Abstract:

Training chemical engineering students in modern literature synthesis and data analytics is undoubtedly helpful for them to navigate the current explosion of scientific research. The volume of published data has expanded so fast that traditional manual meta-analyses now risk significant oversights. To address this challenge, this communication presents a novel framework for dynamic or “living” meta-analyses augmented by artificial intelligence (AI), providing a structured way for students and researchers to systematically curate, extract, and synthesize vast amounts of information in real time.

The proposed framework integrates Large Language Models (LLMs), such as ChatGPT and Gemini, with bibliometric platforms and network visualization tools like VOS viewer. Aligned with the PRISMA guidelines, this methodology ensures a transparent and reproducible data curation process that can be updated as new research emerges.

The general workflow is divided into five progressive milestones:
(i) scope definition and prompt engineering for search string optimization;
(ii) bibliometric mapping to identify thematic clusters and influential research nodes;
(iii) systematic screening to ensure data eligibility;
(iv) AI-driven structured data extraction from complex texts and supporting information; and
(v) human-in-the-loop synthesis to verify outputs and provide theoretical context.

From a pedagogical perspective, this approach is grounded in the Information Problem Solving (IPS-I) model. It shifts the student’s role from performing repetitive, mechanical tasks (such as manual data extraction and bibliography sorting) toward high-level critical thinking and the rigorous evaluation of AI-generated content. Importantly, professors should insist on the “human-in-the-loop” requirement, as our expertise remains indispensable to verify AI outputs, mitigate potential hallucinations, and ensure that findings are properly contextualized within the existing body of knowledge.

This methodology is being implemented at the University of Girona within the chemical engineering curriculum. It helps students prepare for their Bachelor’s theses (“Treball de fi de grau” in Catalan) and further bibliographical tasks after graduating. Results indicate that this approach significantly reduces the cognitive load associated with information overload while simultaneously equipping students with the digital literacy and AI scientific literacy necessary for modern industrial R&D. It is shown that transforming static, one-off literature reviews into dynamic platforms enables a more robust connection between academic learning and industrial innovation. Educators are encouraged to adopt these AI-augmented tools to enhance conceptual understanding and ensure that the next generation of chemical engineers is fully prepared for the data-centric paradigm of modern science and technology.

Keywords:

Artificial Intelligence (AI), Chemical engineering, Information Problem Solving (IPS), Bachelor's Theses, education.