1138973

research-article2022
NMS0010.1177/14614448221138973new media & societySiles et al.

Article

new media & society

Learning to like TikTok . . .

1­–17
© The Author(s) 2022
Article reuse guidelines:
and not: Algorithm sagepub.com/journals-permissions
DOI: 10.1177/14614448221138973
https://doi.org/10.1177/14614448221138973
awareness as process journals.sagepub.com/home/nms

Ignacio Siles , Luciana Valerio-Alfaro

and Ariana Meléndez-Moran
University of Costa Rica, Costa Rica

Abstract
This article analyzes algorithm awareness as a process—a series of activities intended to
reach a goal over time. It examines how a group of Costa Ricans understood, felt about,
and related to TikTok and its algorithms as they began using the app for the first time.
Data come from diary entries completed by 43 participants about their use of TikTok over
a month and seven focus groups with these diarists. The article discusses five activities
through which users expressed developing forms of awareness of TikToks’ algorithms
and enacted various rhythms in the experience of the app: managing expectations about
what TikTok is and how it works; “training” the app; experiencing a sense of algorithmic
personalization; dealing with oscillations in the pertinence of recommendations; and
showing various forms of rejection of TikTok. The article then considers some implications
of bringing time to the fore in the study of algorithm awareness.

Keywords
Algorithms, awareness, diaries, folk theories, Latin America, personalization, process,
rhythm, temporality, TikTok

Introduction
Awareness of algorithms is key in how people understand, feel about, and relate to
platforms. Algorithm awareness also matters as it is tied to an increased capacity to
limit manipulation and exploitation through data extraction practices, protect human
autonomy and privacy, refine Internet skills, and develop critical thinking abilities

Corresponding author:
Ignacio Siles, School of Communication, University of Costa Rica, San José 11501-2060, Costa Rica.
Email: [email protected]
2 new media & society 00(0)

(Dogruel et al., 2021; Gruber et al., 2021; Oeldorf-Hirsch and Neubaum, 2021; Shin
et al., 2022).
Despite its manifold contributions, research on users’ relationship with algorithms has
often espoused a view of awareness as a relatively static possession that people either
have or not. However, as Dogruel et al. (2021) note, “algorithms are subject to evolve-
ment over time and their mutual shaping with user interactions make them difficult to
approach using predefined assumptions about factual knowledge” (p. 16). Focusing on
awareness as a process—a series of activities intended to reach a goal over time—would
thus make it possible to identify the stabilities and changes in how users think about, feel
about, and relate to algorithms. To implement this approach, in this article, we ask, how
do people develop an awareness of TikTok’s algorithms over time? What activities char-
acterize the process of establishing a relationship with the app and its algorithms? Data
come from diary entries completed over a month by 43 new TikTok users aged 18–
66 years old from different sociodemographic backgrounds and seven focus groups with
these participants.
Defined by its parent company as “the leading destination for short-form mobile
video [. . .] to inspire creativity and bring joy” (TikTok, 2020), TikTok offers a compel-
ling case for a study of algorithm awareness. The platform allows users to post videos of
up to 10 minutes. Users can then access these videos through their “For You” page,
described on TikTok’s website as an algorithm-powered “stream of videos curated to
your interests” that constitutes “the magic” of the app (TikTok, 2020). Although it is also
possible to “follow” specific profiles, TikTok emphasizes both materially and discur-
sively the centrality of algorithms in the experience of the app. There is a generalized
belief that, compared to other platforms, TikTok has the most “aggressive” and “addic-
tive” algorithms (Siles and Meléndez-Moran, 2021; Schellewald, 2021). In a succinct
explanation, TikTok indicates that its “recommender system” prioritizes the following
three main factors: user interaction metrics, metadata about the videos, and “device and
account settings” (TikTok, 2020). Kang and Lou (2022) elaborate,

TikTok applies natural language processing to identify textual and audio elements (e.g., sounds)
of the videos that users enjoyed, computer vision to classify the videos’ visual components, and
analysis of the hashtags and captions connected to such videos. TikTok’s algorithm is so
powerful and aggressive that it can learn the vulnerabilities and interests of a user in less than
40 minutes. (p. 4)

Temporality is thus the key in the experience of TikTok, from the incomparable speed
at which users feel its algorithms react to their actions, to the unique pace at which
people swipe content to signal their interests and preferences to these algorithms
(Boffone, 2022).
We begin by discussing recent scholarship on algorithm awareness and the tempo-
rality of social media, and the particularities of our research design. We then examine
five activities through which users expressed developing forms of awareness of
TikToks’ algorithms and enacted various rhythms in the experience of the app: man-
aging expectations about what TikTok is and how it works; “training” the app; expe-
riencing a sense of algorithmic personalization; dealing with oscillations in the
Siles et al. 3

pertinence of recommendations; and showing various forms of rejection of TikTok

and its algorithms. We conclude by considering some implications of bringing time to
the fore in the study of algorithm awareness.

