[Paper Review] Friend or Foe? Exploring the Implications of Large Language Models on the Science System

Benedikt Fecher, Marcel Hebing+3 more|arXiv (Cornell University)|Jun 16, 2023

Artificial Intelligence in Healthcare and Education8 citations

TL;DR

The paper reports a Delphi study with 72 science and AI experts to assess LLM applications, limitations, and implications for the science system, highlighting transformative potential and risks requiring regulation and education.

ABSTRACT

The advent of ChatGPT by OpenAI has prompted extensive discourse on its potential implications for science and higher education. While the impact on education has been a primary focus, there is limited empirical research on the effects of large language models (LLMs) and LLM-based chatbots on science and scientific practice. To investigate this further, we conducted a Delphi study involving 72 experts specialising in research and AI. The study focused on applications and limitations of LLMs, their effects on the science system, ethical and legal considerations, and the required competencies for their effective use. Our findings highlight the transformative potential of LLMs in science, particularly in administrative, creative, and analytical tasks. However, risks related to bias, misinformation, and quality assurance need to be addressed through proactive regulation and science education. This research contributes to informed discussions on the impact of generative AI in science and helps identify areas for future action.

Motivation & Objective

• Assess how large language models (LLMs) and LLM-based chatbots affect scientific practice and the science system.
• Identify potential applications across administrative, creative, and analytical tasks.
• Investigate ethical, legal, and regulatory considerations and required competencies for effective use.
• Highlight risks such as bias, misinformation, and quality assurance and propose actions for education and policy.

Proposed method

• Conduct a Delphi study with 72 experts in research and AI to gather structured expert judgments on LLM use in science.
• Explore domains of application, limitations, and impacts on the science system through iterative surveys.
• Examine ethical and legal considerations and the competencies required for effective use of LLMs in science.
• Synthesize findings to discuss regulatory and educational needs to mitigate risks.

Experimental results

Research questions

• RQ1 What are the potential applications of LLMs and LLM-based chatbots in administrative, creative, and analytical tasks within science?
• RQ2 What limitations and risks (biases, misinformation, quality assurance) accompany LLM use in scientific practice?
• RQ3 What regulatory, ethical, and education-related measures are needed to responsibly integrate LLMs into the science system?
• RQ4 What competencies and skills should scientists develop to effectively and safely use LLMs?

Key findings

• LLMs have transformative potential for administrative, creative, and analytical tasks in science.
• Risks include bias, misinformation, and quality assurance challenges that require proactive management.
• Regulation and science education are needed to address ethical and legal considerations.
• The study contributes to informed discussions and identifies areas for future action in generative AI and science.