💻 computer science

Enabling and Inhibitory Pathways of University Students' Willingness to Disclose AI Use: A Cognition-Affect-Conation Perspective

This study employs a Cognition-Affect-Conation framework and mixed-methods design to reveal that university students' willingness to disclose AI use is driven by enabling factors like psychological safety and institutional support, while being hindered by inhibitory factors such as evaluation apprehension and policy ambiguity.

Original authors: Yiran Du, Huimin He

Published 2026-04-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yiran Du, Huimin He

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a student in a university class. You've just finished a big essay, and you used an AI tool (like a super-smart robot assistant) to help you brainstorm ideas, check your grammar, or explain a tricky concept. Now, you have a choice: do you tell your teacher, "Hey, I used a robot to help me," or do you keep it a secret?

This research paper tries to answer why students choose to speak up or stay silent. The authors, Yiran Du and Huimin He, used a psychological framework called Cognition–Affect–Conation (which is just a fancy way of saying: Thoughts → Feelings → Actions).

Here is the story of their findings, explained with simple analogies.

The Big Picture: The "Safe Zone" vs. The "Minefield"

The researchers found that students' decisions aren't random. They are walking a tightrope between two emotional zones:

The Safe Zone (Enabling Pathway): When students feel safe, they are willing to admit they used AI.
The Minefield (Inhibitory Pathway): When students feel scared, they hide their AI use to avoid getting blown up (punished or judged).

1. The "Safe Zone": How to Get Students to Speak Up

Think of the university as a garden. For a plant (the student) to grow and show its flowers (disclose AI use), the soil needs to be right.

Fairness is the Sun: If students believe the rules are fair and the teacher isn't playing favorites, they feel warm and secure. They think, "If I used AI, the teacher will judge me based on the rules, not just get angry."
Teacher Support is the Water: When a teacher says, "It's okay to use tools to learn," it's like watering the plant. It tells the student, "You are not in trouble; you are learning."
School Support is the Fence: When the whole university has clear, helpful policies (not just vague warnings), it acts like a protective fence. It keeps the wind (fear) away.

The Result: When these three things are present, students feel Psychological Safety. This is like wearing a superhero cape. They feel brave enough to say, "I used AI, and here is how I used it," without fearing they will be punished.

2. The "Minefield": Why Students Stay Silent

Now, imagine the classroom is a minefield where stepping on the wrong spot triggers an explosion (a bad grade or a lecture).

Stigma is the "Bad Reputation" Sign: Some students feel that using AI is like cheating, even if they didn't. They worry people will think, "Oh, they couldn't do it themselves." This makes them feel ashamed.
Uncertainty is the Fog: The rules about AI are changing fast. It's like walking through a foggy forest where you don't know where the path ends. Students think, "If I say I used AI, will the teacher get mad? Is it allowed? I don't know!"
Privacy is the Glass Wall: Students worry that if they admit using AI, the teacher will look too closely at their work, like a detective. They fear, "If I tell them, they'll scrutinize every word I wrote."

The Result: These fears create Evaluation Apprehension. This is like wearing heavy, lead boots. The student is so worried about being judged that they freeze. They decide, "It's safer to just say nothing and hope for the best."

3. The "Strategic Player" (What the Interviews Revealed)

The researchers didn't just ask questions; they talked to 22 students. They found that students aren't just "honest" or "dishonest." They are strategic players, like chess players.

The "Grammar Check" Exception: A student might say, "I used AI to fix my spelling, so I'll tell the teacher."
The "Big Help" Exception: But if the AI wrote half the essay, they might stay silent because the risk is too high.
The "Teacher Vibe" Check: Students look at the teacher. If the teacher seems cool with AI, they speak up. If the teacher looks strict, they stay quiet.

It's like a traffic light.

Green Light (Clear rules + Supportive teacher): "Go ahead, tell them!"
Red Light (Vague rules + Strict teacher): "Stop! Don't say anything!"
Yellow Light (Mixed signals): "Wait... maybe I'll just say a little bit, but not everything."

The Takeaway: What Should Schools Do?

The paper concludes that if universities want students to be honest about AI, they can't just make strict rules and hope for the best. They need to change the emotional environment.

Stop the Fog: Make the rules crystal clear. Tell students exactly when AI is okay and how to report it.
Take off the Heavy Boots: Teachers need to show they aren't going to punish students for using tools responsibly. They need to be the "safe zone," not the "minefield."
Wear the Cape: Create a culture where using AI is seen as a smart learning tool, not a shameful secret.

In short: Students will only take off their masks and admit they used AI if they feel the ground beneath them is solid, not shaky. If they feel safe, they will be transparent. If they feel scared, they will hide.

1. Problem Statement

The rapid integration of Generative AI (GenAI) and Large Language Models (LLMs) into higher education has created a critical tension between enhanced learning efficiency and the imperative for academic integrity. While students increasingly use AI tools for drafting, problem-solving, and content generation, there is a significant gap in understanding why students choose to disclose or conceal this usage.

Existing literature focuses heavily on AI adoption rates or detection mechanisms, neglecting the psychological processes driving voluntary disclosure. Students often hesitate to report AI use due to fears of negative evaluation, policy ambiguity, and social stigma. This study addresses this gap by investigating the specific cognitive and affective mechanisms that enable or inhibit students' willingness to disclose AI use.

