CoChat

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

Open access
Klug, Daniel, Qin, Yiluo, Evans, Morgan, Kaufman, Geoff · 2021 · 13th ACM Web Science Conference 2021

The DOI matches its Crossref record; 56 of the 56 references it cites were found in a registry or an index.

Abstract

The short-form video sharing app TikTok is characterized by content-based interactions that largely depend on individually customized video feeds curated by the app’s recommendation algorithm. Algorithms are generally invisible mechanisms within socio-technical systems that can influence how we perceive online and offline reality, and how we interact with each other. Based on experiences from consuming and creating videos, users develop assumptions about how the TikTok algorithm might work, and about how to trick and please the algorithm to make their videos trend so it pushes them to other users’ ‘for you’ pages. We conducted 28 qualitative interviews with TikTok users and identified three main criteria they assume influence the platform’s algorithm: video engagement, posting time, and adding and piling up hashtags. We then collected 300,617 videos from the TikTok trending section and performed a series of data exploration and analysis to test these user assumption by determining criteria for trending videos. Our data analysis confirms that higher video engagement through comments, likes, and shares leads to a higher chance of the algorithm pushing a video to the trending section. We also find that posting videos at certain times increases the chances of it trending and reaching higher popularity. In contrast, the highly common assumption that using trending hashtags, algorithm related hashtags (e.g. #fyp, #foryou), and piling up trending hashtags would significantly push videos to the trending section was found not applicable. Our results contribute to existing research on user understanding of social media algorithms using TikTok as an example for a short-video app that is explicitly built around algorithmic content recommendation. Our results provide a broader perspective on user beliefs and behavior in the context of socio-technical systems and social media content creation and consumption.

Overview

Integrity signals

Registry

Crossref · 10.1145/3447535.3462512

Cites

56 of 56 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Verification

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Citations

Cites

Research works printed but not linked: 3

  1. danah boyd. 2010. Social network sites as networked publics: Affordances, dynamics, and implications. Networked self: Identity, community, and culture on social network sites (2010), 39–58.
  2. 10.5555/2500954 · doi:10.5555/2500954
  3. Robyn Longhurst. 2003. Semi-structured interviews and focus groups. Key methods in geography 3, 2 (2003), 143–156.

Cited by

See all 392

Metadata

Authors
Klug, D., Qin, Y., Evans, M., & Kaufman, G.
Affiliations
Carnegie Mellon University (US)
Venue
13th ACM Web Science Conference 2021
Publisher
ACM
Published
2021-06-21
Pages
84-92
OpenAlex
W3176799022
Cited by
392
References
54
Citations by year
2026: 105 · 2025: 102 · 2024: 91 · 2023: 64 · 2022: 26 · 2021: 4
FWCI
96.53
Type
conference paper
Language
en
Open access
Locations
ACM Web Science Conference (WebSci) (publishedVersion, cc-by)
Topics
  • Social Sciences › Communication › Media and Digital Communication
  • Physical Sciences › Computer Science › Information Systems › Recommender Systems and Techniques
  • Social Sciences › Communication › Social Media and Politics
Keywords
TikTok, recommendation algorithms, algorithmic transparency, short-form video, video engagement, posting time, hashtags, trending videos, content creation, socio-technical systems, user behavior

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.