Human Experts or Artificial Intelligence? Algorithm Aversion in Fake News Detection

The sheer volume of social media posts is forcing social platforms to supplement human-based fake news detection with machine-learning algorithms. Hence, this study investigates algorithm aversion in the context of fake news detection. Based on an empirical study with 238 subjects using artificial fake social media posts, our results suggest that fake news flags have a weaker warning effect if the warning contains additional information on the method used to detect fake news using either human expert opinions or artificial intelligence (AI). Algorithm aversion was reflected in participants assuming that AI algorithms erroneously classify posts as fake more often than human experts do.Furthermore, results showed that flags based on human experts have a more substantial effect on user behavior than neutral flags. These insights can practically be used to assist social media platforms in how to inform their users about the underlying method used to detect fake news