Board #15 - Does AI Exhibit Human Biases? Visual Framing affects LLM Online Content Assessment and Dissemination in the 2024 USA Presidential Election

• In Event: 72-303 - Poster Session: Emerging Technology and Political Science

Brief Overview

This study uses a survey experiment and simulations to examine ChatGPT-4’s cognitive biases in moderating political content. Analyzing how topic, stance, and modality impact decisions, it finds priming and agreement shape issue ranking and sharing.

Author

• Maya Magee, Dartmouth College