Using conversational AI to reduce science skepticism.

Area of Science:

• Science communication
• Artificial Intelligence
• Public trust in science

Background:

• Widespread mistrust in scientific consensus on critical issues like climate change and vaccination hinders global threat response.
• Generative AI, including large language models (LLMs), poses a risk of spreading misinformation at scale.
• LLMs also present an opportunity to reduce misinformation and conspiracy theories, potentially transforming science communication.

Purpose of the Study:

• To review empirical research on using conversational AI, like ChatGPT, to address mistrust in science.
• To evaluate the potential of LLMs in defusing skepticism around contentious scientific topics.

Main Methods:

• Summarization of the current empirical research landscape on conversational AI and science communication.
• Analysis of studies investigating interactions between users and LLMs on scientific topics.

Main Results:

• Empirical studies show negligible evidence of LLMs reproducing conspiracy theories or misinformation.
• Conversations with LLMs generally decrease participants' science skepticism and belief in misinformation.
• LLM interactions tend to foster greater trust in scientific consensus.

Conclusions:

• Current LLMs show potential as a tool to enhance science communication strategies.
• Safeguards are necessary to maintain informational integrity and public trust when using LLMs.
• LLMs can be a valuable addition to efforts aimed at rebuilding public trust in science.