Methods & Innovation A quarterly digest of the Center’s latest methodological research and data science discoveries · Subscribe ↗ Preelection polling this year accurately depicted the public’s dissatisfaction with national conditions and where Americans placed the blame. It revealed that Democrats were highly motivated to vote, as well as dissatisfaction among parts of Trump’s 2024 electoral coalition. In the coming weeks and months, further analysis will help us better understand why the election unfolded the way it did and hopefully reveal why polling understated the forces that brought Democrats back into power in Congress and in many states. Scott KeeterSenior Survey Advisor The most common question pollsters hear around election time is “Who’s going to win?” A close second in 2026 was “Are the polls going to be wrong again?” With final vote tallies nearly complete, it’s clear that preelection polls captured the direction of this year’s election while generally understating the size of Democratic Party’s advantage, writes Senior Survey Advisor Scott Keeter. As artificial intelligence becomes more advanced, there is growing interest in using it to stand in for real respondents in public opinion surveys. To learn more about this method, we replicated three human surveys from our American Trends Panel using AI respondents. These synthetic surveys did a poor job of reproducing the real results and were especially error-prone for some subgroups, including Republicans and Black adults. As of 2024, almost 70% of national election polls weight on political party affiliation or report vote choice in a previous election. Though this kind of weighting is controversial, evidence shows that it has made polls in recent elections modestly more accurate. But it can sometimes lead to an overstatement of similarity to past elections, which might have contributed to polls underestimating the extent of Democratic gains in 2026. One of the most urgent problems in online opt-in polling is bogus respondents, or survey-takers who make no effort to answer questions truthfully and instead are just looking to finish them quickly and collect rewards. Our recent study compares three methods for identifying and removing bogus respondents from survey samples but finds that a surefire solution remains elusive. A recent Pew-Knight Initiative report uses a multimethod approach to explore how Americans show up in public life. The project draws from a nationally representative survey of 5,393 U.S. adults, which asked about their political activity, civic involvement, religious attendance and attention to news. Pew Research Center’s Decoded methodology blog focuses on the “how” behind our research. We regularly dig into interesting questions related to survey methods, data science, data visualization and more, so researchers can build on and engage with our work. Read some of our latest posts: Check out our Methods 101 video series, an approachable introduction to the basics of survey research that answers common questions about polls and explores how pollsters are tackling challenges in the industry.
Can AI Stand In for Human Survey-Takers?
Full Article
Original Source
Read the full article at Pewresearch →KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.