Vibe Coding Entanglements – Repositioning Boundaries of Intention, Authorship, and Responsibility in Programming with Generative AI

Paper Title

Vibe Coding Entanglements – Repositioning Boundaries of Intention, Authorship, and Responsibility in Programming with Generative AI

Publication Info

• Topic area: Exploratory programming practices with generative AI and their implications for intention, authorship, and responsibility.
• Keywords: Vibe coding, generative AI, programming assistants, human-AI collaboration, intention, authorship, responsibility, co-creation, interaction design, boundary-making.

Background and Problem

• Problem / challenge: Traditional programming practices assume clear boundaries of control and authorship, but generative AI tools challenge these assumptions by introducing shared agency and interpretive dynamics. This creates ambiguity in intention, control, and responsibility.
• Significance: Understanding these dynamics is critical for designing effective AI programming tools and addressing ethical concerns like accountability and authorship in AI-assisted programming.
• Motivation and related work: Prior studies have explored AI tools like Copilot and CodeShaper, which influence productivity and intent articulation. However, they often overlook the relational and co-constitutive nature of programming with AI, leaving a gap in understanding how human and AI agencies interact and reshape programming practices.

Solution

• Proposed approach: The study introduces and examines "vibe coding" as a decentered, co-constituted programming practice where humans and AI collaboratively shape code through iterative interactions.
• Novelty:
  1. Conceptualisation of vibe coding as a relational and decentered programming practice.
  2. Study of three AI-based programming provocations to explore human-AI dynamics.
  3. Identification of design configurations that influence human-AI collaboration in programming.
  4. Ethical framing of accountability and responsibility in vibe coding practices.
• Procedure and key techniques:
  • Designed three programming assistants (provocations) with distinct interaction styles: compliant but biased, critical and argumentative, and visual sketch-based.
  • Conducted a workshop with 12 participants (programmers and designers) to explore these provocations through programming exercises.
  • Used reflexive thematic analysis to examine interactions, focusing on intention, control, and boundary-making.

Results

• Concrete findings:
  • Intention and control are not fixed but emerge relationally through iterative interactions between humans and AI.
  • Outcomes are co-constructed, requiring interpretive work to align user intent with AI-generated code.
  • Visual and textual modalities shape how ideas are expressed and interpreted, with unique challenges in translating dynamic behaviours.
• Advantage over baselines:
  • Highlights the non-neutrality of AI tools, showing how they influence intention and authorship beyond traditional command-response paradigms.
  • Provides a framework for designing AI programming tools that support co-creation and shared agency.
• Experiments / evaluation:
  • Three provocations were tested in a workshop setting with 12 participants divided into four groups.
  • Data sources included video recordings, written reflections, chat logs, and generated code.
  • Reflexive thematic analysis identified themes like intention, control, and shifting roles.
• Limitations and future work:
  • The study was conducted in an exploratory setting, not reflective of professional programming environments.
  • Short timeframe limits insights into long-term dynamics of human-AI collaboration.
  • Future work could explore these dynamics in real-world settings and over extended periods.

Summary

This paper investigates vibe coding as a co-constituted programming practice where humans and generative AI collaboratively shape code through iterative interactions. Using three design provocations, the study reveals how intention, control, and authorship are dynamically negotiated, challenging traditional human-centered programming paradigms. The findings highlight the importance of designing AI tools that support co-creation, transparency, and ethical accountability. By foregrounding process over outcome, the study provides a conceptual and practical framework for understanding and designing AI-based programming practices.