📄 social_science

Do Social Norms Matter in Promoting Students’ Engagement in GenAI-Supported Learning Environments? An Integrated Theoretical Model

This study utilizes data from 2,568 Chinese undergraduates to demonstrate that positive attitudes toward generative AI significantly enhance student engagement through the sequential mediation of usage intention and self-efficacy, a process that is further strengthened by supportive social norms.

Original authors: Chengjin Gao, Dong Yang, Gurpinder Singh Lalli

Published 2026-08-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Chengjin Gao, Dong Yang, Gurpinder Singh Lalli

Original paper licensed under CC BY 4.0 (https://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 walking into a giant, futuristic library where the books can talk back, write essays for you, and even argue about history. This is the world of Generative AI in education. But just because the library is open doesn't mean everyone wants to run inside and start chatting with the books. Some students are excited, some are scared, and some just don't know how to talk to the machines. This is where the science of psychology and education collide.

To understand why a student decides to dive into this AI library, researchers look at a few key ideas. First, there's Attitude: this is just your gut feeling. Do you think the AI is cool and helpful, or is it a scary robot that will steal your homework? Second, there's Self-Efficacy. Think of this as your "I can do it" muscle. It's not about whether you are good at using AI right now, but whether you believe you can learn to be good at it if you try. Third, there's Intention, which is simply the plan to actually use the tool. Finally, there are Social Norms. This is the "vibe check" of your school. Do your teachers, friends, and the school itself seem to say, "Hey, using AI is a great idea!" or do they whisper, "Don't touch that, it's against the rules"?

Why does this matter? Because schools are pouring money into these AI tools, but if students are too scared, too unsure of themselves, or feel like their friends are judging them, they won't use them. And if they don't use them, they miss out on the chance to learn in a whole new way. The big question is: How do we get students from "I guess AI is okay" to "I am deeply engaged and learning with AI every day"?

The Study: The AI Engagement Detective

A team of researchers from China decided to play detective to solve this mystery. They gathered a massive group of 2,568 university students and asked them a bunch of questions about how they felt about AI, how confident they felt using it, what they planned to do, and what their friends and teachers thought. They then used a super-smart mathematical tool (called PLS-SEM, which is like a GPS for finding hidden connections in data) to map out exactly how these feelings connect to one another.

Here is what they found, broken down into a story:

1. The Spark: Attitude is the Engine
The study found that if a student has a positive attitude toward AI (they think it's useful and fun), they are much more likely to be engaged. Engagement here means they aren't just clicking buttons; they are thinking deeply, asking questions, and really getting involved.

The Analogy: Think of AI attitude like the fuel in a car. If the fuel is high-quality (positive feelings), the car is ready to go. If the fuel is bad (negative feelings), the car won't start, no matter how good the engine is.

2. The Middlemen: Confidence and Plans
But having a positive attitude isn't the only thing that matters. The researchers discovered that attitude works through two middlemen: Self-Efficacy and Intention.

The Chain Reaction: A student with a positive attitude first starts to believe, "I can actually learn to use this!" (Self-Efficacy). Once they believe they can do it, they make a plan: "I'm going to use this tool for my next project" (Intention). Finally, that plan turns into action (Engagement).
The Analogy: Imagine you want to ride a skateboard.
- Attitude: You think skateboarding looks awesome.
- Self-Efficacy: You tell yourself, "I can learn to balance on this."
- Intention: You decide, "I'm going to go to the park and try it."
- Engagement: You are actually out there skating.
  The study showed that skipping the "I can do it" belief or the "I will do it" plan makes it much harder to get to the actual skating.

3. The Amplifier: Social Norms
This is the most interesting part. The researchers found that Social Norms act like a volume knob or a turbocharger.

How it works: If a student has a positive attitude but their teacher says "AI is bad" or their friends say "Don't use it," the connection between their attitude and their action gets weaker. But, if the student has a positive attitude and their teachers and friends are cheering them on ("Yes! Use AI!"), that positive attitude explodes into much stronger confidence and much deeper engagement.
The Analogy: Think of your attitude as a small campfire.
- Low Social Norms: If everyone around you is blowing on the fire to put it out (negative vibes), the fire stays small, even if you have good wood (positive attitude).
- High Social Norms: If everyone is fanning the flames and throwing on more wood (supportive vibes), that same small campfire turns into a roaring bonfire. The supportive environment makes the positive feelings work much harder.

4. What They Ruled Out
The study also made it clear that having a positive attitude alone isn't a magic wand. You can think AI is great, but if you don't believe you can use it (low self-efficacy) or if your environment is hostile (negative social norms), you won't end up deeply engaged. The paper suggests that you can't just tell students "AI is cool" and expect them to jump in; you have to build their confidence and make sure their school environment supports them.

5. How Sure Are They?
The researchers are quite confident in these results because they used a huge number of students (over 2,500) and rigorous math to prove the links. They found that the "chain" of Attitude → Confidence → Plan → Action is real and measurable. They also found that the "Social Norms" turbocharger effect is statistically significant, meaning it's not just a fluke; it really does change how students behave.

The Bottom Line
This paper tells us that to get students excited and deeply involved with AI in school, we can't just hand them the technology. We have to:

Make sure they like the idea of using it.
Help them believe they are capable of using it.
Make sure their teachers and friends are cheering them on.

