Synthetic public opinion polling: AI personas don't always change their minds like real voters do
A voter surveyed in October may give a different answer by January. An AI persona might not, depending on how it's built.
Synthetic polling, which uses AI-generated personas to represent voters, may not accurately capture shifts in public opinion over time. Professor Jennifer Stromer-Galley from the School of Information Studies warns that AI personas built on fixed instructions and demographic information tend to maintain consistent answers across multiple surveys, despite real voters changing their minds due to evolving circumstances.
Traditional polling surveys real people at each stage, capturing opinion changes, while synthetic personas are built on consistent instructions, resulting in narrower variance among responses. The concern lies in how AI models may introduce bias, as seen in a Pew Research Center experiment where AI results differed from human results by an average of 12 percentage points.
Synthetic polling may overlook outliers and underrepresent atypical views, potentially distorting public perception. While synthetic polling firms can adjust demographics, they cannot account for individual outliers or model bias.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
