The Social Construction of Reality within Parliamentary Discourse

Principal Aim

This project seeks to develop, test, and ethically evaluate a novel computational methodology — perspectival sentiment mapping — which analyses how individuals and groups construct and contest social realities through pronoun-positioned sentiment toward shared concepts. Using UK Parliamentary Records (Hansard) as a grounded, publicly available testbed, the study will assess whether linguistic perspective structures correlate with observable political behaviour (e.g. voting records), and explore the implications for wider applications in primary care mental health, multi-agency working, and community-level social analysis.

 

Background and Rationale

Public institutions increasingly rely on sentiment analysis to understand public opinion, professional disagreement, and social conflict. However, conventional sentiment analysis treats opinion as an individual psychological attribute rather than a relational and perspectival phenomenon. This limits its usefulness in contexts where disagreement reflects differing constructions of reality rather than simple attitudinal polarity.

Linguistic research, social psychology, and clinical theory converge on the insight that pronoun usage encodes perspective, agency, and social alignment. In political, clinical, and community discourse, pronouns (“I”, “we”, “they”) structure who is positioned as agent, who is aligned, and who is othered. These structures are central to:

• contested political realities,
• professional disagreement in multi-agency systems,
• interpersonal conflict and relational distress in mental health contexts.

This project proposes an ethically constrained methodology that maps sentiment toward shared concepts as expressed from different pronoun positions, producing a structured representation of how realities are linguistically constructed and contested.

Hansard provides an ideal ethical test case: it is public, well-documented, attributed, temporally structured, and linked to independent behavioural records (votes). Demonstrating validity here provides grounding before any application to sensitive domains such as health or community data.

 

Research Questions

1. Can pronoun-positioned sentiment toward shared policy concepts be reliably extracted from parliamentary discourse?
2. Do these perspectival sentiment profiles cluster in ways that align with, but are not reducible to, party affiliation?
3. Do divergences in perspectival sentiment correlate with observable legislative behaviour (e.g. voting divergence, rebellion)?
4. What ethical insights emerge regarding the interpretation of contested realities, responsibility, and agency in institutional discourse?
5. Under what ethical constraints could this methodology be responsibly adapted for community and primary care mental health contexts?

 

Methodology

Data Sources

Primary dataset:

• UK Parliamentary Records (Hansard), accessed via official XML/API archives.

Supplementary dataset:

• Public voting records (e.g. UK Parliament API, Public Whip).

All data are publicly available, non-sensitive, and pertain to elected officials acting in their official capacity.

 

Sampling Strategy (Initial Phase)

The pilot phase will focus on:

• A defined parliamentary period (e.g. one parliamentary session or two-year window).
• 3–5 policy domains with high levels of contestation (e.g. housing, immigration, welfare, policing, climate/energy).
• All MPs contributing to debates within those domains during the sampling window.

This bounded sampling ensures analytical tractability while preserving institutional and ideological diversity.

 

Analytical Procedure

One Excel spreadsheet will be created for every concept identified, with rows populated by MPs and columns populated by the first/second and third person, singular or plural, pronouns they use in reference to the concept identified. A value will be assigned to each cell based on the frequency and intensity of positive or negative sentiments expressed by each MP towards the concept identified. These values will be used to conduct multivariate statistical analyses looking at how MPs positions towards the concepts cluster and change over time. 

1.    Text Processing: Tokenisation, part-of-speech tagging, and dependency parsing using established NLP libraries.

2.    Concept Identification: Policy concepts defined via curated keyword sets and validated through embedding-based expansion.

3.    Pronoun Positioning: Identification of pronoun usage (e.g. I, we, they, you, impersonal/passive forms); Contextual classification of collective pronouns (e.g. government, party, nation) using debate metadata and linguistic heuristics.

4.    Sentiment Attribution: Local sentiment analysis using scientifically approved coding methods, applied at sentence or clause level, linked specifically to concept mentions and pronoun positions.

5.    Perspectival Stance Vectors: For each MP, construct multidimensional vectors capturing sentiment toward each concept from each pronoun position.

6.    Mapping and Analysis: Dimensionality reduction and clustering to visualise perspectival similarity and divergence; Comparison with voting behaviour and party alignment; Temporal analysis to observe shifts preceding behavioural change.

Validation and Evaluation

Validity will be assessed through:

• Structural validity: recovery of known political alignments without supervised labels.
• Explanatory value: identification of meaningful intra-party variation and cross-party alignment.
• Temporal coherence: alignment between linguistic shifts and subsequent voting behaviour.

The project explicitly avoids claims of prediction or causal inference.

 

Ethical Considerations

Data Ethics

• All data are public and pertain to professional political activity.
• No private individuals are analysed.
• No personal or sensitive data are collected.

Interpretive Ethics

• The methodology does not label beliefs, motivations, or psychological traits.
• Findings are framed as structures of discourse, not properties of persons.
• Outputs are aggregated and contextualised to avoid reputational harm or misinterpretation.

Downstream Application Safeguards

This project functions as an ethical proving ground. Any future extension to community or health contexts would require:

• explicit informed consent,
• strict anonymisation and aggregation thresholds,
• separation of analytical insight from decision-making authority,
• prohibition of individual profiling, prediction, or enforcement use.

The methodology is designed to support institutional reflexivity and listening, not surveillance or control.

 

Upscaling and Future Phases

Following successful validation, the methodology could be ethically adapted to:

• Community discourse analysis (e.g. social media, consultations), using aggregate, anonymised data to understand relational sentiment toward public services.
• Multi-agency professional communication, supporting reflective practice and coordination without performance monitoring.
• Primary care mental health research, as a supplementary tool for understanding relational narratives and contested realities, never as a diagnostic instrument.

Each upscaling step would require a distinct ethics review and participatory governance model.

 

Wider Implications

This research contributes to:

• political science (discourse and institutional behaviour),
• computational social science,
• ethics of AI and public-sector analytics,
• mental health and systems-based care theory.

This research seeks to understand how interpersonal attachment relationships within a defined social network evolve over time. By reframing sentiment as perspectival and relational, the project offers a more humane and analytically precise way of understanding disagreement, conflict, and plural realities in democratic and care-based institutions. 

Conclusion

This proposal outlines a carefully bounded, ethically robust study that grounds a novel analytical methodology in a transparent, public, and high-stakes institutional setting. Demonstrating validity and limitations in this context provides a responsible foundation for any future application to more sensitive social and clinical domains.

The project prioritises interpretive caution, ethical governance, and public value throughout.