Learning and Teaching about Ethics in Language Technology

Information Science students engaged in discussions with high school students about the societal impact of language technology. These visits to various high schools in and around Groningen are part of the third-year course "Ethical Aspects in Natural Language Processing," which was added to the Information Science Bachelor's programme in academic year 2021-2022. In this course, future professionals learn to reflect on the ethical implications and the overall societal impact of the language-based AI tools they will contribute to, as well as how to communicate these topics to non-professionals. The course was co-developed by prof. dr. Malvina Nissim .

The release of ChatGPT to the general public has put a spotlight on how pervasive and important language-based artificial intelligence is in society. Recent advancements might feel revolutionary, but progress has actually been gradual and, until now, has mostly gone unnoticed, says Malvina Nissim, Professor of Computational Linguistics and Society at the University of Groningen.

"For several years now, millions of people have been using machine translation on a daily basis, voice assistants have inhabited our houses, and enhanced web searches have been integrated into our online queries," she explains. "These tools have become so ingrained in various aspects of our public and personal lives that we often take their presence and smooth operation for granted."

Questions, concerns, and heated debates

Questions, concerns, and heated debates have arisen surrounding the emergence of this technology. Are these tools safe to use? What are the risks of generating texts that closely resemble human-produced language but have no basis in reality? And who should be held responsible for the potential harmful outputs of language-based technologies?

"The ethical concerns in this field are subjects of ongoing debate and evolve as new applications continue to enter society," says Nissim. "It is crucial that we train young professionals to reflect on and engage dynamically with such concerns, as well as effectively communicate them, especially to those who are likely to be the heaviest users of language technology."

Learning about Ethical Aspects

This is why the course "Ethical Aspects in Natural Language Processing" was introduced to the Information Science Bachelor's programme. "The module aims to establish explicit connections between the practical skills students acquire in natural language processing and the potential impact of applying those skills," explains Nissim. "Instead of treating ethical concerns as an afterthought, the goal is to expose students to new ideas put forward by the research community and encourage them to critically examine the rapidly evolving technologies entering the market. Ultimately, the course aims to be transformative, fostering critical thinking and awareness by allowing space for open questioning and challenging unstated assumptions inherent in technology design and use."

Moreover, it is essential for students to learn how to effectively communicate the impact of language technology on society. To achieve this, six groups of students presented on the ethical aspects of language technology to high school students as the culmination of the course. Nissim: "Presenting to high school students provided students with the opportunity to fulfill the role of experts in the field and take responsibility for informing the general public, particularly the younger generation."

Furthermore, by conveying potentially sensitive information to younger individuals, who are avid users of language technology but may not yet fully grasp how it works or the implications of its use, students had to engage in deeper reflection on the course materials and topics discussed.

The school visits were a valuable experience for the students. "Many students had never presented at a high school before, and they were pleasantly surprised by how much they enjoyed discussing these important topics with the students. It is equally important that the students gained a lot from it, as they are the generation that heavily relies on NLP models," Nissim notes.

About GroNLP

Within the Faculty of Arts, the GroNLP group, the Computational Linguistics group of the Centre for Language and Cognition (CLCG), conducts research in the field of language-based AI. They also organize seminars and workshops, both within and outside the university, to inform and raise awareness about the development and use of language technology, such as "An Evening with ChatGPT."

About the course

The course "Ethical Aspects in Natural Language Processing" was designed and developed by prof. dr. Malvina Nissim, Professor of Computational Linguistics and Society at the Faculty of Arts, and dr. Beatrice Savoldi, a former visiting PhD candidate from the University of Trento & Fondazione Bruno Kessler, who is currently a postdoc at the same institution. The course has been offered since the academic year 2021-2022 and is taught by Nissim and a teaching assistant (in 2021-2022 this was Riemmelth Ruitenbeek). Collaborating with high schools will continue to be part of the course.

Read also: How Human Biases sneak their way into Computer Programs

Quotes from students:

“I learned that it’s better to keep an open mind to this being a process of continual learning that involves acknowledging the potential risks and unintended consequences of NLP technology, actively seeking out diverse perspectives, and being open to reevaluating and revising my opinion as new, ‘better’ insights become available.”

“I thought the process of incrementally building a presentation catered towards younger students was a wonderful educational experience. Processing all the information on a weekly basis and narrowing things down, together with my friends, to the core concepts that mattered most to me really helped me internalize them.”

“It is refreshing to have a course like this where we reflect and think critically about the field we are in. NLP and machine learning in general are having a bigger impact on the world every year, which will probably mean that the tools we create later in our careers will also have an impact on some people's lives. We will have to think about the impact of the tools we build and not just release them without thinking about the consequences.”