2. Methodology

The study employs a sequential explanatory mixed-methods design, combining quantitative structural equation modeling (SEM) with qualitative thematic analysis.

Quantitative Phase:
- Sample: 546 university students from diverse disciplines (STEM and non-STEM) across multiple universities in China.
- Instrument: A structured questionnaire measuring nine constructs based on the Cognition–Affect–Conation (CAC) framework: Perceived Fairness, Teacher Support, Organizational Support, Stigma, Uncertainty, Privacy Concern, Psychological Safety, Evaluation Apprehension, and Willingness to Disclose.
- Analysis: Data were analyzed using R and Structural Equation Modeling (SEM). Confirmatory Factor Analysis (CFA) validated the measurement model (reliability, convergent, and discriminant validity), followed by path analysis to test hypothesized relationships.
Qualitative Phase:
- Sample: 22 students purposively selected from the survey pool, representing high and low willingness to disclose.
- Instrument: Semi-structured interviews exploring experiences with AI policies, instructor attitudes, and decision-making processes regarding disclosure.
- Analysis: Deductive-inductive thematic analysis was used to interpret and contextualize the quantitative findings.

3. Theoretical Framework

The study is grounded in the Cognition–Affect–Conation (CAC) framework, which posits that behavior is a sequential process:

Cognition: Students evaluate their environment (e.g., fairness, support, stigma).
Affect: These evaluations trigger emotional states (e.g., psychological safety, evaluation apprehension).
Conation: Emotional states drive behavioral intentions (willingness to disclose).

The study proposes a dual-pathway model:

Enabling Pathway: Positive cognitive perceptions $\rightarrow$ Psychological Safety $\rightarrow$ Disclosure.
Inhibitory Pathway: Negative cognitive perceptions $\rightarrow$ Evaluation Apprehension $\rightarrow$ Reduced Disclosure.

4. Key Results

Quantitative Findings (SEM Results)

The structural model provided strong empirical support for the proposed hypotheses:

Enabling Pathway:
- Psychological Safety is the strongest positive predictor of Willingness to Disclose ( $\beta = 0.48, p < .001$ ).
- Perceived Fairness ( $\beta = 0.29$ ), Teacher Support ( $\beta = 0.34$ ), and Organizational Support ( $\beta = 0.21$ ) all significantly and positively predict Psychological Safety.
Inhibitory Pathway:
- Evaluation Apprehension significantly and negatively predicts Willingness to Disclose ( $\beta = -0.31, p < .001$ ).
- Evaluation Apprehension also negatively impacts Psychological Safety ( $\beta = -0.26$ ).
- Perceived Stigma ( $\beta = 0.41$ ), Perceived Uncertainty ( $\beta = 0.28$ ), and Privacy Concern ( $\beta = 0.24$ ) are significant positive predictors of Evaluation Apprehension.

Qualitative Findings

Thematic analysis revealed three core themes that explain the quantitative data:

Institutional Clarity and Support: Clear, consistent policies and supportive instructor attitudes reduce ambiguity, making students feel safer to disclose. Inconsistency between instructors, however, creates hesitation.
Stigma and Uncertainty: Students fear being labeled as academically dishonest or lacking ability. Ambiguous policies regarding "acceptable" AI use heighten anxiety about potential punishment.
Strategic Disclosure: Disclosure is often conditional. Students engage in "risk-benefit analysis," disclosing only for minor tasks (e.g., grammar checking) or when the instructor is known to be supportive, while concealing usage in high-stakes or ambiguous contexts.

5. Key Contributions

Theoretical Extension: The study successfully extends the CAC framework to the specific context of AI transparency in education, identifying Psychological Safety and Evaluation Apprehension as critical mediating affective states.
Dual-Pathway Mechanism: It moves beyond a single-factor explanation by delineating distinct enabling (support/fairness) and inhibitory (stigma/uncertainty) pathways that operate simultaneously.
Mixed-Methods Insight: The qualitative data provides nuance to the quantitative model, revealing that disclosure is not a binary decision but a strategic, context-dependent behavior influenced by the consistency of institutional messaging.

6. Significance and Implications

For Policy Makers: Institutions must move beyond vague prohibitions. Clear, consistent, and fair policies are essential to reduce uncertainty and stigma. Policies should explicitly define acceptable AI use to foster psychological safety.
For Educators: Instructor behavior is a primary driver of disclosure. Teachers who demonstrate openness, provide explicit guidance on AI integration, and avoid punitive assumptions can significantly increase student transparency.
For Practice: Promoting "AI Literacy" and ethical frameworks that normalize AI as a tool rather than a cheat sheet can reduce the perceived stigma associated with its use.

7. Limitations

Context: The study was conducted in China; cultural norms regarding authority and academic integrity may limit generalizability to Western contexts.
Design: The cross-sectional nature of the quantitative data prevents definitive causal claims, though the theoretical model supports the directionality.
Self-Report: Reliance on self-reported intentions may be subject to social desirability bias.

Conclusion

The study concludes that fostering responsible AI transparency requires a dual approach: actively building psychological safety through supportive and fair institutional environments while simultaneously mitigating evaluation apprehension by clarifying policies and reducing the stigma surrounding AI use.