一项基于社交平台数据的分析研究指出，在社交媒体上引发病毒式传播的民调往往不具备统计代表性，其传播机制显著倾向于那些对民主党更有利的结果 1。该研究强调了在评估公众舆论时，必须区分具备科学抽样基础的统计代表性与受社交算法驱动的传播偏差 1。
研究人员针对 Bluesky 平台上 @usapolling.bsky.social 账号过去 90 天的帖子进行了系统性抓取与分析 1。由于 Elon Musk 破坏了 Twitter API 的学术用途，研究者转而使用 Bluesky 作为数据源 1。技术层面，团队利用 Python 进行数据采集，通过 OpenRouter 调用 OpenAI GPT-6 Luna 大语言模型提取民调信息，并借助 R 语言完成统计建模 1。
核心发现显示，随着民调结果偏离平均差值增大且对民主党的有利程度增加，相关帖子的转发数量呈现显著上升趋势 1。具体案例对比进一步印证了这一现象：针对得克萨斯州参议院（TX-Senate）选区的民调更容易引发病毒式传播，而宾夕法尼亚州州长竞选相关的民调则关注度相对较低 1。
Social media polls that achieve viral status often lack statistical representativeness and tend to skew in favor of the Democratic Party. 1 A recent analysis revealed that as poll results became more advantageous for Democrats relative to the average margin, the number of shares for those posts increased significantly. 1
To conduct this study, researchers utilized Python to scrape data from the Bluesky account @usapolling.bsky.social covering the past 90 days. 1 They employed OpenRouter to access the OpenAI GPT-6 Luna large language model for extracting polling information, followed by statistical modeling using the R programming language. 1 The shift to Bluesky was necessitated by Elon Musk’s disruption of Twitter API academic usage, which forced investigators to find alternative platforms. 1
Specific case comparisons highlighted distinct patterns in audience engagement across different races. 1 For instance, polls regarding the TX-Senate district were prone to going viral, whereas those related to the Pennsylvania governor’s race attracted lower attention levels. 1 This research underscores the critical importance of distinguishing between social media propagation bias and true statistical representation when evaluating public opinion. 1
评论
还没有评论，欢迎留下第一条。