Framing bias¶
Framing bias refers to the systematic selection and emphasis of certain aspects of an issue while downplaying or omitting others, thereby shaping audience perception and interpretation. Framing is not inherently deceptive—it is a fundamental feature of communication—but biased framing becomes problematic when it systematically favors particular ideological positions or narratives.
Key concepts¶
Frame selection: The choice of which generic frames (e.g., conflict, morality, economics) or issue-specific frames are applied to a topic. A story can be framed as an economic issue, a moral issue, a legal issue, or a public health issue, each emphasizing different aspects and solutions.
Content vs. style framing: Biased framing manifests in what is discussed (content—which entities, statistics, or arguments are mentioned) and how it is discussed (style—the emotional tone, linguistic polarity, and rhetorical strategies used).
Audience effects: Research shows that frame choice influences audience perceptions of causality, responsibility, and appropriate solutions. Liberal and conservative audiences may interpret the same facts differently depending on how they are framed.
Measurement approaches¶
- Manual frame annotation: Coding news articles against established taxonomies (e.g., Boydstun's 15 frame dimensions) to quantify which frames dominate
- Comparative analysis: Analyzing how the same event is framed across outlets with different political leanings
- Linguistic markers: Identifying words, metaphors, and rhetorical patterns that instantiate particular frames
- Entity and topic extraction: Measuring which entities and topics are frequently mentioned (content bias) and with what sentiment (style bias)
- Automated detection: Machine learning models trained on annotated frame corpora to classify frames in new text
Key papers¶
- Measuring Political Bias in Large Language Models: What Is Said and How It Is Said — Measures framing bias in LLM-generated content by decomposing it into content bias (entity/topic selection) and style bias (lexical polarity)
- DELL: Generating Reactions and Explanations for LLM-Based Misinformation Detection — Includes framing detection as a misinformation detection task
Related topics¶
- Media bias detection — Broader detection of bias in news media, of which framing is one dimension
- Framing detection — Identifying and classifying frames in text
- Propaganda — Deliberate use of biased framing for persuasion and influence
- Political bias in fake news detection — Ideological slant that often manifests through frame selection