“ChatGPT can lead to greater equity in education”
The public launch of ChatGPT has led to considerable dismay at schools and universities. However, a position paper authored by more than 20 scientists at the Technical University of Munich (TUM) and Ludwig-Maximilians-Universität (LMU) working in educational, social, computer and data sciences shows that the so-called language models also present many opportunities for education. In this interview, the coordinator Prof. Enkelejda Kasneci explains how the new technology could benefit learners and make teachers’ work easier.
The New York School District has banned the use of ChatGPT. It that the right way forward?
We think it’s the wrong way. It’s also the easy way out. The development of language models like ChatGPT is a technological milestone. There is no going back. The tools are in the world. They will get better and we have to learn how to use them constructively. We’re convinced that they offer big opportunities for the empowerment of previously disadvantaged people. ChatGPT and similar programs can lead to greater educational equity.
Who could benefit from ChatGPT applications?
First, this is a tool that makes it possible for everyone in the world with internet access to learn – regardless of the quality of the education system in their own country. Second, it can help people to express themselves better in writing who otherwise have difficulty doing so, for example due to a disability. That can offer them new opportunities to participate in society.
And how about at school?
Here we see major potential for the personalized use of such tools to overcome the individual weaknesses of each child, to bring out strengths and contribute to constructive learning successes. We’re talking about an AI-based tool that can recognize and produce various forms of text. Pupils could get suggestions for alternative wordings or better text compositions. That can help them improve their ability to express themselves. My research chair team has just developed a tool that can analyze an essay using large language models and provide feedback such as: “It would be better to use a consistent verb tense” or “You could pay more attention to the subjunctive.” This feedback could be adjusted to the age and skill levels of the individual children.
At present people seem to be worried that language learning will deteriorate.
We don’t see it like that. On the contrary, applications like this can promote the understanding of language. But they can also be helpful in other subjects, for example by creating questions on a certain topic. Students preparing for exams at home could use it as a study partner that would focus on points where more work is needed. That is a level of individualized learning that would be difficult to achieve in the classroom.
So AI could make teachers’ work easier, too?
We think so. Artificial intelligence could also support them in correcting schoolwork.
… that are generated with ChatGPT for pupils who want better grades.
Of course nobody can be certain that written homework is done without outside help. But these discussions remind me very much of the debates when Wikipedia was launched. Back then, people were worried that most homework would be copied from the internet. Then, as now, we have to start teaching children from the elementary school level the importance of not relying on the information of just one internet portal and that they need to check information and back it up with sources. If a text automation program can write the answers for an exam, this also says something about the quality of the exam itself. We have to ask ourselves what teaching methods we are using and to what extent we are helping learners to acquire competencies such as critical thinking and problem solving skills.
What is needed to ensure that ChatGPT and similar models are actually used in beneficial ways in the classroom?
Researchers must deliver more robust conclusions on the effects of the language models on learning, how they can be used in a certain learning context and when they will be ready for use. Holistic teaching concepts and continuing education opportunities will also be needed for teachers. All of us need to work together and bring about a rapid response. And the providers must take issues surrounding data protection, security, bias and distortions in machine learning, intellectual property and transparency very seriously.
It will take some time before these goals are achieved. How can teachers deal with ChatGPT and similar tools in the meantime?
We advise all teachers: try out these tools! Discover them together with your students. There are no limits on creativity as long as you maintain a critical perspective.
About Enkelejda Kasneci:
Prof. Dr. Enkelejda Kasneci is heading of the Chair of Human-Centered Technologies for Learning at the TUM School of Social Sciences and Technology. The school studies the interdependencies of technology and society. Prof. Kasneci is setting up the new TUM Center for Educational Technologies that will develop and test teaching and learning methods with new technologies. She is also a member of TUM’s Munich Data Science Institute, which brings together researchers from various disciplines to investigate machine learning issues. Before being appointed to a professorship at TUM, Enkelejda Kasneci studied informatics and conducted research into human-machine interactions at the University of Tübingen.
The participating TUM scientists are conducting research at the TUM School of Social Sciences and Technology, the TUM School of Computation, Information and Technology and the Munich Data Science Institute.
Prof. Dr. Enkelejda Kasneci
Technical University of Munich (TUM)
Chair of Human-Centered Technologies for Learning
+49 89 289 25118
Kasneci, Enkelejda, Kathrin Seßler, Stefan Küchemann, Maria Bannert, Daryna Dementieva, Frank Fischer, Urs Gasser, et al. 2023. ChatGPT for Good? On Opportunities and Challenges of Large Language Models for Education. EdArXiv. January 30. doi:10.35542/osf.io/5er8f
https://mediatum.ub.tum.de/1698539 Images for media coverage
https://www.sot.tum.de/en/ TUM School of Social Sciences and Technology
https://www.cit.tum.de/en/ TUM School of Computation, Information and Technology
https://www.mdsi.tum.de/en/ Munich Data Science Institute