Generative AI and the Future of Higher Education: Challenges and Opportunities

Photo by Alex Knight on Unsplash

A moment in technological history

In 1952, Arthur C. Clark published The Exploration of Space. Its closing words offered a hint of the passions that would animate much of Clarke's career.

We stand now at the turning point between two eras. Behind us is a past to which we can never return … The coming of the rocket brought to an end a million years of isolation … the childhood of our race was over and history as we know it began.

Last week, ChatGPT was seen by many as an equivalent moment in the history of technology.

ChatGPT - which stands for “generative pre-trained transformer” has been hailed as the best artificial intelligence chatbot ever released to the general public. It’s the first chatbot that’s enjoyable enough to speak with and useful enough to ask for information.

Developed by OpenAI, the same company responsible for tools like GPT-3 and DALL-E 2, ChatGPT quickly gained popularity after its release, with over a million people signing up to test it within five days of its launch. However, ChatGPT also raises important questions about the role of AI in higher education.

The future of the essay in higher education

The use of essays in higher education has been a longstanding tradition for evaluating students' critical thinking and writing abilities. However, the emergence of AI technology like ChatGPT, which is capable of generating ideas and arguments and writing them clearly, has disrupted this tradition and raised questions about the continued relevance of the essay in higher education.

One of the key challenges posed by ChatGPT and other generative AI technologies is the fact that they have no understanding of content and do not evaluate information for accuracy or importance. If we are assigning students tasks that can be completed by generative AI without any consideration of meaning, what are we really assessing, and are we helping students learn in the process?

Is the calculator a useful analogy?

One analogy to what ChatGPT means for assessment is the calculator, which allows students to quickly and easily solve mathematical equations without having to do tedious calculations by hand. However, unlike calculators, which are deterministic and always produce the same result for a given set of inputs, ChatGPT is a probabilistic system that generates output based on a statistical model of language.

The probabilistic nature of ChatGPT poses challenges for educators who are seeking to assess student learning. Unlike with a calculator, where there is a single correct answer and the ability to show one's work is sufficient to demonstrate understanding, the output of ChatGPT is not always right or wrong. Instead, accuracy is a reflection of the statistical model on which it is based, the data on which it is trained, the prompts it receives, and the previous conversations it has had.

The benefits of integrating generative AI into higher education

As AI technology continues to advance, many of our students will likely use generative AI in their future jobs. The rise of generative AI has challenged the notion that jobs requiring creativity, compared to those involving more routine, repetitive and highly-structured tasks, are relatively immune to disruption from technology.

One of the key benefits of incorporating generative AI into higher education is that it can help students develop important skills that will be essential to the future of work. For instance, many jobs will involve a "sandwich" workflow, in which a human provides a prompt to the AI, the AI generates a menu of options, and the human chooses an option and adds any final touches. This type of workflow requires students to be able to make decisions, and edit and verify the output of the AI.

To effectively incorporate generative AI into higher education, we need to create assessments that take into account the unique capabilities and limitations of this technology. This may involve using methods of assessment that value the processes and experiences of learning rather than relying on a single artefact like an essay.

Educators can use generative AI as a tool to help their students engage in deep and critical thinking to evaluate and verify the accuracy of (rather than simply regurgitate) what has been generated. Students could be required to engage in metacognitive reflection, asking them to articulate what they have learned and then responding to their reflections as evidence of learning.

Providing students with a “writing companion” could also help them to see how creative writing embraces uncertainty and accommodates multiple voices. If composing prose is a dialogue between the writer and the world, then the act of writing as a game or relay between human and AI brings this dialogue to the surface. Such an approach could enhance a student’s confidence in the final version of their work.

Lessons for the collaboration between the humanities and STEM

While the value of the humanities has long been recognised, it’s decline in higher education continues. However, the emergence of natural-language processing is bringing the humanities and STEM fields together, highlighting the need for a broader, more interdisciplinary approach to education, that includes history, and ethics. The last week has clearly demonstrated that the connection between the humanities and technology requires graduates with a breadth of vision that transcends any particular disciplinary field. It is more important than ever for educational institutions to cultivate and strengthen habits and dispositions, such as curiosity, experimentation and reflection, in an integrated and transformative manner.