A process approach to algorithm awareness

Following Kruger and Kearney (2006), we define awareness as a tridimensional concept
that comprises “what [.  .  .] a person know[s] (knowledge); how [.  .  .] they feel about the
topic (attitude); and what [.  .  .] they do (behaviour)” (p. 291). Our approach to algorithm
awareness builds on research that has accounted for these three dimensions.
Some studies have emphasized cognitive issues involved in the awareness of algo-
rithms. These studies focus on what people know about algorithms when they use plat-
forms (Cotter and Reisdorf, 2020) or how they “understand certain aspects of algorithms”
(Hargittai et al., 2020: 761, emphasis added). Early studies identified the extent to which
users knew that algorithms shaped their interactions with various platforms and con-
cluded that people were little aware of algorithms (Eslami et al., 2015). More recently,
research has shown that this knowledge is not evenly distributed: age, gender, and educa-
tion are the most significant factors explaining differences in people’s understanding of
algorithms (Espinoza-Rojas et al., 2022; Gran et al., 2021). Awareness of algorithms
comes from different sources of information, related to both artifacts themselves (“endog-
enous”) and to “knowledge in the world” (“exogenous”; DeVito et al., 2018: 3).
Researchers have found evidence of this awareness in the “folk theories,” “imaginaries,”
and “stories” that users put forth to account for their relationships with algorithms (Siles
et al., 2020; Schellewald, 2022; Ytre-Arne and Moe, 2021).
In addition to cognition, scholars have also examined the affective dimension of algo-
rithm awareness. In this body of work, algorithms are not only known or understood but
also felt and experienced. Emotional responses to algorithms are considered a key heu-
ristic device to understand how users “perceive, feel, and behave around” (Swart, 2021:
2) them. Scholars have thus considered feelings of “surprise” (Swart, 2021), “apprecia-
tion” (Oeldorf-Hirsch and Neubaum, 2021), “irritation” (Ytre-Arne and Moe, 2021), or
“aversion” (Yeomans et al., 2019) as indicative of algorithm awareness.
A third strand of scholarship has emphasized how awareness of algorithms is tied to
the practices of users. A key insight of these studies is that awareness is formed primarily
through the use of technologies (DeVito et al., 2018). Thus, Eslami et al. (2015) showed
that the more users engaged with the features of platforms, the more aware they became
of algorithmic operations. Work on content creators has arrived at a similar conclusion.
Cotter and Reisdorf (2020) referred to algorithms as “experience technologies” (p. 745),
things that people mostly understand through use. For Cotter (2022), “practical knowl-
edge” of algorithms is “knowledge expressed intuitively in action” (p. 16). This knowl-
edge is not necessarily oriented only toward cognitive interpretations of algorithms but
often focuses instead on establishing social relationships with others. Users thus mani-
fest this form of awareness through specific practices, tactics, competencies, and skills
(Siles et al., 2019; Gruber et al., 2021; Hargittai et al., 2020).
To further theorize algorithm awareness as a process, we turn to scholarship on the
temporality of social media experience. A growing number of studies have examined
4 new media & society 00(0)

how the technicity of platforms (primarily algorithms) is key in the production of

such temporal regimes as “real-time” (“the promise of an experience of the now”
[Weltevrede et al., 2014: 126]) and “right-time” (the perception that the right content
comes at the right time [Bucher, 2020]). Scholars have identified the temporal pat-
terns that shape the experience of these time regimes. Weltevrede et al. (2014) thus
argued that each platform is endowed with its own “pace,” which enables various
forms of “real-timeness.” Similarly, Lupinacci (2022) analyzed the “rhythms” of
algorithms (or “algorhythms”) that orchestrate a sense of time through such mecha-
nisms as instantaneity, freshness, simultaneity, and ephemerality. For Carmi (2020),
algorithmic rhythms matter as they lead to forms of user sociality that benefit tech
companies’ business models. Our study builds on this body of work to further under-
stand how various rhythms in the experience of TikTok are tied to algorithm aware-
ness as a temporal process that incorporates cognitive, affective, and practical
dimensions.

