Trick and Please. A Mixed-Method Study On User Assumptions About the TikTok Algorithm

Klug, Daniel; Qin, Yiluo; Evans, Morgan C.; Kaufman, Geoff

doi:10.1145/3447535.3462512

Cited by 90 publications

(62 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Scholarship on TikTok is still in its early days; however, most research focuses on the role of TikTok's algorithm. Research by Klug et al (2021) found that TikTok users are highly aware of TikTok's algorithm and have developed assumptions about how the algorithm works to make their videos trend. A study by Omar and Dequan (2020) found that self-expression explains active involvement on TikTok with users wanting to express themselves and interact with others.…”

Section: Literature Reviewmentioning

confidence: 99%

TikTok’s Spiral of Antisemitism

Weimann

Masri²

2021

Journalism and Media

View full text Add to dashboard Cite

The growing presence of antisemitism on social media platforms has become more prominent in recent years. Yet, while most of the scholarly attention has been focused on leading platforms like Twitter, Facebook, or Instagram, the extremist immigration to other platforms like TikTok went unnoticed. TikTok is the fastest-growing application today, attracting a huge audience of 1.2 billion active users, mostly children and teenagers. This report is based on two studies, conducted in 2020 and 2021, applying a systematic content analysis of TikTok videos, comments, and even usernames. Data were collected twice, in two four-month periods, February–May 2020 and February–May 2021, to allow for comparisons of changes and trends over time. Our findings highlighted the alarming presence of extreme antisemitic messages in video clips, songs, comments, texts, pictures, and symbols presented in TikTok’s content. TikTok’s algorithm is even more disconcerting since it leads to a spiral of hate: it pushes users who unintentionally view disturbing content to view more. Considering TikTok’s young demographic, these findings are more than alarming; TikTok even fails to apply its own Terms of Service, which do not allow content “deliberately designed to provoke or antagonize people, or are intended to harass, harm, hurt, scare, distress, embarrass or upset people or include threats of physical violence”.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

TikTok’s Spiral of Antisemitism

Weimann

Masri²

2021

Journalism and Media

View full text Add to dashboard Cite

show abstract

“…This content-based component of the algorithm is complemented by a second collaborative filtering one. The authors of Klug et al (2021) investigate the principles behind this recommendation algorithm and find some expedients to trick and make it suggest certain trends to other users. They found that some of the aspects taken into consideration by the TikTok's recommendation algorithm are hashtags, time of posting and user engagement.…”

Section: Related Literaturementioning

confidence: 99%

Extracting time patterns from the lifespans of TikTok challenges to characterize non-dangerous and dangerous ones

Bonifazi

Cecchini

Corradini

et al. 2022

Soc. Netw. Anal. Min.

View full text Add to dashboard Cite

One of the key aspects that distinguish TikTok from other social media is the presence of challenges. A challenge is a kind of competition that starts when a user posts a video with certain actions and a certain hashtag and invites other users to replicate the same video in their own way. Most challenges are fun and harmless, but sometimes dangerous challenges are launched as well. The authors of these challenges use various tricks to bypass TikTok’s controls. In this paper, we analyze the lifespans of some TikTok challenges and show how they are very different for non-dangerous and dangerous ones. Then, we deepen our analysis by identifying some time patterns that characterize the two types of challenges. Finally, we test the accuracy of the results obtained on a large set of challenges different from those used during the detection of time patterns. The focus of this paper is the detection of time patterns allowing the classification of challenges in dangerous and non-dangerous ones. This could represent a first step towards an approach for the early detection of dangerous challenges in TikTok.

show abstract

“…Research in this space began by investigating algorithmic experiences through qualitative methods such as in-depth interviews. Particularly in uncovering folk theories, qualitative studies about Facebook (Bucher, 2017), Google News (Powers, 2017), YouTube (Alvarado et al, 2020), Spotify (Siles et al, 2020), and TikTok (Klug et al, 2021) asked users to share their experiences with these platforms. These methods allow participants to express a wide range of emotions about and expectations of with algorithms in their own use, which can also vary widely by platform.…”

Section: Methodological Starting Pointsmentioning

confidence: 99%

“…Recommender system beliefs recognize the role of an algorithm that recommends videos based on who and what content is similar to what the user likes, while company policy beliefs acknowledges paid recommendations and third-party content. On TikTok, users focus on how content could make it to other users' video feeds, and assumed it was a combination of video engagement, posting time, and using the right hashtags (Klug et al, 2021).…”

Section: Folk Theoriesmentioning

confidence: 99%

What Do We Know about Algorithmic Literacy? The Status Quo and a Research Agenda for a Growing Field

Oeldorf-Hirsch¹,

Neubaum²

2021

Preprint

View full text Add to dashboard Cite

The increasing emergence of algorithms in our daily use of technologies comes with a growing field of empirical research trying to understand how aware and knowledgeable individuals are about algorithms. This field is marked by a certain diversity in terms of how it theorizes and measures people’s literacy when interacting with algorithms. We propose converging on the term algorithmic literacy that covers different dimensions used by previous research. This article summarizes the state of knowledge on algorithmic literacy by systematically presenting initial steps in theory building and measurement development. Drawing on this, we propose an agenda including five different directions that future research could focus on: 1) theory building to understand algorithmic literacy, 2) addressing the algorithmic divide, 3) uncovering the relationship between algorithmic literacy and attitudes, 4) examining algorithmic literacy as predictor for user behavior, and 5) exploring ways to increase algorithmic literacy.

show abstract

Trick and Please. A Mixed-Method Study On User Assumptions About the TikTok Algorithm

Cited by 90 publications

References 40 publications

TikTok’s Spiral of Antisemitism

TikTok’s Spiral of Antisemitism

Extracting time patterns from the lifespans of TikTok challenges to characterize non-dangerous and dangerous ones

What Do We Know about Algorithmic Literacy? The Status Quo and a Research Agenda for a Growing Field

Contact Info

Product

Resources

About