In Progress 2026 - Present

Generative AI Public Discourse on Bilibili

A large-scale computational study of how users on a youth-oriented video platform understand, use, debate, and imagine generative AI.

TopicGPT Generative AI Computational Communication

Current Stage: Analysis & Writing

Topic discovery and human review are complete; current work focuses on longitudinal analysis, interpretation, and manuscript development.

Overview

This project examines public sensemaking around generative AI on Bilibili. It treats Bilibili as a youth-oriented cultural platform while avoiding assumptions about the age of individual commenters.

5,314 Videos
5,112,810 Deduplicated comments
126 Final topics
6 Analytical categories

Method at a Glance

The workflow combines deterministic preprocessing, semantic embeddings, large-scale K-Means topic discovery, TopicGPT-informed interpretation, and human audit. The resulting 119 included topics are organized into 6 analytical categories for descriptive temporal and interaction analysis.

  1. 01

    Prepare

    Clean and deduplicate

  2. 02

    Embed

    Encode comment meaning

  3. 03

    Discover

    Identify topic structure

  4. 04

    Interpret

    Name and audit topics

  5. 05

    Analyze

    Compare time and interaction

Back to all research