Category: ARP

This final post answers the questions I received after the presentation.

1. Could you expand slightly upon the social justice element of the project please?

The social justice element of this project is about access and fairness in how coding is taught and learned on the course. Students arrive with very different levels of prior experience, shaped by factors like schooling, socioeconomic background, confidence, language, and access to technical provision and computers. AI-assisted coding was initially assumed to lower barriers and make code-based work feel more accessible to students who might otherwise feel excluded.

The research however, highlighted that if vibe coding is left informal or implicit, it can exacerbate or create new inequalities. Students with more technical confidence may benefit quickly, while others are left unsure whether using AI is allowed, how to use it well, or whether it undermines their learning. The takeaway from this research is to focus on making those expectations visible and teaching AI-assisted coding as a supported practice, so that it becomes a shared learning and coding tool rather than a hidden advantage. Framing vibe coding pedagogically is about creating fairer conditions for learning, not just faster outcomes.

2. What are your thoughts about the ethics of AI teaching? We acknowledge this could be an essay in its own right – so what are the main considerations in the context of your ARP?

The ethics of AI in teaching, in the context of this project, are mainly about transparency, care, and responsibility rather than whether AI should be used at all. AI-assisted coding is already present in students’ workflows, so the ethical question becomes how educators respond to that reality. UAL’s position on AI and guidance for students already sets out terms for our course to align to; embracing ‘AI through a Curious, Critical, and Compassionate lens’: https://www.arts.ac.uk/about-ual/learning-and-teaching/digital-learning/ai-and-education

Key considerations include being clear about when and how AI use is appropriate, avoiding punitive or ambiguous policies, and ensuring students are not disadvantaged for either using or avoiding AI (again, echoed by UAL’s student guide https://www.arts.ac.uk/about-ual/learning-and-teaching/digital-learning/ai-and-education/student-guide-to-generative-ai). In the context of coding, there is also an ethical responsibility to support learning-to-code rather than encourage dependency, by helping students understand what AI-generated code is doing.

Finally, there are broader ethical concerns around data, surveillance, sustainability, and power, particularly when AI tools are embedded into platforms students are required to use. Within this ARP, the focus is on designing teaching approaches that are transparent, supportive, and aligned with learning aims, while remaining attentive to the wider institutional and wider contexts in which AI operates

3. Thinking about data analysis, what worked – and what was challenging – in this? And how was it analysing a written interview alongside a text interview?

What worked well in the data analysis was the depth of reflection the interviews enabled, even with a very small sample. The semi-structured format allowed participants to surface concerns around confidence, legitimacy, and learning that may not have emerged through a more task-based or survey-led approach. Analysing the interviews comparatively, across a current student and a recent graduate, also helped highlight how quickly issues around AI are shifting.

The participant that responded in writing rather than through a live interview removed opportunities for follow-up questions, clarification, and conversational nuance. At the same time, the written response was often more considered and precise, offering insights that might not have surfaced in real-time discussion, and clearly the participant felt more comfortable with this format. This evidences a need for more inclusive and flexible methods of data collection.

Analysing the written interview alongside the transcribed spoken interview required treating them slightly differently, while still looking for shared themes. Rather than seeing this as a limitation, it reinforced the value of flexible methods and informed my thinking around more open and creative approaches of data collection for future research.

This final blogpost presents the project presentation, shared here as a PDF for in-blog viewing. While the original PowerPoint version includes two short videos, the presentation has been designed so that the PDF remains fully legible and informative without them. Viewing the videos is therefore optional and not required to understand the project.

The presentation outlines the research question and rationale for the study, situating it within the context of BA Fine Art: Computational Arts and the emergence of vibe coding within the course. It then documents the interview-based research method, reflecting on the methodological choices made and how these could be developed further. The final section presents the research findings and considers future pedagogical approaches and potential interventions in response to those findings.

Link to OneDrive for SharePoint Presentation

PDF:

The findings from this small-scale study suggest that AI-assisted coding, or “vibe coding,” cannot be assumed to be neither universally positive nor negative towards learning. Instead, its pedagogical impact is shaped by when and how is introduced on an individual basis. The contrasting experiences of the two participants highlight a key issue for Computational Arts education: how the uneven distribution of prior coding competence greatly affects the experience of using powerful AI tools for the first time.

