Students who value AI ethics may regulate their learning more effectively

A study of 333 university students found that critical thinking, autonomy, and views of AI’s ethical and social value were associated with how learners managed their work with the technology.

Research involving 333 university students linked critical thinking, autonomy, and views of AI ethics with stronger self-regulated learning.

University students who view artificial intelligence through the lens of ethics and social good may be more effective at planning, monitoring, and evaluating their learning with the technology, according to new research.

The study found that students’ perceived autonomy and critical thinking directly predicted self-regulated learning with AI. Their beliefs about responsible AI use and its potential social benefits also helped explain those relationships.

A generally positive attitude toward AI did not significantly predict self-regulated learning in the researchers’ statistical model. The finding suggests that enthusiasm for AI tools is not, by itself, enough to help students use them effectively for learning.

Published in TechTrends, the study involved 333 art and design students who completed a one-week ChatGPT workshop before answering questions about autonomy, critical thinking, AI ethics, social good, attitudes toward AI, and their learning behaviors.

The research was conducted by Qi Xia of Zhejiang University, Norah Salem Aldharman of the University of Bisha, and Thomas K. F. Chiu of The Chinese University of Hong Kong.

Critical thinking and autonomy predicted self-regulated learning

Self-regulated learning covers the actions students take to direct their own progress, including setting goals, selecting strategies, monitoring their understanding, and reflecting on outcomes.

Within the study's statistical model, critical thinking had a significant direct association with self-regulated learning, with a standardized coefficient of 0.324. Perceived autonomy also had a significant direct association, with a coefficient of 0.203.

The critical thinking questions examined students' confidence in addressing complex problems, learning challenging concepts, and thinking precisely. The autonomy measures covered whether students felt able to choose learning activities, tools, skills, and topics when working with AI.

Chiu said in a LinkedIn post: "Students need autonomy support to take ownership of their learning and strong critical thinking skills to evaluate AI-generated content."

The findings indicate that access to an AI tool does not, by itself, show whether students can use it to manage their learning. Students also need to assess the accuracy and relevance of AI responses, decide whether recommendations support their learning goals, and recognize when information requires further checking.

The study measured students' perceptions and reported learning behaviors rather than objective academic results. It did not test whether students who reported stronger critical thinking or autonomy achieved higher grades, retained more knowledge, or produced better work.

Ethics and social good mattered more than enthusiasm for AI

Students' views about the values surrounding AI were also associated with self-regulated learning.

Perceptions of AI ethics significantly predicted self-regulated learning within the model, with a coefficient of 0.143. Perceptions of AI's contribution to social good recorded a coefficient of 0.257.

The ethics questions covered responsible use, privacy, security, and concern about the misuse of AI. Social-good measures included whether students believed AI could support disadvantaged people, promote human well-being, and be used for the common good.

Both factors partly mediated the relationships between critical thinking, autonomy, and self-regulated learning. This means students' views about how AI should be used formed part of the connection between their learning skills and how they reported managing work with the technology.

Positive attitudes toward AI produced a coefficient of 0.112, but the result was not statistically significant. Questions in that section asked whether students felt hopeful about an AI-enabled future and expected the technology to make society or their own lives better.

The result challenges the assumption that students who like AI will automatically learn more effectively with it. A student may be willing to use the technology without consistently setting goals, checking its output, monitoring understanding, or reflecting on the quality of the work produced.

Chiu said: "Crucially, we must cultivate value-based perceptions, helping students see AI as a tool for ethical progress and social benefit."

The researchers recommend that educators address the purpose and social consequences of AI alongside practical and technical skills. This could include asking students to consider privacy, fairness, misuse, accountability, and the impact of AI systems on other people.

The findings do not establish that teaching AI ethics will directly improve academic performance. They show an association between students' reported value perceptions and their reported self-regulated learning within this group and learning activity.

Study focused on one student group and self-reported data

The 333 participants had an average age of 19.62 and were studying subjects including visual communication design, environmental design, and digital media technology.

Of the students surveyed, 40.2 percent were in their first year, 21.3 percent were sophomores, 32.7 percent were juniors, and 5.7 percent were seniors. The sample was 55.3 percent female and 44.7 percent male.

Most participants had previously encountered AI tools but had relatively limited experience using them specifically for learning.

Before completing the survey, all students took part in a one-week workshop on using ChatGPT. They were asked to create a summer travel plan covering factors including budget, group size, and trip duration.

A teacher used autonomy-supportive teaching strategies and allowed participants to decide how they would use ChatGPT during the activity. Students then completed an electronic questionnaire lasting approximately 20 minutes.

The cross-sectional research design identified relationships between the measured factors at one point in time. It cannot establish that stronger critical thinking, autonomy, or ethical views caused students to become better self-regulated learners.

The study also relied on students reporting their own skills, perceptions, and behaviors. It did not observe how they used ChatGPT over a longer period, independently assess the quality of their travel plans, or compare their performance with students who did not use AI.

The researchers acknowledge that the participants' art and design backgrounds may have influenced their perceptions of the technology. They also note that the research examined only three aspects of AI perception and focused largely on potential benefits rather than harmful or counterproductive uses.

The authors recommend that teachers give students meaningful control over parts of their learning while retaining human guidance. Suggested approaches include allowing learners to design study plans, select appropriate tools, and use ChatGPT for additional practice.

They also call for educational AI products to be designed around the requirements of particular disciplines rather than assuming that the same system and learning approach will suit every student.

Future studies are expected to include students from a wider range of subjects, combine survey results with qualitative evidence, and examine negative as well as positive effects of AI-supported learning.