GAAD TAKEOVER WEEK: How Do Disabled Students Use and View Generative AI Tools at the University of Sheffield

 By Xin Zhao, Andrew Cox, Bryan Coleman, Xuanning Chen

GAAD (Global Accessibility Awareness Day) logo, a circle with GAAD in the middle and a computer keyboard on the side of the circle

The appearance of a growing range of generative AI-based services used by students has created controversy across Higher Education. Concerns have been raised about their impact particularly regarding academic integrity, personal privacy breaches, and misinformation  (Zhao, Cox, & Cai, 2024). 

Despite these concerns, these new services also potentially offer  significant benefits, in enhancing writing quality, work efficiency, and employability (OECD, 2023; Ding, Zhao, & Wang, 2024; Zhao, Xu, & Cox, 2024). In particular, they have the potential to level the playing field for disabled student groups (Arora, 2024). Currently, there is limited research about how students use generative AI tools in self-directed learning, especially how disabled students use them.

To inform the University of Sheffield's policies and practices in supporting disabled students, the Information School and the Disability and Dyslexia Support Service (DDSS) conducted a collaborative survey. 

A collection of silhouettes, in different colours, with the brain highlighted on each person


Informed by the results of a previous interview-based study  (Zhao, Cox, & Cai, 2024), the questionnaire was organised in three main sections: Demographic data, Disability-related questions, General GenAI usage and views. This blog covers some of the key findings of the survey report and highlights our recommendations for staff who are interested in or have a responsibility for supporting disabled students at the University of Sheffield. A detailed report on the survey can be found here.

Participants  

As of February 2024, the University of Sheffield has 7188 disabled students.The survey gathered 124 valid responses from disabled students. Most participants identified themselves as female (57%) or male (28%). The majority of respondents were pursuing a bachelor's degree (69%, n=85), followed by 20% (n=25) studying for a master's degree. Students came from a wide range of departments, with the largest representation from Bioscience (12%, n=15). 95% (n=118) of participants reported that English was their native language. In terms of digital competence, most participants rated themselves as having an intermediate (61%, n=75) or advanced level of skills (27%, n=33) 

Disability Conditions Affecting Students' Academic Writing

The three groups of disability that were most common among questionnaire respondents were:  1) neurodiversity, including ADHD (34.7%, n=43), 2) specific learning difficulties such as dyslexia and dyspraxia (29%, n=36), or 3) and social/communication impairments such as autism spectrum conditions (21.8%, n=27). The response rate from these groups suggests that they that see value in generative AI. It is important to note that some students disclosed having more than one condition in this survey. 

Main Barriers to Writing Associated With the Identified Conditions

Given that a significant portion of survey participants reported neurodiversity, with ADHD (34.7%) and specific learning difficulties like dyslexia and dyspraxia (29%), the main barriers encountered in the writing process may stem from these conditions. Open-text data on challenges in writing suggest students face the following barriers: 

  • Difficulty in understanding assignment briefs and rubrics, particularly in grasping expectations and interpreting hints or prompts.
  • Mental blocks, characterised by difficulty in translating thoughts into written words, experiencing brain fog, and mind blanks.
  • Low motivation or difficulty in maintaining concentration levels for writing, often attributed to fatigue and stress.
  • The need for extra time for reading and writing tasks.
  • Struggles with the reading and proofreading/editing stages.
  • Challenges in word choice.
  • Writing structure issues, such as organising sentences and paragraphs effectively.
  • Finding an appropriate tone of voice and writing style.
  • Time management challenges

Main GenAI Tools Used By Students During the Writing Process

Among all participants, 77% (n=96) reported that they routinely incorporate GenAI tools into their learning. Whereas only 23% (n=28 students) report seldom or never using GenAI tools for learning. The majority of students (91%) do not spend money on subscribing to GenAI tools. The most commonly used GenAI services are ChatGPT, Google Translate and Grammarly Go. But by far the most commonly used of these was ChatGPT. Interestingly, Google Gemini - despite being recommended by the University - only ranks 4th.

