Research shows risks and opportunities of GenAI for dyslexic, neurodivergent and disabled students

In November 2022, ChatGPT made significant headlines around its release due to its advanced language processing capabilities. There has since been a significant increase in both academic and research conversations across the globe about its potential use and implications in the higher education.

The ability of ChatGPT to process natural language and generate human-like conversational explanations is often claimed to make learning more accessible and engaging for students.

This recent rise in the use of Generative AI (GenAI) tools has sparked the interest of support teams across schools and universities in considering recommending them to aid the learning of dyslexic, neurodivergent and disabled students. For example, for use with tasks such as explaining concepts and ideas and summarizing articles.

We worked with an undergraduate student to get the views of these students on GenAI in higher education. Specifically, we were interested in understanding how the students use this tool in their daily lives and the benefits and risks to their learning.

We received 54 complete responses from students who are known to the university’s disability and inclusion team. Our preliminary analysis revealed several interesting trends that could benefit learning and teaching practice, student support, and university policies on using AI.

Training on specialist software and assistive technology

The existing provision of specialist software and assistive technology training through DSA’s (Disabled Students’ Allowance) may provide various useful lessons regarding the initial take-up and sustained usage of assistive technology.

However, ensuring that the training provided is accessed, effective, adapted to individual needs and is available at the right time during students’ studies is challenging.

Support for student learning and daily lives

Of the respondents, 46% felt that GenAI tools were well designed to accommodate their specific needs; 59% of the respondents reported that they have used GenAI to explain concepts.

The potential for using GenAI tool such as ChatGPT as a “private tutor” is an area of growing interest, however when asked directly in our survey, only 9% of students claim to have used GenAI in this way.

While some students report being willing to experiment with virtual advice, less than 30% are currently likely to use GenAI for non-study related advice such as social, relationship, and mental health.

Learning and teaching

For the learning and teaching aspect, students feel that giving personal or pastoral advice, delivering teaching and marking summative assessments are the least appropriate uses of GenAI by a university.

We found that 50% of the respondents were more worried about being accused of plagiarism since the emergence of GenAI, while many feel “about the same” (48%). Most students are not sure if any GenAI detection technology can work as claimed (59%).

Future aspirations and concerns

Most respondents (91%) felt that “keeping up with current technologies” was somewhat or very important to their course and career. Most were unsure about the impact of GenAI on their current career aspirations, although this view tended to get more negative in the longer term (>5 years’ time).

Students have a clear level of concern around the wider effects and abuses of GenAI technology: from misinformation, scams, impact on artists and copyright, to uses in government, surveillance, harassment and more.

Conclusion and next steps

The preliminary results of our study provide insights into the experiences of dyslexic, neurodivergent and disabled students with GenAI technology. When comparing the results with another survey which did not specifically target disabled students, our respondents are more likely to use GenAI for study tasks such as explaining concepts, suggesting ideas and summarizing articles.

However, both groups of students have a similar view on the use of GenAI in assessment, that is GenAI should be used sparingly or not at all in assessments. From the teaching practice perspective, assessments may be designed to be “AI-proof,” however there is clearly a concern that any alternatives may result in assessments that are potentially less inclusive or less amenable to individualized reasonable adjustments where needed.

Social and ethical concerns will likely continue to require ongoing debate and discussion, particularly from courses that seek to make greater use of GenAI to enhance student learning experience.