Potentials and risks of GenAI in sex education - a comparative case study of three GenAI chatbots

• In Event: Highlighted Session: Possibilities and Problematics for Gender Equity across Digital Technologies

Proposal

This case study analyses the discourse around pornography, and violence against women, created by three different chatbots applying generative AI technology. The study investigates discourses of ChatGPT (OpenAI), Gemini (Google), and an GenAI chatbot developed by a sex education company called Beducated (beducated.ai). The paper discusses the discourses through a feminist lens, with the promotion of gender equality in mind. The study aims to contribute to the field of sexuality education, which will also transform more and more in a digital society.
The rise of Large Language Models (LLMs) powered by GenAI such as ChatGPT has had a massive impact on learning, communication, and knowledge creation (Fütterer et al, 2023). GenAI can possibly have a long-term transformative effect on teaching and learning in practice, and they have various potentials as well as risks and limitations. They might create the potential for more personalized and adaptive learning facilitation (Baidoo-Anu et al., 2023). These language models can also contribute to inclusive education as they can help with disadvantages in education such as visual impairment, dyslexia. They can also make knowledge more equally accessible across different locations, education systems, and schools.
While ChatGPT has become a very useful tool to find information quickly and made text production incredibly fast, it also comes with risks and limitations. The accuracy of information is a major one, as GenAI can create convincing texts but with incorrect information (Zhang, 2023). Furthermore, these language models come with privacy issues, as they are trained on large, publicly available data that is scraped from the internet (Khowaja et al., 2024). The training data also determines what knowledge is presented, and social biases can be therefore easily reproduced by GenAI. Educational scientists point to a use of GenAI as a tool which are accompanied by facilitation, critical thinking and interrogation of sources (Fütterer et al, 2023). This will require a thorough understanding of AI both by teachers and students. It is also a question at this point how it will impact learning on the long term, whether these models can become a helpful tool which indeed allow more space for personalized learning, unbiased feedback, and interactive learning.
A burning question is how GenAI can be helpful with complex, widely debated topics, which are hardly about yes or no questions – such as sex. The internet is a primary source for children and young people to learn about sex, as well as a main scene of the sexual culture where sexual socialization happens (Ragonese et al., 2017). Research about sex education shows that schools often offer inconsistent sex education, often not addressing students’ questions and interests sufficiently. A recent systematic review on sex education programs included in scientific literature found that pornography, and gender-based violence are neglected topics (Aznar-Martinez et al., 2024). The study also shows that programs about porn mainly addressed 13-17 year olds. Since evidence shows that the first encounter with porn is predicted to be at a lower age, porn should be addressed earlier, adequately and appropriately. Another finding was that the programs addressing pornography and gender-based violence failed to have a holistic and a community-based approach – for example not addressing parents, not including other stakeholders from the schools (Aznar-Martinez et al., 2024). Pornography is a challenging topic, because adults are often not aware of what children experience in their online presence and how the sexual culture looks today in the online world. Or even if they are aware, they often don't know how to address it with young people. Evidence shows that professional learning is a crucial element for effective sex education and violence prevention programs. Could GenAI contribute to the training of adults in these matters, or form a bridge between adults and young people for these discussions? If yes, what kind of knowledge do we encounter on current GenAI chatbots?
A study exploring teachers’ perception of incorporating GenAI chatbots into sex education shows positive attitudes from the teachers, and highlights their main concerns and conditions. These included accuracy and reliability of the information, accessibility of the technology, values presented by the model, and safety measures in regarding children’s use of the model (Lin et al., 2023).
This paper uses feminist discourse analysis to investigate what the current discourses of GenAI chatbots are regarding pornography, and violence against women. There is no evidence available yet whether and how teachers, parents, or young people use these tools to learn about sex, however, we can assume that they are used, and will be more widely used in the future. Therefore, it is important to carry out analysis of the knowledge produced by GenAI and investigate it through a feminist lens. GenAI both applies and reproduces existing discourses by creating knowledge, just as curricula or education policies are (Ball, 2015). Analyses of the discourses can lead to improvement of GenAI models, of their application in education, and in future policies around them.

Author

• Bogata Kardos, University of Gothenburg