Participant A’s experience points to a potential social justice issue for students without a privileged background knowledge of coding. Encountering vibe coding at the very start of their studies, without foundational coding knowledge or structured guidance, led to confusion and a loss of confidence. AI-generated code appeared opaque and difficult to claim ownership over, raising ethical and creative concerns about authorship and understanding. In this context, vibe coding risked accelerating towards finished assignments without supporting comprehension, producing what felt like progress without learning. A crude analogy is that of a beginner driver being told that they can just take taxis as it will get them to their destination faster. Beginners learning alongside more competent coders could feel further discouraged and lacking confidence if they felt like they needed the AI crutch to fulfil the same assignments.  

Participant B’s experience, by contrast, shows how vibe coding can function as an accelerator once a baseline level of competence is established. Having learned to code prior to the availability of AI assistants, Participant B was able to use conversational coding tools to extend their practice, overcome creative plateaus, and shift focus from overcoming technical limitations towards conceptual intent. Here, AI supported learning rather than replacing it.

These findings suggest a clear pedagogical implication: vibe coding should not be treated as neutral nor left for students to discover for themselves. Doing so risks reproducing existing inequalities, where students with prior experience benefit disproportionately, while novices experience anxiety, dependency, or disengagement. An inclusive approach requires designing teaching according to different levels of competence.

One potential intervention is the introduction of early diagnostic activities that assess students’ confidence, prior exposure, and understanding of coding concepts to inform individual support. Alongside this, vibe coding could be introduced in a structured and reflective way. A new priority would be to bring it in first as a learning tool that supports understanding, before introducing more advanced functions that autocomplete assignments, foregrounding understanding over output. From a pedagogical perspective, the teaching team may need to conceptually separate coding as a technical skill, and coding as a creative medium, to not always run ahead with the latter. 

Such an approach also responds to wider critiques of AI in education, including those articulated in Current Affairs (Purser, 2025). The article warns of “cognitive debt,” where reliance on AI produces an illusion of engagement while eroding underlying skills. While this critique is persuasive it risks flattening all uses of AI into a single systemic failure and issue for HE. The findings of this study suggest a more nuanced position: the danger lies not in AI itself, but in pedagogical models that either turn a blind eye to AI and leave students discovering it for themselves without support, or models that place too much focus on final outputs and assessment criteria and that are fundamentally ill-equipped to monitor and support developmental trajectories.  

In Computational Arts, where learning is already exploratory, non-linear, and practice-based, the challenge is not to ban vibe coding, but to bring about reflection, intentionality, and also to explore the benefits of coding coding without AI support as part of the learning experience. Schoolteacher Carl Haefemeyer makes a useful analogy saying that “[learning] is like weightlifting. You wouldn’t bring a forklift to the gym. The goal isn’t to get the bar up in the air, the goal is to build your muscle by lifting the bar.” (MPR News, 2025).

Ultimately, a more inclusive pedagogy that helps computational artists grapple with vibe coding requires shifting away from its image as a shortcut to that of a software tool to be dissected in class with support and critical reflection built into the curriculum. The field of Software Studies may provide a useful theoretical framework, with authors such as Matthew Fuller (2008) and Lev Manovich (2013) equipping us with the conceptual tools for unpacking complexities of software applications. Additionally, some pedagogical  and pastoral attention needs to be placed on preexisting inequalities within student cohorts, as these are only amplified by AI tools, rather than resolved by them. 

References

Fuller, M. ed. (2008). Software studies: A lexicon. Mit Press.

Manovich, L. (2013). Software takes command (p. 376). Bloomsbury Academic.

Purser, R. (2025). AI is Destroying the University and Learning Itself. [online] Currentaffairs.org. Available at: https://www.currentaffairs.org/news/ai-is-destroying-the-university-and-learning-itself. [Accessed 8 Dec 2025]

MPR News, (2025). AI in schools: St. Paul teacher says it’s ‘like bringing a forklift to the gym’. [online] MPR News. Available at: https://www.mprnews.org/episode/2025/09/09/ai-in-schools-st-paul-teacher-says-its-like-bringing-a-forklift-to-the-gym# [Accessed 3 Feb. 2026].

