How Pew Research Center is – and is not – using AI in our work

At Pew Research Center, we’re approaching synthetic intellect from many perspectives.

As public-facing social scientists committed to innovation, we’re exploring what AI power add to our toolkit. As public opinion researchers, we’re studying and explaining how group are reacting to the advancement of AI technologies. And as information providers whose essential values contain accuracy and methodological rigor, we’re moving alongside awesome deliberation to justify the norm of our activity remains high.

In all of these spaces, our chief dedication is that our activity remains people-centered:

• Real people, not machines, answer our surveys. We don’t use AI to create or example synthetic community opinion. Our study results are according to the views that genuine group study to us. For U.S. polls, for example, we depend on a multimode, probability-based study panel made up of approximately 10,000 adults who are selected at random from throughout the complete country.
• Humans decide what we study. Choosing investigation topics and deciding what to ask the community are among the most consequential things we do. Our researchers create those decisions.
• Humans, not AI, compose and assessment our investigation reports. Humans decide what’s analytically crucial and interesting.
• Accuracy and rigor remain paramount. Humans oversee all aspect of our work, and humans clasp the supreme duty for its norm and accuracy.
• We defend respondents and their data. We never disclose our study respondents’ individual identifying data to any open, community AI tool.
• The photographs that escort our activity are of, and by, genuine people. They’re selected by individual editors. We don’t use AI to create or modify photos. The identical goes for our illustrations.

How we currently use AI

• In the manufacturing of our website. Our engineers use AI code assistants to compose the code underlying pewresearch.org.
• In parts of our investigation process. Researchers may use AI assistants to compose the code needed to prepare, arrange or run inspection on study datasets. AI tools additionally assistance alongside tasks involving textual data, specified as coding open-ended study responses into categories or abrasion websites for key data. But equal whenever we use AI, researchers scheme the inspection plans and construe the data for our audiences.
• In the final stages of our editorial process. Many extensively accessible tools already use AI to assistance spotless up grammar and punctuation. We’re experimenting alongside its use in first phases of copy editing, although humans volition continue to assessment and endorse our final copy.
• In creating derivative products. We’re additionally experimenting alongside using AI to assistance in creating derivative products according to human-authored research, specified as social media posts. AI helps modify existing findings for distinct formats and audiences, but it does not create the Center’s first inspection or current new interpretations. Humans assessment all AI-assisted satisfied previous to publication.

We’ll be transparent

• If we create meaningful use of AI in investigation manufacturing – for example, in analyzing textual data or coding open-ended responses – we’ll depict its use in the report’s methodology section.
• If AI use expands in ways that materially power how our external products are produced, we’ll revisit how we disclose that use publically and update this part accordingly.

We’d affection to comprehend your thoughts, hopes and concerns on this topic.

Note: This is an update of a article initially published on Aug. 27, 2024.