For Reference
Hypotheses & Research Questions Restated
Does the AI training programme increase attendees' AI literacy?
Is there any difference of interest in AI study based on SES?
Do STEM students show higher AI literacy than non-STEM students before and after the training?
What do participants perceive as barriers and enablers for them to study AI following attendance of the training programme?
How do career expectations influence their choices around postgraduate study after attending the training programme?
How does institutional policy impact interest in AI studies after attending the training programme?
Mixed-Methods Evaluation
Key Conclusions
This study provides a mixed-methods evaluation of a short AI training delivered by the Jean Golding Institute (University of Bristol, UK), offering insight into how AI training may shape students' AI literacy, perceptions, and, potentially, educational aspirations. The main aim is to understand whether AI training may impact real and/or perceived barriers to AI doctoral studies. By combining survey data, qualitative interviews, and content analysis, the findings highlight a consistent pattern: whilst training is effective in increasing AI knowledge/literacy, it is not sufficient, in isolation, to transform longer-term educational pathways, particularly regarding postgraduate study.
Literacy Gains
Quantitative findings demonstrate a clear improvement in AI literacy following training — supported by both raw scores (+16.2pp, p=0.024) and IRT-based measures (+0.5 SD).
Confidence as User
Qualitative insights show that increased familiarity with concepts enhances confidence in engaging with AI — primarily as users within existing disciplines, rather than towards a potential PhD.
Employability Frame
AI literacy enhancement does not necessarily translate into increased interest in AI postgraduate study. Participants frame AI engagement as a means of improving employability, not as an academic pathway.
A key contribution is identifying the gap between knowledge acquisition and postgraduate educational aspirations. Across qualitative data, three interrelated factors emerge: the perceived relevance of AI to one's current field, confidence in technical skills (particularly coding), and uncertainty about career pathways in AI. For non-STEM students, AI continues to be perceived as implying a shift in competencies, limiting willingness to pursue further AI study despite learning outcomes.
The Inequality Risk
In contrast, STEM students not only begin with higher baseline literacy but may also benefit more from the training, suggesting that AI training may inadvertently reinforce existing inequalities rather than reducing them. The highly skewed sample — with minimal representation from lower socio-economic backgrounds — highlights that engagement with AI training opportunities may already be unequal at the point of access.
This study also shows the impact of contextualisation. Participants frequently expressed uncertainty about how AI is linked to their discipline and career aspirations, suggesting that technical training alone may not be sufficient without clear links to real-world career and disciplinary relevance.
Limitations
- —Small sample (n=13) limits generalisability and constrains statistical analysis, particularly for socio-economic differences.
- —Short timeframe between pre- and post-training measures means only immediate effects could be measured; no long-term follow-up.
- —Undergraduate sample limits the ability to explore questions related to AI doctoral study and admissions policies.
- —SES skew: 12/13 participants from Higher SES — formal H3 hypothesis testing was not possible.
- —Gender inference via LLM relies on probabilistic name associations, not self-identification, risking misclassification.
Future Directions
- +Longitudinal design to better understand how AI engagement evolves over time and across diverse populations.
- +Purposive sampling targeting underrepresented groups and current/prospective PhD candidates who aspired to AI studies but were unable to proceed.
- +Training with contextualisation — explicitly highlighting career and disciplinary applications of AI to support sustained engagement.
- +Intersectional analysis of race/ethnicity/migration and career pathway decisions, which emerged as a theme warranting further investigation.
Summary Statement
This study demonstrates that AI training may be an effective tool for AI literacy, but it does not, in isolation, address the broader barriers shaping students' engagement with AI as a field of study. Addressing structural and institutional barriers to AI doctoral studies requires a more holistic approach that takes into account literacy development and contextualisation, whilst also considering inclusivity and support for diverse learners.