Story
A conceptual framework for ideology beyond the left and right
Key takeaway
Researchers propose a framework to measure ideologies beyond a simple left-right spectrum, which could help better understand complex political views.
Quick Explainer
The proposed conceptual framework models ideologies as attributed, multi-level socio-cognitive networks. Key components include attitudes, beliefs, rationalizations, and values, which exhibit coherence across subject matter, time, social groups, and concepts. This framework goes beyond the traditional left-right political spectrum, allowing for the computational study of ideologies as complex, nuanced constructs. By connecting disparate NLP tasks like stance detection and framing analysis through the lens of ideology, the framework offers a more theoretically-grounded and data-driven approach to understanding the role of ideology in language and cognition.
Deep Dive
A Conceptual Framework for Ideology Beyond the Left and Right
Overview
This paper proposes a conceptual framework for studying ideology that goes beyond the typical left-right political spectrum. The key contributions are:
- Defining ideology as an attributed, multi-level socio-cognitive concept network, and explaining how ideology relates to other constructs like framing and identity
- Providing actionable guidelines for operationalizing this framework to computationally study ideology in discourse
- Outlining new directions for detecting ideologies in humans and language models, and measuring their downstream impacts
- Showing how this framework can unify disparate NLP tasks like stance detection, framing analysis, and natural language inference
Key Concepts of Ideology
The framework defines the following as core components of ideologies:
- Attitudes: Tendencies to evaluate an object positively or negatively
- Beliefs: Determinations about the truth of a statement or proposition
- Rationalizations: Reasons that justify the truth of relevant beliefs
- Values: More abstract and prescriptive concepts that guide attitudes and beliefs
Ideologies are distinguished from related constructs like:
- Social identity: One's social position, role, or group membership
- Expressed issue stance: Stated preferences over a governmental policy
- Frames: The discursive choices that emphasize particular aspects of a situation
Modeling Ideological Structure
The framework models ideologies as attributed, multi-level networks with two key relationships:
- Entailment: Cognitive associations where a view on one concept is expected to bring to mind a view on another
- Composition: Where one concept is a necessary component of another (e.g. beliefs require rationalizations)
Ideologies should exhibit four types of coherence:
- Subject matter coherence: Ideologies only exist around particular domains of social life
- Temporal coherence: Ideologies have a core of temporally stable concepts and relations
- Partial social coherence: Ideologies are socially shared but not universally held
- Conceptual coherence: Concepts within an ideology should be more strongly related to each other than to external concepts
Implications for NLP
This framework suggests new ways to computationally study ideology:
- Moving beyond fixed ideological spaces toward discovering coherent ideological networks
- Connecting disparate NLP tasks like stance detection, framing, and natural language inference through the lens of ideology
- Reinterpreting existing work on ideological biases in language models, considering their multiplex and complex nature
Overall, the framework advocates a more nuanced, theoretically-grounded, and data-driven approach to understanding ideology through computational methods.
