PDF Publication Title:
Text from PDF Page: 005
lation coefficiency (ICC) between the three labelers. Based on this analysis, we can determine that all three labelers were in agreement when it comes to the high-level labels, e.g., left-right-center. Besides, there is a high coefficiency in the majority of the granular categories. On the left side of the graph, we can see the intraclass correlation coefficiency values, the estimates of the ”real” information captured by our classification, which ranges from 0 to one. The larger the number, the more similar the tags were. One the right side of the Figure, we see the reviewer agreement in percentages. The ICC values above 0.75 are considered excellent, be- tween 0.75 and 0.59 are good and above 0.4 are consid- ered as fair [48]. In our categorization, few classifications measure under 0.4. However, we believe that the explanation for this convergence is related to the nature of these cate- gories. The low coefficiency scoring of groups,’Provocateur’, ’Anti-whiteness’ and ”Revolutionary,’ could be explained by the labeler’s hesitation to apply these rather extreme labels where consistent evidence was lacking. Besides, since each channel was allowed four different ’soft tags’ defining these subcategories, the channels were likely tagged by the other, milder tags. The rationale behind the lack of agreement on the ’Educational’ label is best explained by the fact that this category classification might be somewhat superfluous. Political content, even educational one, often has a clear bias, and the content already belongs to one or more stronger categories, such as Partisan Left or to channels that are non- political. Fig. 1. The intraclass correlation coefficiency between the three labelers However, if one looks at the percentages of the agreement, the agreement if very high in most cases. The only category where disagreement seems to be significant is the left-right- center categorization. However, this disagreement can be ex- plained by the weighing applied when calculating the ICC factor. To assign a label, we investigated which topics the channels discussed and from which perspective. Some channels are overtly partisan or declare their political stances and support for political parties in their introductions or have posted several videos where such topics are discussed. For example, libertarian channels support Ron and Rand Paul (Libertarian politicians affiliated with the Republican party) or discuss Austrian economics with references to economists such as Frederick von Hayek or Ludwig von Mises or the fictional works of the author Ayn Rand. Comparably, many channels dedicated to various social justice issues title their videos to reflect the content and the political slant, e.g., ”Can Our Planet Survive Capitalism” or ”The Poor Go To Jail And The Rich Make Bail In America” from AJ+. Nevertheless, other channels are more subtle and required more effort to tease out their affiliation. In these cases, we analyzed the perspective that these channels took on political events that have elicited polarized opinions (for example, the nomination of Brett Kavanaugh in the U.S. Supreme Court, the Migrant Caravan, Russiagate). Similarly, we also analyzed the reactions that the channels had for polarizing cultural events or topics (e.g., protests at university campuses, trans activism, free speech). If the majority of these considerations aligned in the same direction, then the channel was designated as left- leaning or right-leaning. If there was a mix, then the channels were likely assigned to the centrist category. The only way to conduct this labeling was to watch the content on the channels until the labelers found enough evidence for assigning specific labels. For some channels, this was relatively straightforward: the channels had introductory videos that stated their political perspectives. Some of the intros are very clearly indicating the political viewers of the content creator; some are more subtle. For example, a polit- ical commentator Kyle Kulinski explicitly states his political leanings (libertarian-left) in channel SecularTalk description. In contrast, a self-described Classical Liberal discussion host Dave Rubin has a short introduction of various guests, pro- viding examples of the political discussions that take place on his channel The Rubin Report. In other cases, the labelers could not assign a label based on introduction or description but had to watch several videos on the channel to determine the political leanings. On average, every labeler watched over 60 hours of YouTube videos to define the political leanings without miscategorizing the channel and thus misrepresenting the views of the content creators. Based on the eighteen classification categories, we created thirteen aggregate groups that broadly represent the political views of the YouTube channels. The eighteen ’soft tags’ were aggregated from ideological groups and better differentiated between the channels. For more details on tagging aggregation, please see the Appendix A-B. These groupings were applied in the data visualization rather than the more granular eighteenPDF Image | Algorithmic Extremism: Examining YouTube’s Rabbit Hole of Radicalization
PDF Search Title:
Algorithmic Extremism: Examining YouTube’s Rabbit Hole of RadicalizationOriginal File Name Searched:
1912-11211.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com (Standard Web Page)