.cornell.edu/generative-artificial-intelligence/cu-committee-report-generative-artificial-intelligence-education