Democratic institutions face a growing yet understudied threat from transformative artificial intelligence (TAI). While much research examines AI's economic impacts, its potential to accelerate democratic backsliding—the gradual erosion of democratic norms and processes—remains poorly understood. This gap is particularly concerning given recent democratic declines in the U.S. and other nations, combined with AI's expanding political applications.

Bridging Two Critical Fields

One approach could involve systematically connecting political science research on democratic backsliding with emerging AI capabilities. This would entail:

• Cataloging established backsliding mechanisms (e.g., media manipulation, judicial erosion)
• Analyzing how TAI might amplify these pathways through deepfakes, microtargeting, or automated governance
• Developing testable hypotheses about AI-specific threats to democratic institutions

Unlike existing democracy indices that measure current conditions, this would focus on anticipating future institutional vulnerabilities created by AI's political applications.

Strategic Execution Pathways

The research could progress through phased implementation:

1. Literature review of backsliding theories and AI governance studies
2. Expert interviews across political science and AI safety fields
3. Scenario development for U.S.-specific institutional risks

A minimal version might focus solely on AI's impact through information ecosystems before expanding to examine effects on electoral systems, checks and balances, or civil society.

Unique Value Proposition

This interdisciplinary approach could offer several advantages over existing work:

• Applying rigorous political science frameworks to analyze AI risks rather than treating them as purely technological challenges
• Focusing on institutional erosion rather than just individual manipulation or misinformation
• Providing policymakers with concrete prevention strategies grounded in historical backsliding patterns

By systematically mapping how AI capabilities might interact with known democratic vulnerabilities, this line of research could help develop more robust safeguards for democratic institutions in the AI era.