Many students said they use GenAI to summarise reading materials, a preference that may stem from difficulties with word reading accuracy or fluency. Another notable finding is that students reported GenAI tools help them overcome mental blocks. This is followed in frequency of mention by the use of these tools to support the idea-generation process, structuring assignments, rewrite and paraphrase text. These usages correlate with the challenges that students reported during their writing process in the above section.


Students’ views towards GenAI tools

Students generally hold positive views towards using GenAI but they also have concerns about these tools. They consider that students should be involved in the University’s process of formulating GenAI policies. They suggest that the university should offer training on how to use GenAI effectively. They consider learning to use GenAI important for their future careers. Only a small group believes GenAI should be banned. The majority of participants are concerned with issues including deepfakes, inaccuracies, breaches of academic integrity, and risk of automating thinking processes.

Support and Training Students Wish to Receive from the University

Students expressed a strong interest in AI training from the university. And we specifically asked them about the support they would like the university to offer. Our open-text data suggest the following areas for training:  
  • General/introductory training on GenAI
  • Training on how to use GenAI effectively
  • Training on avoiding unfair means and unethical uses
  • Writing prompts
  • How to use GenAI to search for information
  • How to fact-check or reference information from GenAI
  • How to use GenAI for summarising texts effectively
  • Introduction to a wider range of tools available for students

Recommendations on supporting disabled students

Based on our analysis of the survey, we have the following recommendations to support disabled students in the writing process:
  • Who needs support.
Given the groups who responded to the survey, it seems appropriate to provide tailored support to students with ADHD, specific learning difficulties such as dyslexia and dyspraxia, and social/communication impairments such as autism spectrum conditions.  Because understanding assignment briefs is a particular problem it might be useful to reinforce existing messaging from Learning Support Plans (LSPs) to departments about setting clear assignment briefs and allowing clarification appointments.
  • Policy
There seemed to be an ongoing lack of clarity about appropriate and inappropriate uses of GenAI, which was inhibiting use. A policy statement specifically on allowed uses for those with disabilities might be helpful.
Students themselves should have a voice in policy making.
  • AI literacy
Examining the results from the perspective of our AI literacy model (Zhao, Cox and Chai, 2024) there are gaps around a) awareness of the wider range of AI tools and probably in prompt engineering b) socio-ethical concerns, such as environmental impact.
  • Training.
Given that students were asking for training it would be useful to
  • Signpost students to existing GenAI training provided by the university, including advice on effective prompt engineering targeting support for students in summarising reading, overcoming mental blocks, brainstorming ideas, and structuring assignments.
  • Design training on university  recommended GenAI tools and how to effectively use them in the context of disability, particularly focusing on Google Gemini (although ChatGPT is the favourite one), Translation tools (Google Translate, DeepL), Proofreading and Rewrite Tools (GrammarlyGO, Quillbot), addressing barriers including translating thoughts into writing and motivation.
  • Design training on how to critically engage with information provided by GenAI, including fact-checking, finding and referencing original sources.
  • Equity in access to GenAI tools. 
  • Given their popularity and usefulness, it might be appropriate to provide subscriptions for students to tools such as GrammarlyGO and possibly translation tools. 

Please join us on the 16th of May 2024 for GAAD (Global accessibility awareness day) to hear more about this study. You can book onto Supporting Disabled Students In The Era Of Generative AI now through MyDevelopment.

References:
Arora, P. (2024). From Pessimism to Promise: Lessons from the Global South on Designing Inclusive Tech. MIT Press
Ding, Z., Zhao, X., & Wang, W. (2024). The Transformative Impact of AI-Powered Writing Assistants in Education: A Comprehensive Systematic Literature Review. iConference 2024 Proceedings.
Zhao, X., Xu, J., & Cox, A. (2024). Incorporating Artificial Intelligence into Student Academic Writing in Higher Education: The Use of Wordtune by Chinese International Students. Proceedings of the 57th Hawaii International Conference on System Sciences. 
Zhao, X., Cox, A., & Cai, L. (2024). ChatGPT and the digitisation of writing. Humanities and Social Sciences Communications, 11(1), 1-9.
Zhao, X. Cox, A, Chen, X, Coleman, B. (2024). A Report on the Use and Attitudes Towards Generative AI Among Disabled Students at the University of Sheffield Information School. The University of Sheffield. Report.