When you build software with an AI, what would make you truly own that code rather than just say you do?
When do students feel authentic ownership of code they co-created with AI?
This explores what has to be true for a student's sense of ownership over AI-co-created code to be real rather than just claimed, and the corpus shows much more about when that feeling is an illusion than about when it is earned.
This explores what has to be true for a student's sense of ownership over AI-co-created code to be real rather than just claimed. The corpus has no study of the positive case. It is much better at showing when ownership is an illusion, so most of what follows maps the traps, with a few clues about the way out.
Ownership comes in two kinds that come apart. Users routinely declare authorship socially ('I built this') without the felt, cognitive ownership that comes from knowing how the thing came to be Do users truly own the AI-generated content they produce?. The cause isn't dishonesty. The intermediate steps are opaque, and people build the story of what they did afterward. Student coding behavior shows the same gap. Across 19 vibe-coding students, 63.6% of interactions were testing the running prototype and only 7.4% touched code directly, and 90% of those were reading rather than editing Where do vibe coding students actually spend their debugging time?. The ownership these students report probably attaches to what the program does, not to what it is made of. It would only be tested when they have to explain or change the code.
The feeling of ownership still comes easily because fluent output reads as a signal of your own ability. Processing ease works as a metacognitive cue for 'I understand this' Does processing ease mislead users about their own competence?. Four mechanisms feed each other: blurry attribution, fluency, outsourced thinking and opaque pipelines How do AI tools trick users into overestimating their own skills?. One paper calls the result the LLM fallacy, where you fold AI output into your picture of your own skills, especially when the seam between you and the model can't be seen Do AI-assisted outputs fool users about their own skills?. Checking is costly, and studies find roughly 80% of AI output adopted unchallenged When do users stop checking whether AI output is actually backed?. Users in every language studied also follow confident answers whether or not they're right Do users worldwide trust confident AI outputs even when wrong?. So the strength of the feeling is a poor guide. The smoother the collaboration, the more convincing and the less trustworthy the sense of ownership.
The clues about authentic ownership all involve making the seam visible. Paired writers in shared editors wanted to see when, how and where AI had been used, both to understand each other's thinking and to verify the text Do writers want to see each other's AI prompts in shared editors?. A newsroom tool won adoption by binding every number and quote to its origin, so provenance rather than polish decided trust Can source traceability make AI writing trustworthy?. For a classroom, the closest thing to a test is whether a student can point to which parts were theirs, which were the model's, and why they trust each. That is an extrapolation, not a finding. No study here follows students' felt ownership of code after they debug, rewrite or defend it.
Two shortcuts don't work. Students working with a chatbot did better on practical tasks but contributed less dialogue and far fewer subjective perspectives, so they gained performance and lost some of their own voice Does chatbot interaction trade authenticity for better problem-solving?. You also can't read ownership off how someone talks to the AI. Conversation traits explained coding outcomes beyond prior achievement, but they weren't stable or transferable enough to count as learnable skills Can conversation patterns predict coding outcomes better than prior skill?.
Sources 11 notes
Research shows users declare authorship at a social level while lacking genuine cognitive ownership of AI-generated content. This dissociation arises from opaque intermediate steps and post-hoc narrative construction, not dishonesty, and leads to inflated self-assessments of independent competence.
Across 19 students, 63.6% of interactions involved testing the prototype while only 7.4% touched code directly. Of code interactions, 90% were reading rather than editing, suggesting students remain distant from implementation details.
High-quality AI output triggers a metacognitive heuristic: users experience fluency as a signal of their own capability, even though they didn't generate it. This self-directed fluency illusion systematically inflates perceived competence because LLMs optimize for fluency regardless of user understanding.
Attribution ambiguity, fluency illusion, cognitive outsourcing, and pipeline opacity combine to systematically misattribute AI outputs as user competence. The effect is multiplicative—each mechanism amplifies the others.
Research identifies a systematic cognitive attribution error where individuals integrate AI-generated outputs into their capability identity, believing they possess skills they don't actually have. This occurs when task output is seamless and fluent, obscuring the human-AI boundary.
Show all 11 sources
Users systematically accept AI outputs without verification because checking is costly and fluent output builds false confidence. This receiver-side surrender—measured in studies showing 80% unchallenged adoption—is what enables inflationary token systems to function at scale.
Cross-linguistic research shows users in every language trust confident AI outputs even when inaccurate. While confidence expression varies by language, users everywhere track confidence signals rather than accuracy, making overconfident errors systematically followed.
Sixteen paired writers showed strong preference for higher levels of prompt visibility in shared editors, valuing awareness of when, how, and where AI was used. Benefits included understanding collaborators' thinking and verifying AI-generated text, though some found full sharing intrusive and self-conscious.
Data2Story's Inspector binds every number, quote, and asset to its origin, making provenance rather than fluency the adoption gate. Across 18 samples, human raters favored this approach, showing that verifiable derivation—not surface polish—enables professional newsrooms to adopt agent output.
An empirical study found students working with chatbots achieved better practical performance and more knowledge-based dialogue than peer groups, but contributed significantly less dialogue overall and expressed far fewer subjective perspectives.
Machine learning identified interpretable traits from coding-agent conversations that explained outcomes beyond prior achievement. However, these traits lacked the stability and transferability required to qualify as learnable human-AI collaboration skills.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- The LLM Fallacy: Misattribution in AI-Assisted Cognitive Workflows
- Language Models Learn to Mislead Humans via RLHF
- How AI Impacts Skill Formation
- Does AI Assistance Leave a Temporal Fingerprint? Detecting Overreliance in AI-Assisted Writing and Programming
- Has the Creativity of Large-Language Models peaked? —an analysis of inter- and intra-LLM variability —
- Linguistic markers of inherently false AI communication and intentionally false human communication: Evidence from hotel reviews
- Evidence-centered Assessment for Writing with Generative AI
- Exploring Student-AI Interactions in Vibe Coding