AI in academia: LLMs shown to improve teamwork and analytical reasoning

The study finds mounting evidence that large language models can promote collaboration and teamwork in both physical and digital classrooms. Instructors are using AI chat systems to simulate group brainstorming sessions, generate discussion prompts, and scaffold cooperative project design.

A major literature review published in Systems has found that artificial intelligence systems, particularly large language models (LLMs), could reshape higher education by strengthening critical thinking, teamwork, and problem-solving skills. The study systematically examines the growing body of research on how generative AI tools influence student learning outcomes, teaching strategies, and academic integrity in modern universities.

The paper, titled “Can Large Language Models Foster Critical Thinking, Teamwork, and Problem-Solving Skills in Higher Education?: A Literature Review,” is based on 22 peer-reviewed studies published between 2023 and 2024 and maps the early evidence from classrooms around the world where students and instructors have adopted LLMs such as ChatGPT, Bard, and Claude to enhance learning and assessment.

How AI is changing the foundations of higher education

The research team followed the PRISMA 2020 methodology, applying a systematic and transparent process for selecting and analyzing academic literature indexed in Web of Science. From an initial pool of 203 records, the authors identified 22 studies that met inclusion criteria, each directly investigating the pedagogical or cognitive impact of LLMs in higher education.

The review addresses four key questions: whether LLMs can (1) foster critical thinking, collaboration, and problem-solving; (2) bridge theoretical knowledge with real-world applications; (3) provide scalable, personalized feedback in large classes; and (4) modernize assessments toward skills-based evaluation.

The findings suggest that while the technology remains in early adoption, LLMs are already reshaping the structure and dynamics of university learning. Students who engage critically with generative AI, especially through guided prompting, reflection, and iterative feedback, develop higher-order thinking skills that mirror authentic cognitive processes such as analysis, synthesis, and evaluation.

In several of the reviewed studies, learners who interacted regularly with LLMs reported greater confidence in argumentation, reasoning, and peer communication. Faculty members who incorporated structured AI use into coursework observed increased engagement, faster iteration on written assignments, and deeper conceptual understanding when students were asked to verify or critique the model’s responses.

The review concludes that LLMs serve best as cognitive partners rather than as information providers. By challenging students to interrogate outputs, identify biases, and refine prompts, these models inadvertently reinforce metacognitive awareness, the mental process of thinking about one’s own thinking, that underpins critical reasoning.

LLMs as collaborative learning partners

The study finds mounting evidence that large language models can promote collaboration and teamwork in both physical and digital classrooms. Instructors are using AI chat systems to simulate group brainstorming sessions, generate discussion prompts, and scaffold cooperative project design.

When used in team settings, LLMs help students combine diverse perspectives into coherent solutions by acting as mediators that organize information, summarize debate points, or generate alternatives during problem-solving exercises. In one cited study, engineering students used an AI assistant to draft initial prototypes of project proposals; subsequent human discussion improved clarity and innovation, while the AI’s suggestions accelerated early-stage ideation.

This effect depends on active guidance. When educators embed AI use into a structured pedagogical framework, such as design thinking or problem-based learning, students are more likely to use the tool as a collaborative co-thinker rather than a shortcut to answers.

Moreover, AI can bridge language and disciplinary barriers in multicultural classrooms. Studies reviewed in the paper show that multilingual students benefit from the model’s capacity to rephrase complex terminology or generate examples relevant to their cultural contexts. This improves inclusivity and teamwork by ensuring equal participation in collaborative assignments.

However, the review also acknowledges significant challenges. Overreliance on LLMs risks flattening creativity if students bypass cognitive struggle or fail to critically appraise machine-generated responses. Additionally, questions of authorship and originality continue to complicate group assessment and academic integrity standards.

The authors propose that universities adopt clear guidelines to encourage responsible AI collaboration, highlighting documentation of prompts, peer review of AI-assisted outputs, and reflective commentary as part of coursework evaluation.

Personalized feedback and the future of assessment

Perhaps the most transformative potential identified in the review lies in feedback generation. Across the analyzed studies, LLMs demonstrated capacity to provide instant, individualized feedback on writing, reasoning, and problem-solving tasks, capabilities that could address one of academia’s oldest challenges: the time constraints faced by instructors managing large classes.

AI tools can evaluate student work in real time, highlight structural weaknesses, suggest improvements, and simulate formative dialogue. When calibrated to specific rubrics or course objectives, these systems enable continuous learning loops that reinforce retention and comprehension.

In addition to improving efficiency, AI-generated feedback helps democratize access to academic support. Students who might hesitate to seek help in person can test ideas, practice skills, or iterate on drafts privately with a language model, receiving tailored guidance that boosts confidence and autonomy.

The review also finds early evidence that LLMs can enhance assessment design itself. Educators are using generative AI to build competency-based evaluation frameworks, creating adaptive tests, scenario-based assignments, and reflective essays that assess reasoning and creativity rather than rote knowledge.

By integrating AI into formative assessment, instructors can monitor learning progress dynamically and provide just-in-time interventions. The technology supports shift from static testing to continuous, data-informed evaluation, aligning with emerging educational models that prioritize lifelong learning and employability.

Nonetheless, the authors warn that these benefits depend heavily on transparency and oversight. Black-box algorithms may embed unrecognized biases in scoring or feedback. Therefore, the review urges institutions to adopt explainable AI (XAI) systems that make decision logic visible and allow human moderation of automated feedback.

Ethical, pedagogical, and institutional implications

The review situates the pedagogical promise of LLMs within a broader institutional debate about the ethics and governance of AI in education. It recognizes that generative models introduce new dilemmas related to plagiarism, intellectual authorship, data privacy, and the balance between automation and human mentorship.

The authors note that while AI enhances efficiency, universities must ensure that its integration preserves academic integrity and human-centered learning. Faculty development programs are essential to help instructors design activities that use AI ethically while preserving cognitive rigor and originality.

Policy frameworks should evolve in tandem with technology. Institutions need clear codes of conduct defining acceptable AI use in coursework, research, and examinations. Additionally, data governance measures must protect student interactions with AI tools from commercial exploitation or unauthorized monitoring.

Going ahead, the authors argue for a hybrid educational model that combines AI’s analytical speed with the emotional intelligence and ethical discernment of educators. Such integration could redefine the teacher’s role, not as a transmitter of information but as a facilitator of inquiry, creativity, and reflection.