Analysis of Comments on the Video Interview Featuring Huang Kuo-Chang and Puma Shen by Chifei Fan

Recently, a bill addressing legislative overreach, proposed by the Kuomintang and the People’s Party in the Legislative Yuan, has generated significant controversy. In response, Chifei Fan (范琪斐) invited Huang Kuo-Chang (黃國昌) and Puma Shen (沈伯洋) to appear separately on a program for a corresponding discussion. Given the fervent nature of the comments below the discussion, an analysis was undertaken to ascertain whether there was coordinated inauthentic behavior among the commentators.

Considering that only the accounts of commenters are visible on YouTube, and it is not feasible to identify viewers or users who liked the video, analysis is presently limited to the commenters.

Initially, we utilized the YouTube API to obtain the list of comments on the videos. The interview video featuring Huang Kuo-Chang accrued 11,258 comments from 9,162 users, whereas the interview video with Puma Shen amassed 6,766 comments from 5,376 users.

Initially, we perform an analysis based on the publish dates of the commenters’ channels (i.e., the users’ registration dates). This constitutes a preliminary examination, as bot accounts are frequently registered in batches. By employing this method, we aim to identify and analyze collectively registered bot accounts. However, there is a possibility that individual users may coincidentally register at similar times, making it challenging to distinguish them. Consequently, this analysis may inadvertently include ordinary users and should therefore only be used for a general status assessment.

We have determined that the registration dates for the commenters listed below in the interview video featuring Huang Kuo-Chang and Puma Shen were more than ten days prior:

In the comments section of the interview video featuring Huang Kuo-Chang and Puma Shen, the dates corresponding to the top 10 registration dates for both individuals are highlighted in yellow.

Notably, a substantial number of accounts were registered during a similar timeframe, which warrants considerable suspicion.

We utilized the registration dates of the initial 30 accounts to extract comments from the two interview videos in question. Our findings indicate that among the accounts registered on these dates, there were 209 bot accounts associated with the interview video featuring Huang Kuo-Chang, constituting 2.28% of the total commenting accounts, and a cumulative 248 comments, representing 2.2% of the total comments. In contrast, the interview video featuring Puma Shen had 236 bot accounts, accounting for 4.45% of the total number of commenting accounts, with a cumulative 287 comments, comprising 4.24% of the total comments.

We proceeded to meticulously annotate the comments from these accounts individually and compiled the identified targets of their attacks. The findings are as follows:

The identified targets of bot accounts attack

Based on the analysis of the engagement metrics in the interview videos featuring Huang Kuo-Chang and Puma Shen, it is clear that the operating strategies of the accounts on these two channels are aligned. In the video interview with Huang Kuo-Chang, the principal target of critique from bot accounts was Chifei Fan, who faced accusations of bias (53%), while a smaller percentage of the critiques were aimed at Huang Kuo-Chang (8.5%) and the Democratic Progressive Party (DPP) (4.9%). Conversely, in the interview video with Puma Shen, the majority of criticisms from bot accounts were directed towards Puma Shen (33.2%), followed by critiques of Chifei Fan (17.8%) and the DPP (9.8%).

The aforementioned constitutes a preliminary analysis of the remarks made in response to these two videos.

--

--