AI chatbots are infiltrating social-science surveys - and getting better at avoiding detection
Briefly

AI chatbots are infiltrating social-science surveys - and getting better at avoiding detection
"People get paid to participate in online surveys - anywhere from pennies to US$100 or more per hour. And an industry was created to administer the surveys and manage vast pools of potential respondents. Between 2015 and 2024, the use of online surveys in published studies increased four-fold, and with that explosion came people trying to game the system, from simply giving fake answers to deploying bots that impersonate individuals; the industry has had to build in checks and tools to root out fraud."
"In November, Sean Westwood, a political scientist at Dartmouth College in Hanover, New Hampshire, demonstrated an AI chatbot that can reliably impersonate a human participant and evade most known mechanisms built into surveys to detect fake responses. Westwood used OpenAI's o4-mini, an AI-based reasoning tool, to build a bot and set it loose on a survey that he designed for the purpose of testing it."
Online surveys have become essential infrastructure for social‑science fields such as ecology, psychology, economics and politics, enabling paid participation from people's desktops. An industry emerged to administer surveys and manage large respondent pools. Growth from 2015 to 2024 quadrupled use in published studies, prompting fraud challenges from fake answers and bots. AI enables sophisticated chatbots that can impersonate human respondents and evade detection. A demonstration showed an AI built with OpenAI's o4-mini passed attention checks nearly perfectly. Such impersonation risks corrupting or invalidating thousands of studies unless survey platforms strengthen defenses.
Read at Nature
Unable to calculate read time
[
|
]