Artificial Intelligence and Collaborative Learning: Impacts on Creativity, Critical Thinking, and Problem-Solving

Synthesis: Artificial Intelligence and Collaborative Learning: Impacts on Creativity, Critical Thinking, and Problem Solving

Key Findings

• Mixed-methods study of 102 undergraduate students in interdisciplinary project-based courses, of which 96 responses (94.1%) were valid for analysis; qualitative input came from 80 open-ended survey responses and 15 semi-structured interviews, triangulated with analysis of student project outputs.
• 83% of respondents reported that LLMs enhanced creativity (generating new ideas, offering different perspectives during brainstorming), and 78% said LLMs increased their group's overall productivity — although 20% felt Large Language Models (LLMs) use decreased individual accountability and contribution, creating workload imbalances.
• 81% noted LLMs acted as neutral mediators that helped resolve differences and speed consensus building, while 17% indicated the presence of LLMs diminished the need for critical engagement, with AI suggestions sometimes accepted without discussion.
• Correlation analysis showed the strongest relationship between consensus-building impact and productivity impact (r = 0.71), followed by collective intelligence and productivity (0.69); AI comfort correlated with perceived creativity impact (0.56) and problem-solving impact (0.59), and digital proficiency showed only moderate correlations (0.52, 0.47, 0.50).
• No statistically significant differences in perceived LLM impact were found across groups based on prior AI use, AI comfort, or digital proficiency — a t-test (t = 1.414, p = 0.2929) and ANOVA (f = 5.33, p = 0.1028) at α = 0.05 both failed to reject the null hypothesis.
• Respondents were predominantly business and management majors (62%), aged 18–26 with an even gender split; 90% were very comfortable with AI tools and 96% had prior LLM experience, with digital proficiency rated intermediate (54%) or advanced (40%).

Study Design & Method

The study used a descriptive, mixed-methods design with no variable manipulation. Quantitative data came from a structured Likert-scale survey measuring perceived impacts of LLMs on group collaboration, creativity, and problem-solving efficacy. Qualitative data came from open-ended survey items, audio-recorded semi-structured interviews transcribed for thematic analysis, and inspection of student project outputs for originality and collaborative innovation. Statistical analysis combined a Pearson correlation matrix with t-tests and ANOVA (implemented in Python) to compare perceptions across groups defined by AI comfort, digital proficiency, and prior LLM use. Because the hypothesis tests returned non-significant results, the authors additionally explored alternative factors (specific courses taken, exposure to AI tools, learning styles) and proposed k-means clustering as a way to segment student interaction patterns.

Key Results

• Perceived benefits: students credited LLMs with improving problem-solving (e.g., quick summaries of long texts freed groups to focus on higher-order tasks), structuring group discussion, synthesizing individual contributions into cohesive action plans, and boosting collective intelligence.
• Perceived risks: overreliance on AI to build consensus was flagged as a threat to interpersonal skill development; some group members became passive and deferred to AI-generated solutions; instructors were advised to monitor group dynamics and intervene when workload distribution becomes imbalanced.
• Digital competence does not shape perceptions: the absence of significant group differences suggests familiarity and digital skill are weak predictors of how students perceive LLM effects on creativity, problem-solving, consensus building, and productivity.
• Balanced integration: the authors warn that heavy dependence on generative AI can impair self-regulated learning, intrinsic motivation, and performance, citing the phenomenon of "metacognitive laziness" — outsourcing cognitive effort to AI reduces engagement with deeper learning strategies and can erode independent analysis, synthesis, and evaluation over time.

What this means for practice

• Instructors. Require groups to justify decisions independently of the AI before adopting its suggestions: 17% of respondents reported that LLM presence diminished the need for critical engagement, with AI suggestions sometimes accepted without discussion.
• Instructional designers. Build individual accountability into group tasks, because 20% of respondents said LLM use reduced individual accountability and contribution, with workload imbalance when teammates leaned on the AI for work human members should have done.
• Instructors. Build students' comfort with AI rather than assuming digital proficiency covers it: AI comfort correlated with perceived creativity impact (0.56) and problem-solving impact (0.59) more strongly than digital proficiency did (0.52, 0.47, 0.50).
• Instructional designers. Keep the human elements the study treats as essential — challenging assumptions, negotiating meaning, and collective knowledge construction — and treat speed of consensus as a warning sign rather than a success metric.

Limitations

• The design is descriptive and cross-sectional with no manipulation, drawn from one survey wave of 102 undergraduates in interdisciplinary project-based courses, and only 96 responses were usable after six were dropped for missing major discipline or prior AI familiarity.
• Its hypothesis tests failed to reject the null — t = 1.414 (p = 0.2929) and F = 5.33 (p = 0.1028) at alpha = 0.05 — so the claim that AI comfort, prior LLM use, and digital proficiency do not shape perceptions is an absence of evidence in this sample rather than a demonstrated null.
• The sample skews toward AI-familiar, business-oriented students: 62% were business and management majors, 90% reported being very comfortable with AI tools, and 96% had prior LLM experience, which can restrict the range on exactly the subgroup comparisons that produced non-significant results.
• The impact measures are self-reported perceptions of creativity, productivity, consensus building, and collective intelligence, supported by 80 open-ended responses and 15 interviews rather than any independent performance measure.

Citation

Agnaou, A., & El Asri, H. (2025). Artificial intelligence and collaborative learning: Impacts on creativity, critical thinking, and problem-solving.