At Pew Research Center, we approach artificial intelligence from a variety of perspectives.
as social scientists in public We are committed to innovation and are exploring what AI can add to our toolkit. as public opinion researcher, We study and explain how people are responding to advances in AI technology. and as information provider Our fundamental values include accuracy and methodological rigor, and we work meticulously to ensure the quality of our work is high.
In all these spaces, Our main commitment is that our work is always human-centered.
- A real person, not a machine, answers the survey. We do not use AI to create or model aggregate public opinion. Our findings are based on opinions reported by real people. For example, U.S. public opinion polls rely on a multimodal, probability-based survey panel of approximately 10,000 randomly selected adults from across the country.
- Humans decide what to learn. Choosing a research topic and deciding what to ask the public is one of the most important things we do. Our researchers make those decisions.
- Humans, not AI, create and review research reports. It is human beings who analytically decide what is important and interesting.
- Accuracy and rigor remain paramount. Humans oversee every aspect of our work, and they are ultimately responsible for its quality and accuracy.
- We protect our respondents and their data. We do not disclose personally identifying information of survey respondents to AI tools.
- The photographs accompanying our work are of real people. These have been selected by human editors. We do not use AI to generate or modify photos. The same goes for our illustrations.
How we use AI today
- In the production of our website. Our engineers use AI code assistants to write the underlying code for pewresearch.org.
- Part of the research process. Researchers can use AI assistants to write the code needed to prepare, organize, and perform analyzes on research datasets. AI tools can also assist with tasks involving text data, such as coding free-form survey responses into categories or scraping websites to retrieve important data. But even when using AI, researchers still design the analysis plan and interpret the data for their audience.
- The final stage of the editing process. Many widely used tools already use AI to clean up grammar and punctuation. We’re experimenting with its use in the early stages of copy editing, but humans will continue to review and approve the final copy.
- In the creation of derivative products. We are also experimenting with using AI to help create derivative products based on human-authored research, such as social media posts. AI can help adapt existing research findings to different formats and audiences; do not have The Center generates its own analysis or introduces new interpretations. All AI-powered content is reviewed by humans before being published.
we become transparent
- If you make meaningful use of AI in your research production (for example, analyzing textual data or coding open-ended responses), describe its use in the methodology section of your report.
- If our use of AI expands in a way that materially impacts how we make external products, we will reconsider how we communicate that use to the public and update this article accordingly.
We would love to hear your thoughts, wishes, and concerns about this topic.
Note: This is an update of a post originally published on August 27, 2024.
