Dissertation Oriane Camille Iris Pierrès
«Using Artificial Intelligence in Higher Education: The Perspective of Students with Disabilities»
Abstract
The recent rapid development and availability of large language models (LLMs) such as ChatGPT by OpenAI has spurred a lively debate in how they can benefit or threaten learning in higher education [133, 196, 205]. Currently, many higher education institutions (HEI) are discussing how they should regulate the use of artificial intelligence- based (AI) tools, how they can teach their students to best use them, and how and whether at all they should adopt these tools [141, 307]. Pressure on HEI is particularly high as students are already adopting tools like ChatGPT [249].
Research on AI-driven educational technology (EdTech) dates back to the 1980s [313]. Since the 2010s, the number of original research articles and reviews on its use in higher education has grown significantly [36, 322]. In this thesis, AI EdTech refers to technologies leveraging machine learning techniques for educational purposes. Zawacki-Richter et al. [322] distinguish four uses of AI: 1) profiling and prediction, 2) assessment and evaluation, 3) adaptive systems and personalization, and 4) intelligent tutoring systems. AI EdTech promises numerous benefits to learners, educators, and administrators, including enhanced learning, earlier intervention for those in need, and improved progress monitoring [325].
However, educators and experts diverge on how to deal with newer AI models such as ChatGPT, Gemini or Deepseek. On the one hand, some worry that students may be cheating and not learning effectively [196, 267]. This has led to the development of tools claiming to detect AI-generated content like Turnitin or GPTZero [101]. However, these tools have been criticized for their lack of accuracy [58, 190, 224]. They could also disadvantage non-native-English speakers [224] and students with disabilities (SWD) [71]. Others reject the adoption of AI due to its impact on the environment and the exploitation of workers in developing countries [78].
On the other hand, proponents of AI adoption argue for the need to increase students’ digital literacy to prepare them for their future job and promote the responsible use of AI [101, 141, 196, 230]. Additionally, some educators call for the adaptation of HEI, especially when it comes to assessment [134, 196]. Considering these different perspectives, much research is needed to understand how to best gain from AI in tertiary education while minimizing negative consequences. Many researchers have also pointed out the lack of ethical considerations in the field of AI EdTech [36].
This thesis contributes to this ongoing discussion by highlighting the risks and challenges for the inclusion of students with disabilities and collecting the opinions of this group on the subject. While several researchers have gathered the opinions of the general student population [48, 143, 274, 294], the perspective of students with disabilities is often neglected [281]. Yet, as will be explained in section 3, HEIs have a legal and moral obligation to consider the perspective of students with disabilities in this debate. Additionally, including the voice of those who face daily barriers in higher education could drive innovation.
The following sections will first define the concepts of disability and neurodivergence. Then, I will clarify the concept of AI EdTech before presenting the ‘universal design’ and ‘universal design for learning’ frameworks as experts often advocate for its implementation in higher education to include SWD [56]. After this clarification of concepts and frameworks, I will argue that the perspective of SWD is crucial to ensure a fair and responsible development of AI EdTech. Further, I will outline and justify the chosen methodology. Finally, this introductory chapter will conclude with a short presentation of the conducted studies.