Recognising and embracing AI in research

• Generative AI in academia offers significant potential for augmenting research workflows, provided it is used to support rather than automate human intellectual work.
• Critical risks include "hallucinations," fabricated citations, and the dissemination of misinformation stemming from low-quality training data.
• Purpose-built academic AI tools often struggle with paywalled content and limited full-text analysis, necessitating integration with traditional research methods.
• Academic integrity increasingly requires transparent disclosure of AI assistance and the development of personal ethical frameworks for its use.

The Automation vs. Augmentation Divide

• Automation involves full delegation of tasks like drafting, which threatens scholarly rigor and critical thinking.
• Augmentation focuses on assistance, such as refining outlines and summarizing materials, keeping the human researcher in control.

Risks to Integrity

• AI models are prone to hallucinating facts and creating fake academic references.
• Misinterpretations can propagate misinformation; for example, the non-existent term "Vegetative Electron Microscopy" appeared in over 20 published papers due to AI error.
• Large language models are trained on unverified internet content, lacking the rigorous standards of scholarly oversight.

Specialized AI Tools

• Platforms like Scite, Research Rabbit, Elicit, and Inciteful are designed for academic workflows, assisting in literature discovery and citation mapping.
• Limitations include:
  • Exclusion of paywalled literature.
  • Reliance on abstracts rather than full-text analysis, missing methodological nuance.
  • Difficulty in distinguishing between scholarly agreement and contradiction.

Strategic Implementation

• AI is particularly beneficial for non-native English speakers and researchers in the Global South by bridging linguistic and access gaps.
• Best practices for researchers:
  • Use tools like NotebookLM and SciSpace for conceptual synthesis alongside traditional library and search methods.
  • Establish personal ethical frameworks regarding bias and plagiarism.
  • Comply with journal policies regarding the disclosure of AI tool usage to maintain transparency and trust.