Project Description
This collaborative research project addresses the critical challenge of disinformation by integrating advanced computer science methodologies with social science analysis. The core goal is to test the efficacy of AI-driven counter-speech against emotionally persuasive propaganda circulating on YouTube across four pre-selected influencer channels, each serving as a representative case study of disinformation dissemination. The project is driven by three objectives. The first is Technical Validation of AI Agents: the research team will collaboratively build and deploy a standardized Large Language Model (LLM) Agent using the n8n workflow engine. This customized agent must be capable of generating highly consistent outputs, including the specialized rhetorical coding of disinformation and the creation of standardized counter-narratives. The second objective is Comparative Narrative Analysis: four student researchers will analyze how the four stages of Disinfolklore—Crisis, Betrayal, Call to Action, and Moral Restoration—are emphasized and adapted across their four assigned political contexts, assessing the universality of these persuasive structures within the English-speaking digital sphere. The final objective is the Measurement of Platform Resistance, which involves quantifying how the target YouTube channels or the platform's algorithms react to corrective comments over time
Tasks and Responsibilites
The student tasks are divided into three distinct phases: Phase I: Agent Configuration is the intensive setup period where students define the Agent's "brain" and its rules. They collaboratively design the n8n workflow, defining the System Prompts that guide the LLM's reasoning and integrating external web search tools for verification. Phase II: Data Generation & Intervention involves running the experiment. Each student uses the finalized Agent to execute the Disinfolklore Coding for their channel, generate the F–M–F–F comment, and then manually post the approved comment on the target channel. This phase concludes with a seven-day monitoring period where students log all moderation actions and community pushback into a centralized database. Phase III: Analysis and Reporting focuses on synthesizing the results. The team conducts a comparative analysis of the four channel datasets, leading to the co-authorship of a Joint Report. Each student contributes their specific findings, technical details of the Agent design, and the final comparative discussion on platform tolerance, ensuring the completion of all required academic documentation.
Desired Qualifications
None listed.