Skip to content

Taxonomy and typology

Systematic frameworks for categorizing types of misinformation, deception, and information disorder. A taxonomy organizes known phenomena into mutually exclusive and collectively exhaustive categories, enabling precise description, measurement, and comparison of research findings across studies.

Taxonomy dimensions in misinformation research

Taxonomies in this field vary in scope and grain size:

  • Source intent (Wardle & Derakhshan, 2017): intentional (disinformation) vs. accidental (misinformation)
  • Content type (Tandoc et al., 2018): news, rumor, urban legend, propaganda, satire, etc.
  • Mechanism (this work and prior): how false narratives are constructed — mismatch, manipulation, fabrication, exaggeration
  • Multimodal dimension (Dufour et al., 2024; this work): image type, text-image relationship, emotional tone
  • Domain: health, politics, science, finance, etc.

Design challenges

  • Ambiguity: Real-world instances often exhibit multiple characteristics or fall between category boundaries, requiring clear operational definitions
  • Scale: Small taxonomies risk oversimplification; large taxonomies become unwieldy and require expert judgment to apply reliably
  • Mutually exclusive vs. overlapping: Some misinformation instances display more than one mechanism; taxonomies must decide whether to allow multi-label assignment or enforce single labels
  • Grounding: Taxonomies developed by researchers without grounding in real-world data risk missing actual patterns; those grounded only in expert intuition risk researcher bias

Key papers

  • MMMMM: A Unified Taxonomy for Investigating the Mechanisms of Multilingual MultiModal Misinformation — Proposes unified taxonomy grounded in 26,979 real-world multimodal instances across seven languages, with six orthogonal axes to separate mechanism (Mismatch, Manipulation, Fabrication, Exaggeration), image properties, rhetorical function, topic/narrative, and emotional content; operationalizes via Vision Language Model annotation pipeline
  • Dufour et al. (2024) — AMMeBA: Comprehensive multimodal misinformation taxonomy covering three top-level categories (Content manipulation, Context manipulation, Text-based manipulation) with multiple orthogonal sub-axes