“A recommender system is a software technology that proactively suggests items of interest to users based on their objective behaviour or their explicitly stated preferences. It provides benefits to users and enhances websites’ revenue”
(Pu, 2011).
The algorithm on TikTok is a recommender system, as stated in its policy, recommendations are made by analysing user interactions, shares, likes, watch full or skips, followed accounts, comments and personal content created.
However, even knowing how the recommendations work, the shared sentiment of being watched outside the app is a “joke” among users with hashtags such as #tiktokisspyingonus, or #tiktokiswatchingme, or videos with “TikTok is spying on me” as video description. Not to be confused with scopophobia, this sentiment could be explained as a sentiment of feeling like TikTok knows things that are only available in your dep thoughts or daily experiences that have not been shared with anyone.
As a user, I have had a similar experience. After being approached by homeless cats twice a week, TikTok recommended me a video under the hashtag “cat distribution system”. Without further thought this instance can be interpreted as TikTok’s algorithm reading my mind or following my steps. However, a simple look at my search history helps figure out how this situation came to occur. To explain this phenomenon, it is needed to discuss genres of content and how the algorithm interprets them.
This is the trend most viewed that week by my user:
And the next is the first video I got recommended under the “cat distribution system” hashtag.
Firstly, following Amanda Kavoori’s (2011) system of genre division on YouTube content, the first video will fall under the “the phenom” category as most videos under the sound followed a similar stylistic of editing pictures of beloved pets, and the second will fall under “the witness” category as its filmed in a documentarist manner. Secondly following film and media classic genres (Eastern Arizona College Library, 2023) the first video could be define as comedy and family as it intends to take a smile out of the audience and has no warnings for its consume, meanwhile the second video will fall under drama and adventure as it is full of emotion ions and we follow, even if on a short period of time, a brave action from the witness.
Those videos are vastly different in terms of genre; however, they have two commonalities: cats and a million videos.
Going back to TikTok’s explanation on how their algorithm works, both recommendations were made based on my consumption preference and the virality of those videos. For numeric reference, my account has a total of 115 videos liked and 35 of these, 30.4%, are pet centered, mostly cats. I entered the app 20 times in a day and in 6 of said tries the first video recommended to me was of cats, and the longest scrolling I had to do to get a pet video recommended was 13 scrolls (commercials included).
Algorithms on TikTok divide videos into simpler categories: countries, ethnicities, neurodivergence, cats…. It is unable to differentiate between the genre of the content and makes the content in the video the genre itself, hence why ‘cats’ is the genre from the perspective of the algorithm.
Most people on cat distribution system videos are already cat owners or cat lovers in general, some went out of their way because they heard a cat sound, even the act of stopping when seeing a cat on the streets is of caring for said species. With that stated it can be assumed that cat lovers have a higher probability of being in such a situation and having the possibility to film it; by my history on the app, it can also be assumed that I am a cat lover. The algorithm could also make those same assumptions and recommend me a video that was registered as a “CAT” category on their system.
To conclude, disappointingly and against popular believes, the TikTok algorithm is not a genie nor an FBI spy, what happened is a coincidence of timing between my incident in real life fitting with the algorithm its inability to differentiate between genre of content, and at the same time its ability to figure out the personal preferences of their users.
REFERENCE:
- Pu P., Chen L., and Hu R. (2011) A user-centric evaluation framework for recommender systems. In Proceedings of the fifth ACM conference on Recommender systems (RecSys ’11). Association for Computing Machinery, New York, NY, USA, 157–164. Available from: https://doi.org/10.1145/2043932.2043962 (Accessed 22/10/2023)
- Kavoori, A., (2011). Reading YouTube: The Critical Viewers Guide. Available from: https://learningzone.dmu.ac.uk/ (Accessed 22/10/2023)
- Kavoori, A., (2015). Making sense of YouTube. Global Media Journal, 13(24), pp.1-25.
- EAC Course Guides (n.d.) Film and Media Studies. Available from: https://eac.libguides.com/c.php?g=723550 (Accessed 22/10/2023)
BISILA 