If you get all three of those right, the students won't just use the AI; they will truly engage with it, learning and growing in ways we are just starting to understand.

Technical Summary: Social Norms and Student Engagement in GenAI-Supported Learning

Problem Statement

While Generative Artificial Intelligence (GenAI) is rapidly transforming higher education through tools like intelligent recommendation systems and automated assessment, the mechanisms linking AI acceptance to actual student engagement remain insufficiently understood. Existing research often focuses on technical functionalities or initial technology adoption, neglecting the psychological processes that translate acceptance into sustained, deep learning behaviors. Furthermore, the role of social norms—perceived expectations from teachers, peers, and institutions—as a boundary condition in this process has received limited attention. There is a gap in understanding how individual attitudes, motivational factors (intention), and capability beliefs (self-efficacy) interact within a social context to drive engagement in dynamic GenAI environments.

Methodology

The study employed a quantitative, cross-sectional design to test an integrated theoretical model combining the Theory of Planned Behaviour (TPB) and Social Cognitive Theory (SCT).

Participants: Data were collected from 2,568 students (including undergraduates and postgraduates) across ten universities in China, representing diverse educational levels (universities and junior colleges) and disciplines (Humanities, Sciences, Engineering, etc.).
Instrument: A structured online questionnaire administered via the Wenjuanxing platform included five latent constructs measured on five-point Likert scales:
1. AI Attitudes: Cognitive and affective evaluations of AI (6 items).
2. AI Self-Efficacy: Confidence in using AI tools effectively (4 items).
3. Intention to Use: Willingness to continue using AI for learning (4 items).
4. Student Engagement: Multidimensional engagement (behavioral, emotional, cognitive, agentic) in GenAI contexts (19 items).
5. Supportive Social Norms: Perceived encouragement and resource support from the social environment (4 items).
Analysis: The researchers utilized Partial Least Squares Structural Equation Modelling (PLS-SEM) via SmartPLS 4. The analysis involved:
- Assessing measurement model reliability and validity (Cronbach's $\alpha$ , Composite Reliability, AVE, HTMT).
- Testing direct, mediating, and moderating effects using bootstrapping (5,000 resamples).
- Evaluating effect sizes ( $f^2$ ) and predictive relevance ( $R^2$ , SRMR).

Key Contributions

The study makes three primary theoretical contributions:

Theoretical Integration: It extends the integration of TPB and SCT specifically to GenAI learning contexts, proposing a sequential pathway where attitudes shape self-efficacy, which then influences intention, ultimately driving engagement. Notably, it positions self-efficacy as temporally prior to intention in this specific context, arguing that learners must perceive capability before forming sustained intentions.
Outcome Expansion: It shifts the focus from mere technology adoption to student engagement as the primary outcome, capturing the depth of interaction required in GenAI environments.
Contextual Boundary Conditions: It conceptualizes supportive social norms not merely as direct predictors but as critical moderators that strengthen the translation of attitudes into beliefs, intentions, and behaviors.

Results

The structural model demonstrated satisfactory fit ($SRMR = 0.033$, $NFI = 0.885$, $R^2 = 0.688$ for engagement).

Direct Effects: AI attitudes significantly and positively predict student engagement ( $\beta = 0.330$ ).
Mediating Effects:
- Self-Efficacy partially mediates the relationship between attitudes and engagement ( $\beta = 0.129$ ).
- Intention to Use partially mediates the relationship between attitudes and engagement ( $\beta = 0.151$ ).
- Sequential Mediation: A significant chain mediation exists: AI Attitudes $\rightarrow$ Self-Efficacy $\rightarrow$ Intention to Use $\rightarrow$ Student Engagement ( $\beta = 0.068$ ).
Moderating Effects: Supportive Social Norms significantly moderate the relationships between AI attitudes and:
- Self-Efficacy ( $\beta = 0.081$ )
- Intention to Use ( $\beta = 0.111$ )
- Student Engagement ( $\beta = 0.162$ )
- Interpretation: The positive effects of AI attitudes on these outcomes are significantly stronger when supportive social norms are high compared to when they are low.
Effect Sizes: AI attitudes showed a large effect on self-efficacy ( $f^2 = 0.657$ ) and medium effects on intention and engagement. The interaction effects (moderation) were small in magnitude but statistically significant, indicating they meaningfully influence the strength of the main relationships.

Significance and Claims

The authors claim that this study provides a comprehensive explanation of how learners transition from perception to action in AI-supported learning environments.

Mechanism Clarification: The findings confirm that positive AI attitudes are foundational, but their translation into engagement is mediated by the learner's confidence (self-efficacy) and motivation (intention). Crucially, the study posits that self-efficacy acts as a prerequisite for intention in this context, where confidence in capability precedes the formation of sustained usage intentions.
Contextual Importance: The study underscores that individual psychological factors do not operate in a vacuum. Supportive social norms act as critical boundary conditions that amplify the impact of positive attitudes. In collectivist contexts, the alignment of behavior with perceived social expectations significantly enhances the likelihood of engagement.
Practical Implication: The authors suggest that effective AI integration requires a coordinated approach: fostering positive attitudes, building learner confidence through scaffolded experiences, and actively cultivating supportive institutional and peer environments.

The paper concludes that while the cross-sectional design limits causal inference, the integrated model offers a robust framework for understanding the dynamic interplay of cognitive, motivational, and contextual factors in the era of Generative AI.