A Study on the Relationship Between the Epistemological Orientation of Parliamentary Discourse and Deliberative Democracy

This study quantifies the epistemological orientation of 15 million parliamentary speeches using the EMI score, finding that evidence-based reasoning is significantly positively correlated with the quality of deliberative democracy and governance quality, providing quantitative evidence for the relationship between political discourse and democratic governance. It focuses on the impact of epistemological orientations (evidence-based vs. intuition-based reasoning) on democratic systems, using computational social science methods to conduct large-scale cross-country analysis.

The core of democratic deliberation is the pursuit of truth, but real-world political discourse has two epistemological orientations:

Evidence-based reasoning: Relies on verifiable factual data and authoritative sources, with rigorous logic and a willingness to revise views; Intuition-based reasoning: Relies on personal beliefs and emotions, and resists counterevidence.

This difference deeply affects the operation of democracy, but there has long been a lack of systematic measurement methods.

The Stanford team proposed the Evidence-Minus-Intuition (EMI) score, with the framework as follows:

Measurement Components

1. LLM Scoring: Multi-dimensional evaluation (citation of evidence, logical clarity, verifiability, reverse scoring for subjective claims);
2. Semantic Similarity: Comparing the embedding similarity between the target speech and evidence-based/intuition-based corpora;
3. Composite Score: EMI = (Evidence-based score + semantic similarity) - (Intuition-based score + semantic similarity), ranging from -10 to +10.

Advantages

Scalable (handles massive text), consistent (avoids human subjectivity), fine-grained (sentence-level analysis), and cross-language support.

Data Scope

• Time: 1946-2025 (nearly 80 years);
• Countries: US, UK, Germany, France, Japan, Canada, Australia;
• Text: 15 million parliamentary speeches;
• Languages: English, German, French, Japanese.

Trend Findings

• 1950s-1980s: High EMI, rational discourse;
• 1990s-2010s: EMI declined, intuition-based discourse increased;
• Post-2010: Intensified divergence (Germany and France maintained high EMI stably, while the US and UK saw significant declines).

Japan showed a 'polite intuition' pattern: low EMI but still high governance quality.

Hypothesis Testing

1. Deliberative Democracy: Correlation coefficient with the V-Dem Deliberative Democracy Index is +0.47 (p<0.001), which remains significant after controlling variables; EMI changes lead by 2-4 years;
2. Governance Quality: Significantly positively correlated with the World Bank's Rule of Law (+0.52), Regulatory Quality (+0.44), and Control of Corruption (+0.38).

Causal Inference

• Panel Data: In years when EMI rises in the same country, the deliberative democracy index tends to increase (for every 1-point increase in EMI, the index increases by 0.3 points);
• Instrumental Variable: Education expansion supports the causal effect of EMI on democratic quality;
• Regression Discontinuity: Changes in electoral systems indirectly affect democratic quality through EMI.

1. Information Environment: High EMI discourse makes fact-checking easier and misinformation spread slower;
2. Political Elite Behavior: In a high EMI environment, policy-making relies on experts, debates focus on effects, and compromise consensus is easier to reach (e.g., cross-party cooperation on Germany's energy policy);
3. Citizen Participation: High EMI is positively correlated with high-quality participation (municipal meetings, reading policies) and negatively correlated with emotional participation (protests, cyberattacks).

Theoretical Implications

Traditional democratic theory focuses on procedures/rights; this study emphasizes the epistemological dimension: the quality of democracy depends not only on 'who has the final say' but also on 'how they say it'.

Policy Recommendations

• Optimize parliamentary rules (speech limits, questioning procedures) to improve discourse quality;
• Balance public/commercial media structures to shape a rational discourse environment;
• Strengthen education to enhance civic literacy and cultivate evidence-based dialogue habits.

Technical Challenges

In the era of social media, information fragmentation and algorithmic recommendations reinforce intuition-based discourse; the post-2010 EMI decline is correlated with the rise in social media penetration.

Limitations

1. Language Limitation: Mainly analyzes Western countries; applicability to non-Western contexts needs verification;
2. Causal Identification: Endogeneity issues may still exist;
3. EMI Measurement: LLM scoring may have cultural biases; cross-language comparability needs verification.

Future Directions

• Expand to emerging democratic countries;
• Develop more refined epistemological classifications;
• Experimentally test the impact of discourse rules on EMI and democratic quality;
• Analyze the impact of AI-generated content on the epistemology of political discourse.