Titulo Estágio
VibeCoding - Exploring Human-AI Interaction in Programming
Local do Estágio
Associação Fraunhofer Portugal Research, Rua Alfredo Allen, 455/461, 4200 - 135 PORTO
Enquadramento
As AI-powered tools for code generation, debugging, and documentation rapidly gain traction, programming is increasingly becoming a socio-technical activity shaped not only by the programmer’s intent but also by the behaviours and affordances of AI assistants. “Vibe Coding” is an emergent, loosely-defined practice among developers who work in dynamic, exploratory, and often playful ways with AI assistants (such as GitHub Copilot, ChatGPT, or Replit Ghostwriter), using them as thought partners, collaborators, or even provocateurs. These interactions are not solely task-oriented; they often involve affective, improvisational, and aesthetic dimensions.
Recent studies have highlighted the transformative impact of AI tools on programming workflows. For instance, GitHub Copilot has been shown to significantly enhance developer productivity by automating repetitive coding tasks and providing intelligent code suggestions (Bird et al., 2022). However, the integration of AI into programming also raises concerns about over-reliance on AI-generated code and the potential erosion of fundamental programming skills (Yilmaz & Yilmaz, 2023). Additionally, the use of AI assistants in educational settings has revealed mixed outcomes, with some students benefiting from improved computational thinking skills and motivation, while others struggle with understanding the underlying logic of AI-generated solutions (Lau & Guo, 2023). Zviel-Girshin (2024) explored the impact of AI coding tools on novice programmers, noting increased familiarity and satisfaction but also concerns about cheating and over-reliance. Sarkar et al. (2022) emphasised the distinct nature of LLM-assisted programming compared to traditional methods, highlighting challenges such as intent specification and code comprehension for non-expert users. Sergeyuk et al. (2025) provided a systematic review of Human-AI Experience in Integrated Development Environments, identifying key trends and the need for further research on personalization and ethical considerations.
As AI becomes a more active participant in the coding process, it is important to understand the changing role of the human developer. This dissertation aims to fill this gap by exploring the affective, aesthetic, and improvisational practices involved in vibe coding, and by deriving design implications for creating more expressive, inclusive, and satisfying AI tools.
Objetivo
This dissertation has two main goals:
- To investigate how programmers experience and make sense of coding with AI assistants, particularly in relation to affective, aesthetic, and improvisational practices (vibe coding).
- To derive design insights for creating AI tools that support more expressive, inclusive, and satisfying forms of programming.
Innovative aspects
This research will contribute to the growing body of work on human-AI collaboration in software development by adopting a human-centred and qualitative approach. It will examine coding not only as a technical activity, but as a lived experience shaped by values, habits, moods and relationships. The project will explore the role of “vibe” in coding interactions—a term encompassing mood, flow, aesthetic pleasure, and conversational tone—towards shaping more humane and engaging tools for developers. It will also build on the preliminary work conducted within Fraunhofer AICOS, where mixed-methods research has revealed rich insights into how GenAI is embedded into real workflows, and contribute to the work being developed in project ACHILLES.
Plano de Trabalhos - Semestre 1
Workplan:
1st semester
1- Literature review on coding with AI assistants: Conduct a review of existing literature to understand the current state of research on AI-assisted programming.
2- Write the intermediate document of the dissertation.
Plano de Trabalhos - Semestre 2
Workplan:
2nd semester
1- Fieldwork with developers: Engage with developers through methods such as surveys, contextual inquiries, observations, diary studies, probes and interviews to understand their practices with AI assistants.
2- Qualitative analysis: Analyse the collected data to uncover key patterns and tensions in how developers engage with AI tools.
3- Development of speculative design artefacts: Co-create speculative design artefacts (e.g., interface sketches, tool concepts) with participants (programmers) and/or explore them beforehand to present for reflection and critique.
4- Participation in ACHILLES work meetings: Accompany ACHILLES work meetings as necessary to stay aligned with ongoing research and developments.
5- Identification of design insights: Identify design insights for building AI-assisted development environments that support improvisational coding workflows.
6- Write the remaining chapters of the dissertation and review the intermediate document.
Condições
Student profile:
Inquisitive, curious and open mind;
Interest – and, ideally, experience – in human-centred design processes;
Willingness to perform field work in user research- co-creation and prototype assessment;
Proficiency in English and good domain of Portuguese for field work.
Observações
Bibliography:
Bird, C., Ford, D., Zimmermann, T., Forsgren, N., Kalliamvakou, E., Lowdermilk, T., Gazit, I. (2022). Taking flight with Copilot: Early insights and opportunities of AI-powered pair-programming tools. Queue, 20(6), 35–57.
Yilmaz, R., Yilmaz, F.G.K. (2023). The effect of generative artificial intelligence (AI)-based tool use on students’ computational thinking skills, programming self-efficacy and motivation. Computers & Education: Artificial Intelligence, 4, 100147.
Lau, S., Guo, P. (2023). From “Ban it till we understand it” to “Resistance is futile”: How university programming instructors plan to adapt as more students use AI code generation and explanation tools such as ChatGPT and GitHub Copilot. In Proceedings of the 2023 ACM Conference on International Computing Education Research-Volume 1, 106–121.
Zviel-Girshin, R. (2024). The Good and Bad of AI Tools in Novice Programming Education. Education Sciences, 14(10), 1089. https://doi.org/10.3390/educsci14101089
Sarkar, A., Gordon, A.D., Negreanu, C., Poelitz, C., Srinivasa Ragavan, S., Zorn, B. (2022). What is it like to program with artificial intelligence? arXiv preprint arXiv:2208.06213v2.
Sergeyuk, A., Zakharov, I., Koshchenko, E., Izadi, M. (2025). Human-AI Experience in Integrated Development Environments: A Systematic Literature Review. arXiv preprint arXiv:2503.06195v1.
Orientador
Ricardo Manuel Coelho de Melo
ricardo.melo@aicos.fraunhofer.pt 📩