Eight grand challenges for artificial intelligence and education

Eight grand challenges for dealing with the impact of artificial intelligence on education.

University of Sydney computer science professor Judy Kay has eight grand challenges for artificial intelligence in education.

Educators and artificial intelligence specialists face eight grand challenges in dealing with the impact of AI on education, according to University of Sydney computer science professor Judy Kay, who heads the Human Centred Technology Research Cluster in the engineering and IT faculty.

Speaking to a conference, ChatLLM23, on the ethics of generative AI last week, Professor Kay said she was “right in the thick” of education AI issues. She is the editor-in-chief of the International Journal of Artificial Intelligence in Education.

Her eight challenges, which she presented to the conference, are:

Curriculum change: Students need to be taught skills in the new AI tools that will be widely used in business and the professions to generate documents, write computer code, solve mathematical problems, generate education material and enable collaboration between humans and machines. New job areas are also appearing, such as “prompt engineering” to give AI systems the best prompts to generate the required output.

Human control and responsibility: “It’s fine to use AI. But the real point is that ultimately you’re responsible. We need deep awareness of limitations of the tools,” Professor Kay said. She pointed out that we do not even have a word for “fact-checking” AI output.

Educators building confidence: “We have to get educators to the point where they can do the first two,” she said.

Harness AI in educators’ current work: Educators must use AI and explore what it can and can’t do. For example, ask it to write an exam, and then critically examine how good the result is.

Accessibility implications: AI can make things better for those with a disability. For example, it can write alt text, which describes web page images to vision-impaired people.

AI regulation fit for education: Because most education is aimed at children and young people, particular care needs to be taken to regulate AI-driven education. “What is informed consent when using AI,” she asks.

The changing nature of research: AI is posing challenges for research in universities. Professor Kay said her colleagues at the International Journal of Artificial Intelligence in Education had experimented with asking AI to generate fake research papers. “It will generate the (research) data for you too. Not real data but plausible data, so it’s pretty frightening,” she said. Two principles should be applied, she said. One is that AI should not write author academic papers. The second is that, where academic authors use AI, it should be acknowledged.

Rethinking AI education research: “I think we have an incredible opportunity to build these (AI) systems so we can learn more about how people learn,” Professor Kay said. She also said university education researchers needed to work co-operatively with commercial developers of AI products in the education field.