This study explored how AI-assisted coding (“vibe coding”) is experienced by students within Computational Arts pedagogy, with particular attention to learning, confidence, creativity, and inclusivity. Two contrasting participant accounts offer insight into how prior experience and timing of exposure shape these outcomes. Participant A is a current Year 1 student (2025–26), encountering vibe coding at the very start of their higher education. Participant B is a recent graduate who began their studies in 2021–22, prior to the widespread adoption of AI-assisted coding, and later integrated these tools into their practice.

A key difference between the participants concerns initial confidence and comprehension. Participant A, with no prior experience of writing code, described early use of ChatGPT as alienating: autogenerated code “didn’t really mean anything” and reduced rather than increased confidence. This sense of opacity prompted the student to move away from ChatGPT and use Google Gemini’s Guided Learning feature, which supported understanding and increased confidence when asked questions. As Participant A reflected, this reduced the anxiety of producing work they could not explain, highlighting a strong ethical orientation towards authorship and responsibility.

By contrast, Participant B reported a largely positive trajectory, underpinned by existing coding literacy developed before the emergence of vibe coding. Having already internalised foundational programming logic, they used ChatGPT conversationally to extend their practice, moving from “janky” single-script programs towards modular systems with functions, multiple files, and classes. Participant B described this as helping them overcome a learning plateau, increasing both confidence and ambition: they felt capable of developing more complex work than they would previously have attempted.

Both participants emphasised the importance of understanding code beyond AI output, though for different reasons. Participant A argued that creative ownership requires knowing “what I am writing and why,” likening coding to traditional artistic mastery. Participant B framed foundational knowledge as necessary for maintaining creative control, diagnosing problems, and preventing AI from imposing its own aesthetic or structural logic. In both cases, AI was positioned not as a replacement for learning, but as something that must be carefully integrated into the learning process.

In terms of creativity and accessibility, both accounts suggest that vibe coding lowers barriers to realising code-based ideas. Participant A described AI as a catalyst for ideation, enabling the recombination of “mediocre” suggestions into personally meaningful outcomes. Participant B described a shift from starting with technical feasibility (“what can I do?”) to starting with conceptual intent (“what do I want to make?”), which they identified as transformative for their artistic practice. However, Participant B also noted the risk of distraction, overproduction, and aesthetic homogenisation without clear creative guardrails.

Taken together, these findings suggest that vibe coding can enhance inclusivity and creative ambition, but that its pedagogical value is highly contingent on timing, prior experience, and framing. For novice students, unmediated AI-generated code risks undermining confidence and understanding, whereas guided, reflective use may support learning. For more experienced students, vibe coding appears to function as an accelerant rather than a shortcut. These contrasts underscore the need for intentional pedagogical strategies that treat AI-assisted coding not as neutral infrastructure, but as a situated learning tool requiring critical and ethical engagement.

This research project investigated student experiences of AI-assisted coding within BA Fine Art: Computational Arts, using semi-structured interviews (Wilson, 2012). The study identified a social justice issue within computational arts education: how uneven access to prior technical training, language proficiency, and confidence with coding can shape students’ ability to benefit from emerging AI tools. These disparities directly affect student experience, particularly in a programme that aims to be inclusive while supporting highly individualised artistic practices.

Semi-structured interviews were selected as an appropriate research design for exploring complex, subjective experiences that cannot be meaningfully reduced to quantitative data (Beck and Manuel, 2008). The interview questions provided a flexible structure that enabled participants to reflect on prior coding experience, first encounters with AI tools, and perceived impacts on learning and creative agency. However, the study revealed important methodological considerations. One participant requested access to the questions in advance due to English not being their first language, highlighting accessibility needs that were not fully anticipated. This participant ultimately chose to respond in writing rather than through an online interview, demonstrating how research methods may need to adapt to participants’ circumstances.

