Original Data and Results：Creator-driven human-computer interaction on UGC video platforms

This dataset is collected from Bilibili, a mainstream UGC video platform in China, using Python crawlers and official open APIs. The final dataset consists of 353,501 video records created by 25,147 unique content creators, with multi-layered observational indicators covering both creator-level and video-level attributes, designed to empirically explore the correlations between creator-initiated interactive widgets and two core user engagement outcomes: online tipping (video coins) and comment-based content co-creation. The data collection process starts with generating billions of random user IDs to verify valid platform accounts. We only retain creators who have published at least one video, and remove invalid samples including removed videos, paywall-locked content and records with missing features. The core independent variables are two binary dummy variables representing affective interaction and cognitive interaction, operationalized by Bilibili’s built-in triple-interaction pop-up and voting pop-up respectively. The dependent variables are the number of coins and comments each video receives, log-transformed to mitigate skewness and heteroskedasticity. Multi-dimensional control variables are included: video-level covariates cover video resolution, publication duration, video length and total views; creator-level covariates include total upload quantity, gender dummy variables, follower count, account level, VIP membership and official verification tier. We also label videos as entertainment or knowledge categories to test content-type moderation, and add a grouping indicator to identify creators who deployed interactive widgets for cross-content spillover analysis.本数据集采集自中国主流用户生成内容（User Generated Content, UGC）视频平台哔哩哔哩（Bilibili），采集工作结合Python爬虫与官方开放API完成。最终数据集包含25147位独立内容创作者产出的353501条视频记录，涵盖创作者层面与视频层面的多层观测指标，旨在实证探究创作者发起的互动组件与两类核心用户互动结果的关联：在线打赏（视频硬币）与基于评论的内容共创。 数据采集流程首先生成数十亿随机用户ID以验证平台有效账号。研究仅保留至少发布过一条视频的创作者，并剔除各类无效样本，包括已删除视频、付费墙锁定内容以及存在特征缺失的记录。核心自变量为两类二元虚拟变量，分别指代情感互动与认知互动，其操作化定义分别对应哔哩哔哩内置的三连互动弹窗与投票弹窗。因变量为单条视频获得的硬币数与评论数，我们对二者进行对数变换以缓解偏态分布与异方差问题。 数据集纳入多维控制变量：视频层面协变量涵盖视频分辨率、发布时长、视频时长与总播放量；创作者层面协变量包括总投稿量、性别虚拟变量、粉丝数、账号等级、VIP会员身份与官方认证层级。此外，我们将视频标记为娱乐类或知识类，以检验内容类型的调节效应；同时新增分组标识，用于识别部署了互动组件的创作者，从而开展跨内容溢出效应分析。