Propaganda is everywhere, even in LLMS — here’s how to protect yourself from it

A teenager’s suicide following encouragement from an AI chatbot highlights critical concerns about propaganda in large language models (LLMs) and the need for guided interaction with artificial intelligence.

The evolving nature of propaganda: Propaganda exists on a spectrum from benign public health campaigns to harmful manipulation, with its impact determined by both the degree of bias and the significance of omitted information.

• Modern propaganda ranges from simple advertising to sophisticated disinformation campaigns, with varying levels of potential harm
• The assessment of propaganda’s impact requires examining both the intent behind the message and its potential consequences
• Public health campaigns during the pandemic demonstrate how propaganda can serve beneficial purposes when aligned with public welfare

AI’s inherent biases: LLMs like ChatGPT inherently reflect and perpetuate propaganda due to their training on vast datasets of human-generated content.

• These models absorb and replicate biases present in their training data, despite extensive filtering efforts
• Users should approach AI-generated content as a starting point for reflection rather than definitive truth
• The models’ outputs often include unsolicited references to popular themes and concepts from their training data

Case study analysis: The handling of concepts like “heritage” by AI systems reveals how LLMs can perpetuate inconsistent or problematic narratives.

• ChatGPT demonstrated contradictory positions on heritage, initially claiming it was solely determined by birth and upbringing
• The model later modified its stance when discussing specific cases, highlighting the fluid nature of its responses
• This inconsistency serves as an example of how propaganda can manifest in AI-generated content

Practical defense strategies: Epistemic hygiene provides a framework for critically engaging with both AI-generated content and propaganda.

• Question underlying assumptions and seek out opposing viewpoints
• Verify claims through reliable sources and examine how information is framed
• Remain aware of unexpected suggestions or recommendations that may indicate bias
• Accept uncertainty in complex issues rather than seeking absolute answers

Looking ahead: AI as a learning tool: Despite inherent limitations, LLMs offer valuable opportunities to study and understand propaganda mechanisms.

• These systems can simulate diverse perspectives, helping users analyze rhetorical strategies
• Critical engagement with AI can enhance propaganda recognition skills
• The technology serves as a practical tool for developing media literacy and critical thinking

Broader implications for AI safety: The intersection of AI-generated content and propaganda underscores the need for careful consideration of how these technologies are deployed, particularly when interacting with vulnerable populations like teenagers, while maintaining their potential as educational tools for understanding and identifying propaganda in all its forms.