Rather than viewing this deviation as a failure, the project draws on Mike Michael’s concept of “idiotic methodology” (Michael, 2012; 2013), which encourages researchers to embrace unexpected methodological turns as productive. In retrospect, a more open-ended, creative instrument may have been better suited to an arts education context. Future iterations of the study would therefore adopt a cultural probe approach (Gaver and Dunne, 1999), reshaping the research questions into drawing, annotating, and reflective exercises that can be completed offline. Drawing on my prior experience contributing to probe-based research (Gaver, 2016), this approach would support more inclusive and imaginative forms of participation.

Findings from the interviews suggest that challenges surrounding vibe coding are less about the presence of AI itself and more about uneven starting points in coding education. Both participants emphasised the need for a clear foundational understanding of code in order to use AI tools critically and effectively. This highlights a gap in the course team’s awareness of students’ prior training, with implications for curriculum design and support structures.

Although the study was limited to two participants due to timing constraints, the inclusion of both a current first-year student and a recent graduate enabled a valuable cross-generational comparison. When situated alongside larger-scale studies of student–AI interaction (e.g. Geng et al., 2025), this project demonstrates the value of small-scale, qualitative inquiry while also identifying the need for broader participation and task-based methods in future research.

References

Beck, S. E., & Manuel, K. (2008). Practical research methods for librarians and information professionals. New York, NY: Neal-Schuman.

Gaver, B., Dunne, T. and Pacenti, E. (1999). Design: cultural probes. interactions, 6(1), pp.21-29.

Gaver, W., Ovalle, L. and Plummer-Fernandez, M. (2016). Tilly and the Myth of Energy Independence.

Geng, F., Shah, A., Li, H., Mulla, N., Swanson, S., Raj, G.S., Zingaro, D. and Porter, L. (2025). Exploring Student-AI Interactions in Vibe Coding. arXiv preprint arXiv:2507.22614.

Michael, M. (2012). De‐signing the object of sociology: Toward an ‘idiotic’methodology. The Sociological Review, 60, pp.166-183.

Michael, M. (2013). The idiot. Informática na educação: teoria & prática, 16(1).

Wilson, V. (2012). Research methods: interviews.

Wakeford, N. ed. (2012). Inventive methods. London: Routledge.

The literature informing this project spans overlapping areas: the emergence of vibe coding as a technical paradigm, its implications for education and higher education, and methodological approaches relevant to creative and exploratory research contexts.

Recent technical literature defines vibe coding as a shift from direct code authorship towards conversational intent mediation through AI systems (Meske et al., 2025; Sarkar and Drosos, 2025). These accounts characterise vibe coding as lowering the cognitive load of programming, allowing users to prioritise goals and outcomes over syntax. While this reframing is often positioned as empowering, several authors caution that it introduces new dependencies and long-term maintenance challenges, particularly where users lack foundational coding knowledge (Maes, 2025; Ray, 2025).

Emerging educational research begins to examine how these dynamics play out in learning contexts. Studies of student–AI interaction suggest that AI-assisted coding can increase confidence and task completion for novice programmers, while also obscuring gaps in conceptual understanding (Geng et al., 2025). Horvat (2025) and Chow and Ng (2025) highlight the pedagogical opportunity of repositioning learners as creative directors rather than technical operators, though these benefits are shown to be highly context-dependent. Within academia more broadly, Crowson and Celi (2025) frame vibe coding as a pragmatic response to resource constraints, raising questions about equity, authorship, and assessment. Finally, critical perspectives on AI in higher education warn against unreflective adoption that risks eroding learning, agency, and trust (Purser, 2025).

Within computational art education, earlier work emphasises code as a creative medium rather than a purely technical skill (Levin and Brain, 2021), with tools and interfaces playing a significant role in shaping student engagement (Mcnutt et al., 2023). This provides a useful lens for considering vibe coding not simply as a shortcut, but as another approach in the evolving relationship between artists and software tools.

Taken together, this literature suggests a need for situated, student-centred research that examines how AI-assisted coding intersects with accessibility, inclusion, and creative practice, and highlights a gap in research concerning computational arts students. 

