Methods
Qualitative Analysis
A deductive thematic analysis approach was applied to interview transcripts (Clarke and Braun, 2017). Codes were defined according to prior literature on AI policies, admission policies, AI literacy, and socio-economic barriers. Abductive content analysis was used for free-text survey responses.
Thematic Analysis
Applied to interview transcripts. Codes derived from prior literature and pilot interviews, reviewed asynchronously and in dedicated team meetings.
Content Analysis
Abductive content analysis of free-text survey responses (Humble and Mozelius, 2022). Frequency analysis enabled efficient categorisation across the dataset.
⚠ Important Caveat
The JGI training focused on AI literacy rather than all barriers identified in the literature. Some codes (e.g., gender inequality, confidentiality) were not addressed in post-training interviews. Additionally, the undergraduate sample constrained analysis of doctoral admission policies.
AI Literacy: Barriers & Enablers
1.1 Technical Skills Acquisition
Technical skills were the most frequently mentioned factor in participants' evaluation of what encourages or discourages them from engaging with AI studies. Being knowledgeable of general concepts such as LLMs or neural networks had a positive influence on attitudes towards engaging with AI as users within areas related to their studies.
Having prior coding skills and experience helped participants feel more confident in expanding their AI knowledge and considering further studies. Some expressed being less inhibited after taking the hands-on session that introduced them to practical aspects of coding. Participants particularly appreciated having AI substreams broken down for an easier learning experience, and emphasised the importance of understanding the mechanism behind AI — helping them make better connections to the work they do.
Conversely, having little or no coding background and limited AI knowledge was a factor that made participants less willing to explore further — discouraging engagement with the training and dampening interest in AI postgraduate studies. It is important to consider that these skills are developed over years through prior educational decisions, personal interest, and environmental influence.
"So the way I use LLMs was also affected in terms of realising and understanding how something works. You know how to better use it based on its underlying mechanism."
Theme: Technical Understanding"Maybe it's because of like I don't have that technical background of it. So like I got the base of it, but still like I'm not — probably I won't go for an AI postgrad."
Theme: Technical Barrier"Back in my high school... I took like Politics, Geography, History and Economics. I never got any chance to like study Python... because it seems so complicated to a person from a social science major."
Theme: Educational Pathway1.2 Capacity Building & Training Accessibility
Participants highlighted that being part of the university allowed them to access training programmes that better introduced them to AI. Whilst one training some attended was more about the ethics of using AI, the "Getting Started in AI" training was more technical. Interviewed participants valued this exposure as they identified its impact on their awareness of potential applications of AI within their current studies and the future. Even though that was not a decisive factor in their decision to pursue AI studies, it offered awareness of the field and what it might include for students who were considering a related postgraduate degree.
Being invited through email and posters, promotion of a free training, and being in a friendly environment encourages people to engage with AI more — allowing people to be better informed and technically equipped, increasing confidence in pursuing a degree in AI. Self-efficacy improved from 36% (pre-training) to 63% (post-training) in free-text responses, consistent with prior literature on confidence and AI engagement (Cowit and Fiesler, 2024).
"I also took the course to kind of help me decide whether I'd like to pursue a Masters of Science in AI rather than Masters of Science and Engineering, which was my original path."
Theme: Decision-Making Support"Being invited through email and posters, promotion of a free training, and being in a friendly environment encourages people to engage with AI more."
Theme: AccessibilityCareer Expectations & Postgraduate Choices
2.1 Relevance to Current Pathway
For some interviewed participants, engaging with AI as a user of those systems was more important than studying AI itself. The application of AI was not clear to them in their fields, e.g., social sciences and marketing. This is also connected to the skills they have obtained in their pathway, which are less hard-science oriented. For these participants, the concern was to use AI ethically, understand its many applications in their work, and have the knowledge to make full and optimal use of it. This aligns with Yin and Goh's (2024) finding that a key driver for learning AI is utility rather than pure technical interest.
"I don't know if taking a career, but learning more about it, getting more into more courses like this one."
2.2 Future Career Ambiguity
Whilst future career may not be explicitly linked to barriers to doctoral AI studies, all participants discussed their career expectations as part of their chosen degree and potential postgraduate studies. The ambiguity about future careers was one of the main factors influencing decisions about AI postgraduate study. Concerns varied between the competitiveness of the field, uncertainty regarding the required expertise within it, and the lack of understanding of how AI connects to areas such as the social sciences. For participants with an engineering background, the concern was about the limited number of job opportunities in AI in comparison to a more general engineering specification.
"I chose [this degree] because... I thought it's very broad here in England and I can go into any industry that I want. They will always need a mechanical engineer for something."
Theme: Career Security"There's always the risk of something being true right now, but then completely changing in a year's time because of the nature of the field where everything is changing rapidly."
Theme: Field Volatility
Rapid AI change creates uncertainty about whether skills remain relevant — a key psychological barrier to long-term commitment.
"And actually, one of the guys [delivering the training] was [from abroad], so we had a chat in [their home language]... I'm not from England, so that's really good to know that anyone can get into it."
Theme: Intersectionality
This quote may imply a degree of intersectionality between race/ethnicity/migration and chosen career path, hence higher education studies. However, the small number of participants and the narrow focus of the study mean that more research should be conducted to further understand these intersections.
Institutional Policy Impact
3.1 Perceptions of UoB AI Policy
Most interviewed participants found the UoB AI policy to be clear. The examples provided in the policy addressing the limits of AI use offered guidance for students. However, for others, they indicated that they would appreciate more clarity. These perspectives were related to exposure to the policy — participants who were familiar with the guidebook were more positive about its clarity.
Exposure to the policy through lecturers, emails, and personal curiosity influenced understanding. Responses fell on a spectrum from "too strict" to "a fair policy that respects academic integrity without inhibiting AI use." For most students, their usage of AI was not affected by the policy — this was especially evident with students who used AI less than what the policy allows for. One interviewee highlighted how more flexibility with the policy could help international students with language barriers.
"Seeing that it had been considered by the university as well and it wasn't that black or white, yes or no type of situation, encouraged me. I was happy to see that."
Theme: Policy Clarity"Sometimes I feel a bit limited because, for coursework obviously I'm not going to use it like to copy and paste... So even if I want to use it for personal use... I'll always be thinking of, oh, what if there's a policy against what I'm doing?"
Theme: Over-generalised RestrictionScope of Policy Analysis
This study cannot claim a thorough and structured policy evaluation. Admission policies related to doctoral studies in AI could not be investigated due to sampling constraints (undergraduate sample). A rare input on this matter also mentioned that the restrictive use of AI in the university raises suspicions about the ethical considerations in using AI outside the university for work or personal matters. The literature on admission policies addresses these topics more thoroughly.