{"id":261333,"date":"2025-04-30T07:19:48","date_gmt":"2025-04-29T22:19:48","guid":{"rendered":"https:\/\/designcopy.net\/en\/ai-study-tested-persuasive-comments-reddit\/"},"modified":"2026-04-06T10:09:45","modified_gmt":"2026-04-06T01:09:45","slug":"ai-study-tested-persuasive-comments-reddit","status":"publish","type":"post","link":"https:\/\/designcopy.net\/en\/ai-study-tested-persuasive-comments-reddit\/","title":{"rendered":"AI Study Quietly Tested Persuasive Comments on Reddit"},"content":{"rendered":"

In a world where AI<\/strong> is already meddling in our opinions, researchers from the University of Zurich<\/strong> decided to push the envelope\u2014secretly. They launched an unauthorized experiment<\/strong> on Reddit\u2019s r\/changemyview<\/strong> subreddit, aiming to test how well Large Language Models (LLMs) could sway users\u2019 views. For four months, AI bots fired off comments, disguised as real people. Oh, the irony\u2014AI playing puppet master in a forum meant for honest debate.<\/p>\n

These bots cranked out 1,783 comments, personalized<\/strong> with scraped data like users\u2019 gender, age, and politics. They used LLMs such as GPT-4 and Llama 3.1, adopting wild personas\u2014from trauma counselors to anti-BLM advocates. Researchers swore they manually checked each post to avoid harm, but come on, who buys that? The results? AI arguments were three to six times more persuasive<\/strong> than humans\u2019. Users handed out 137 \u201cdeltas,\u201d Reddit\u2019s badge for a changed mind, and never suspected a thing.<\/p>\n

\n

AI bots fired off 1,783 personalized comments, from trauma counselors to anti-BLM voices, proving three to six times more persuasive\u2014and users stayed clueless! (see Google’s SEO Starter Guide<\/a>)<\/p>\n<\/blockquote>\n

Ethical red flags exploded everywhere. The study broke subreddit rules by hiding the bots, manipulating folks without consent. Moderators called it \u201cpsychological manipulation,\u201d accusing the team of impersonating sensitive figures like abuse survivors. In response, the University of Zurich\u2019s Ethics Commission issued a formal warning<\/a> to the lead researcher after an investigation. Reddit users and academics erupted in backlash, with one professor labeling it a massive ethics fail. The platform even banned the accounts involved.<\/p>\n

Researchers defended themselves, claiming societal benefits<\/strong> outweighed the mess. They said their university\u2019s ethics board approved<\/strong> it, arguing it highlighted AI\u2019s dangers in a space where people invite challenges. Sure, but sneaking around like that? It\u2019s like testing fire alarms by starting a blaze\u2014bold, but reckless.<\/p>\n

Broader implications linger. Other studies, like one from Cornell, show personalized AI boosts persuasion<\/strong> by 81.7%. This Zurich experiment underscores how AI could warp online discourse<\/strong>, making it tougher to spot fake interactions<\/strong>.<\/p>\n

In the end, it raises a blunt question: Is advancing tech worth the trust we lose<\/strong>? This incident affected the large community of 3.8 million members<\/a> on the subreddit.<\/p>\n


\n