Bibliography

Vibe coding: definitions and technical perspectives

Cress, L. (2025). ‘Vibe coding’ named word of the year by Collins Dictionary. BBC News. Available at: https://www.bbc.co.uk/news/articles/cpd2y053nleo.

Horvat, M. (2025). What is Vibe coding and when should you use it (or not)?. Authorea Preprints.

Kaparthy, A. (2025). There’s a new kind of coding I call “vibe coding”… 2 February. Available at: https://x.com/karpathy/status/1886192184808149383 [Accessed: 1 Dec 2025].

Maes, S.H. (2025). The gotchas of AI coding and vibe coding: It’s all about support and maintenance. Available at https://www.researchgate.net/profile/Stephane-Maes-2/publication/391568491_The_Gotchas_of_AI_Coding_and_Vibe_Coding_It’s_All_About_Support_And_Maintenance/links/6832a3e76b5a287c3044caeb/The-Gotchas-of-AI-Coding-and-Vibe-Coding-Its-All-About-Support-And-Maintenance.pdf [Accessed: 2 Dec 2025]

Meske, C., Hermanns, T., von der Weiden, E., Loser, K.U. and Berger, T. (2025). Vibe coding as a reconfiguration of intent mediation in software development: Definition, implications, and research agenda. arXiv preprint arXiv:2507.21928.

Ray, P.P. (2025). A Review on Vibe Coding: Fundamentals, State-of-the-art, Challenges and Future Directions. Authorea Preprints.

Sarkar, A. and Drosos, I. (2025). Vibe coding: programming through conversation with artificial intelligence. arXiv preprint arXiv:2506.23253.

Vibe coding in Higher Education and academia

Chow, M. and Ng, O. (2025). From technology adopters to creators: Leveraging AI-assisted vibe coding to transform clinical teaching and learning. Medical Teacher, pp.1–3.

Crowson, M.G. and Celi, L.C.A. (2025). Academic Vibe Coding: Opportunities for Accelerating Research in an Era of Resource Constraint. arXiv preprint arXiv:2508.00952.

Geng, F., Shah, A., Li, H., Mulla, N., Swanson, S., Raj, G.S., Zingaro, D. and Porter, L. (2025). Exploring Student-AI Interactions in Vibe Coding. arXiv preprint arXiv:2507.22614.

Horvat, M. (2025). What is Vibe coding and when should you use it (or not)? Authorea Preprints.

Purser, R. (2025). AI is Destroying the University and Learning Itself. [online] Currentaffairs.org. Available at: https://www.currentaffairs.org/news/ai-is-destroying-the-university-and-learning-itself. [Accessed 8 Dec 2025]

Computational art and creative coding education
Levin, G. and Brain, T. (2021). Code as creative medium: A handbook for computational art and design. MIT Press.

Mcnutt, A.M., Outkine, A. and Chugh, R. (2023). A study of editor features in a creative coding classroom. Proceedings of the CHI Conference on Human Factors in Computing Systems, pp.1–15.

Methodological references
Beck, S.E. and Manuel, K. (2008). Practical research methods for librarians and information professionals. Neal-Schuman.

Wilson, V. (2012). Research methods: interviews.
Gaver, B., Dunne, T. and Pacenti, E. (1999). Design: cultural probes. interactions, 6(1), pp.21–29.

Michael, M. (2012). De-signing the object of sociology: Toward an ‘idiotic’ methodology. The Sociological Review, 60, pp.166–183.

Michael, M. (2013). The idiot. Informática na educação: teoria & prática, 16(1).

Wakeford, N. (ed.) (2012). Inventive methods. Routledge.

Gaver, W., Ovalle, L. and Plummer-Fernandez, M. (2016). Tilly and the Myth of Energy Independence.

This post contains the research documents that underpin my study, including the research questions, participant information sheet, and consent form. Together, these materials define the ethical and methodological framework for an exploratory investigation into student experiences of AI-assisted coding within BA Fine Art: Computational Arts.

