Large Language Models Exhibit Communication Bias Impacting News
Authors Stefan Schmid and a technology law scholar report in a forthcoming Communications of the ACM paper that large language models exhibit communication bias, subtly emphasizing certain viewpoints while minimizing others and showing persona-based steerability. The paper finds such framing can shape public opinion even when factual accuracy holds, and argues that EU rules like the AI Act and DSA do not adequately address these nuanced biases, recommending competition, transparency, and user-driven accountability.
Key Points
- 1Demonstrate that large language models exhibit communication bias, emphasizing some viewpoints while downplaying others
- 2Show that persona-based steerability and sycophancy can subtly alter public opinion formation over time
- 3Recommend competition, transparency, and user-driven accountability beyond regulation to mitigate subtle framing biases
Scoring Rationale
Peer-reviewed evidence and wide platform implications drive score, but paper offers limited immediate technical mitigation steps.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Logistics & Shipping problems
