背景:我们旨在分析对使用基于体重的歧视术语的推文的反应,并将其与相同用户在X(以前的Twitter)上发布的其他推文进行比较。
方法:我们收集了以基于体重的歧视术语(研究组)为特征的推文,并通过从每个帐户中随机选择最多五个未提及任何此类术语的推文创建了一个对照组。描述性统计,情绪分析,和逻辑回归模型用于比较研究组和对照组的喜欢和转发数量,并理解与这些推文相关的情绪。
结果:我们的分析包括22,075个研究组推文和50,341个对照组推文。对41,403条(57.2%)推文进行了情绪分析,65.7%的研究小组推文被发现含有负面情绪。研究组喜欢(1[0-4])和转发(0[0-0])的中位数高于对照组(1[0-2]和0[0-0],分别,研究组在两个比较中都获得了更高的平均排名,P<0.001)。多变量逻辑回归分析显示,使用基于权重的判别项的推文获得了更多的喜欢(OR=1.22;95%CI:1.16-1.28)和转发(OR=1.61;95%CI:1.49-1.74),独立于,例如,验证状态,追随者计数,推文的年份和季节,和推文的情感表达。
结论:关于脂肪恐惧症的推文,身体羞辱,类似的条款比相同账户发布的其他条款获得更多的反应。
BACKGROUND: We aimed to analyze reactions to tweets that employed weight-based discrimination terms and to compare them to others posted by the same users on X (formerly Twitter).
METHODS: We collected tweets featuring weight-based discrimination terms (the
study group) and created a control group by randomly selecting up to five tweets from each account that did not mention any such terms. Descriptive statistics, sentiment analysis, and logistic regression modeling were used to compare the numbers of likes and retweets of the study and control groups, and to understand the emotions associated with these tweets.
RESULTS: Our analysis included 22,075
study group tweets and 50,341 control group tweets. Sentiment analysis was conducted on 41,403 (57.2%) tweets, with 65.7% of the
study group tweets being found to contain negative sentiments. The
study group had a higher median of likes (1 [0-4]) and retweets (0 [0-0]) than the control group (1 [0-2] and 0 [0-0], respectively, with the
study group obtaining higher mean ranks in both comparisons, P < 0.001). Multivariable logistic regression analysis revealed that tweets using weight-based discrimination terms gained more likes (OR = 1.22; 95% CI: 1.16-1.28) and retweets (OR = 1.61; 95% CI: 1.49-1.74), independent of, for example, verification status, follower count, year and season of the tweet, and emotional expression of the tweet.
CONCLUSIONS: Tweets concerning fatphobia, body shaming, and similar terms gain more reactions than others posted by the same accounts.