The interview questions are designed to elicit student perspectives on learning, confidence, creativity, and inclusion, while the participant documentation sets out how issues of consent, confidentiality, data protection, and participant wellbeing are addressed. The questions were developed for use in semi-structured interviews (Wilson, 2012); however, as discussed in my reflective post, their circulation in advance also prompted an alternative mode of response, leading to an unanticipated adaptation of the research method.

These documents were shared with participants prior to data collection, ensuring that ethical considerations were embedded from the outset of the study.

References
Wilson, V. (2012). Research methods: interviews.

Survey Questions

Consent Form

Participant Information Sheet

This post presents the Ethical Action Plan for my research project, outlining the key ethical, methodological, and practical considerations that underpin the study. The project explores how AI-assisted coding (“vibe coding”) is reshaping student experience within BA Fine Art: Computational Arts, with particular attention to accessibility and inclusivity. The plan addresses issues of informed consent, data protection, participant wellbeing, and my own positionality as course leader and researcher. It also clarifies the scope of the project setting out a feasible research design intended to inform future pedagogical development .

Title:
Vibe Coding in Computational Arts pedagogy: Addressing the accessibility, inclusivity, and learning implications of AI-assisted coding in Fine Art education.

Here is the Ethical Action Plan as a downloadable PDF:

Vibe Coding was recently named Word of the Year by Collins Dictionary (Cress, 2025). The term was coined in February 2025 by OpenAI co-founder Andrej Karpathy to describe how AI can enable programmers to “forget that the code even exists” while building software (Karpathy, 2025).

This shift has significant implications for students learning to code. While AI-assisted coding may lower technical barriers and increase accessibility for students who do not identify as programmers, it also raises questions about dependency, authorship, and how coding skills are valued and assessed. These tensions are particularly acute in computational arts education, where students may need to build functional programs in order to realise creative work, without necessarily aspiring to technical mastery. Understanding how vibe coding shapes student confidence, agency, and inclusion therefore represents a social justice issue within this academic context.

This study will explore these questions through interviews with students at different stages of the BA Fine Art: Computational Arts programme, enabling comparative insights into how prior experience, confidence, and progression shape perceptions of AI-assisted coding. By focusing on student accounts, the research aims to foreground lived experience rather than abstract debates about technology.

As Course Leader, my role is to introduce coding alongside other technical skills in ways that are accessible, inclusive, and sustainable. AI-assisted coding is already in use on the course, with a marked increase in uptake during 2025–26. This project responds directly to that change, seeking to inform pedagogical decision-making while remaining attentive to equity and inclusion.

While debates around AI-generated code are beginning to emerge in higher education (Meske et al., 2025; Maes, 2025; Ray, 2025; Sarkar and Dross, 2025), there remains limited research focused on student experience and inclusive practice within creative coding contexts (Geng et al., 2025). This study addresses that gap.

References

Cress, L. (2025). ‘Vibe coding’ named word of the year by Collins Dictionary. https://www.bbc.co.uk/news/articles/cpd2y053nleo.

Geng, F., Shah, A., Li, H., Mulla, N., Swanson, S., Raj, G.S., Zingaro, D. and Porter, L. (2025). Exploring Student-AI Interactions in Vibe Coding. arXiv preprint arXiv:2507.22614.

Kaparthy, A. (2025) There’s a new kind of coding I call “vibe coding”… 2 February. Available at: https://x.com/karpathy/status/1886192184808149383 [Accessed: 1 Dec 2025]

Meske, C., Hermanns, T., von der Weiden, E., Loser, K.U. and Berger, T. (2025). Vibe coding as a reconfiguration of intent mediation in software development: Definition, implications, and research agenda. arXiv preprint arXiv:2507.21928.

Maes, S.H. (2025). The gotchas of ai coding and vibe coding. it’s all about support and maintenance [online]

Ray, P.P. (2025). A Review on Vibe Coding: Fundamentals, State-of-the-art, Challenges and Future Directions. Authorea Preprints.

Sarkar, A. and Drosos, I. (2025). Vibe coding: programming through conversation with artificial intelligence. arXiv preprint arXiv:2506.23253.