YoronChizu 2024 Open Data

Dataset Overview

• Release Date: November 18, 2024
• Publishing Organization: Mielka, Japanese non‑profit organization
• Data Source: User voting data during the 2024 Japanese House of Representatives election
• Data Types: CSV files and images
• Data Volume: 4,403 independent users participated
• License: CC BY 4.0

Dataset Content

• Voting Data: Anonymous votes on specific issues extracted from party manifestos.
• Image Data: Visualizations of user opinion distribution.

Technical Details

• Service Tool: Polis used as the server‑side deliberation tool.
• Frontend Development: Re‑engineered from scratch for the Japanese context, optimized for mobile access.
• Features:
  • Visualize party positions using party icons
  • AI‑driven opinion cluster explanations
  • Real‑time map updates of user opinion locations

Data Analysis

• Unrepresented Opinion Groups: Certain economic policy opinion groups are not fully represented by existing political parties.
• Vote‑Cluster Relationship: Experiments show that even with 4,400 votes, the four clusters remain clearly distinct.
• Digital Democracy Opinions: Public opinion on digital democracy is divided into four teams: Red (advertising regulation & AI scepticism), Blue (digital integration & efficiency advocacy), Yellow (AI utilization & personalization), Green (privacy protection advocacy).

Future Challenges

• Opinion Extraction: Extracting unbiased opinions from party manifestos requires manual effort; LLM hallucinations pose risks.
• Data Representation: LLMs exhibit errors when explaining complex opinion distributions, necessitating further optimization of data representation.
• User Opinion Submission: Real‑time submission faces quality assurance, misinformation, and dimensionality‑increase challenges.