Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Exploring Emerging Norms of AI Disclosure in Programming Education

Runlong Ye, Oliver Huang, Jessica He, Michael Liut

TL;DR

This study investigates how CS students perceive and disclose AI involvement in programming work, addressing a gap where attribution norms from prose may not apply to code. Using a factorial vignette design with 94 participants across 102 scenarios, it isolates the effects of AI autonomy, refinement, and task context on perceived ownership and disclosure demands. Key findings show that AI autonomy and post-AI refinement primarily shape attribution judgments, while policy expectations align with authorship perceptions but individuals’ personal disclosure decisions do not, indicating a gap between ethics and behavior. The authors advocate a shift to process-oriented attribution, treating disclosure as a record of the solution's evolution to foster critical engagement with AI rather than a liability flag, with implications for pedagogy and policy in programming education.

Abstract

Generative AI blurs the lines of authorship in computing education, creating uncertainty around how students should attribute AI assistance. To examine these emerging norms, we conducted a factorial vignette study with 94 computer science students across 102 unique scenarios, systematically manipulating assessment type, AI autonomy, student activity, prior knowledge, and human refinement effort. This paper details how these factors influence students' perceptions of ownership and disclosure preferences. Our findings indicate that attribution judgments are primarily driven by different levels of AI assistance and human refinement. We also found that students' perception of authorship significantly predicts their policy expectations. We conclude by proposing a shift from statement-style policies to process-oriented attribution, transforming disclosure into a pedagogical mechanism for fostering critical engagement with AI-generated content.

Exploring Emerging Norms of AI Disclosure in Programming Education

TL;DR

This study investigates how CS students perceive and disclose AI involvement in programming work, addressing a gap where attribution norms from prose may not apply to code. Using a factorial vignette design with 94 participants across 102 scenarios, it isolates the effects of AI autonomy, refinement, and task context on perceived ownership and disclosure demands. Key findings show that AI autonomy and post-AI refinement primarily shape attribution judgments, while policy expectations align with authorship perceptions but individuals’ personal disclosure decisions do not, indicating a gap between ethics and behavior. The authors advocate a shift to process-oriented attribution, treating disclosure as a record of the solution's evolution to foster critical engagement with AI rather than a liability flag, with implications for pedagogy and policy in programming education.

Abstract

Generative AI blurs the lines of authorship in computing education, creating uncertainty around how students should attribute AI assistance. To examine these emerging norms, we conducted a factorial vignette study with 94 computer science students across 102 unique scenarios, systematically manipulating assessment type, AI autonomy, student activity, prior knowledge, and human refinement effort. This paper details how these factors influence students' perceptions of ownership and disclosure preferences. Our findings indicate that attribution judgments are primarily driven by different levels of AI assistance and human refinement. We also found that students' perception of authorship significantly predicts their policy expectations. We conclude by proposing a shift from statement-style policies to process-oriented attribution, transforming disclosure into a pedagogical mechanism for fostering critical engagement with AI-generated content.
Paper Structure (26 sections, 4 figures, 4 tables)

This paper contains 26 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Attribution as a Pedagogical Guardrail
  4. Attribution Practices and The Gap in Programming Education
  5. Methods
  6. Procedure and Materials.
  7. Dependent Measures.
  8. Data Analysis.
  9. Results
  10. RQ1a: Perceptions of Authorship rely on AI Autonomy and Human Refinement
  11. RQ1b: Higher AI Autonomy and Implementation Tasks Trigger Stricter Disclosure Expectation
  12. RQ2: Authorship Concerns Drive Policy Requirements, But Do Not Predict Personal Intent to Disclose
  13. Policy Requirements are Driven by Authorship, Not Liability
  14. Personal Intention Diverges from Policy
  15. Discussion
  16. ...and 11 more sections

Figures (4)

  • Figure 1: Interaction between AI assistance level (B1: None, B2: Q&A, B3: Local generation, B4: Systemic/agentic) and human post-AI effort (E1: Used As-Is, E2: Significantly Modified) on six student-perception outcomes; points show mean Likert ratings (1–7) and error bars show standard error (SEM). No AI (B1) is shown as a baseline.
  • Figure 2: Distribution of disclosure preferences by AI Assistance Level (Left) and Activity Type (Right). AI assistance (Factor B) drives a shift towards "Co-authorship," while Production tasks (Factor C) trigger stricter citation requirements than Planning. We omit No AI (B1) because disclosure is trivially 'None' in that condition.
  • Figure 3: Main effect of AI assistance level (factor B) on mean Likert ratings for authorship, responsibility, grading fairness, and learning, collapsed across other factors; error bars show standard error (SEM).
  • Figure 4: Main effect of human post-AI effort (factor E) on mean Likert ratings for authorship, responsibility, grading fairness, and learning, collapsed across other factors; error bars show standard error (SEM).