Research design
The use of TikTok has grown exponentially in Latin America over the past 2 years. Some
sources estimate it is the third most used algorithmic platform in the region, after
Facebook and Instagram (LABS, 2022). More specifically, Costa Rica offers various
characteristics that make it an ideal case for conducting this study. The country is a
regional leader in the use of various apps and is one of the most digitized countries con-
sidering infrastructure, culture, and use. This condition creates fertile grounds for such
algorithmic platforms as TikTok (Siles, 2023).
We began our study by sampling individuals who responded to various criteria. We
looked for participants of different age groups (rather than relatively young informants
only) and people with various degrees of education (rather than college students or par-
ticipants with college degrees only). In addition, we focused on individuals who had
never used TikTok under the premise that their unfamiliarity with the app made it easier
to recognize the development of algorithm awareness.
We circulated a call for participation on the social media profiles of the university
where the research was conducted. Interested participants filled out an online question-
naire that inquired into their previous experiences with TikTok, their use of social media
platforms, and background sociodemographic information. We selected 43 individuals
for the study: 12 participants were aged 18–29 years; 21 were aged 30–39 years; five
were aged 40–49 years; and five were aged 50–66 years. The average age of participants
was 34 years old. In total, 29 participants identified themselves as women and 14 as men.
The sample of participants included people without formal education, college students,
and professionals. Participants were familiar with the use of other algorithmic platforms.
All of them indicated they regularly used Facebook, Instagram, and YouTube. Most said
they also used Netflix and, to a lesser extent, Spotify and Pinterest.
We then asked selected participants to share diary reports of TikTok use. In keeping
with previous studies (Moe and Ytre-Arne, 2021; Risi et al., 2020), we asked participants
to share daily reports during the first 2 weeks of the study and invited them to submit
weekly reports instead over the following 2 weeks. Of 43 participants, nine stopped
Siles et al. 5

sending their reports at some point. When this occurred, we inquired into their reasons
for abandoning the study and thus turned the non-use of the app into a finding in itself.
The remaining 34 participants submitted a total of 275 diary reports. Data collection of
diary entries took place between February and March 2022. We offered participants two
options to submit their responses: an online form where they could share written obser-
vations and WhatsApp chats where they could submit audio recordings. We transcribed
the audio messages we received.
We triangulated our research methods to minimize “the inevitable artificiality and
performative dimension of the diary process” (Markham and Couldry, 2007: 675). Thus,
we conducted seven focus groups with participants: four before they submitted their
diary reports and three afterward. In the initial focus groups, we explained to participants
our research design and clarified questions regarding the submission of the reports. We
also asked participants to discuss their motivations to take part in the study and to
describe their perceptions about TikTok before they began using the app. These conver-
sations took place in February 2022, lasted for an average of 35 minutes, and were con-
ducted on Zoom (because of limitations to carrying out in-person meetings in the context
of COVID-19). The focus groups that took place after the submission of diary reports
served primarily to triangulate our data and add nuance and context to our findings.
These conversations, also conducted on Zoom, took place in March 2022 and lasted for
an average of 52 minutes.
All the data were collected in Spanish (translations are our own). To analyze the
data from the questionnaire, diaries, and focus groups, we employed an abductive
approach that “rested on the cultivation of anomalous and surprising empirical find-
ings against a background of multiple existing [. . .] theories and through systematic
methodological analysis” (Timmermans and Tavory, 2012: 169). During an initial
round of open coding, we identified the various expressions of awareness manifested
by participants, and compared them to similar findings in previous studies. We then
examined these forms of awareness through axial coding, incorporating a process
dimension designed to make visible both stabilities and changes in participants’
awareness over time and to further identify the novelty of our findings. Finally, during
a round of selective coding, we developed five categories (which we conceptualized
as activities) to capture the shifting patterns of cognitive, affective, and practical
awareness we identified in previous rounds of coding. During this stage, we also
accounted for the rhythms enacted in participants’ relationships with TikTok in light
of previous theoretical categories.

Algorithm awareness on TikTok

In what follows, we examine five activities through which users expressed specific forms
of awareness of TikToks’ algorithms: expectations, “training,” personalization, oscilla-
tions, and rejections. These are fluid activities that can be variously enacted rather than
strictly sequential stages. We discuss the trajectories of the following two archetypal
participants: a 55-year-old saleswoman who lives in Costa Rica’s capital (Larissa) and a
23-year-old university student from a rural zone (Federico). We add nuance to the discus-
sion of these two cases with examples from other diarists.
6 new media & society 00(0)

Expectations
Awareness does not emerge in a vacuum but is rather a product of broader contexts. We
explored these contexts by analyzing the expectations that participants had about TikTok
before they started using it. These expectations provided the background against which
participants sought to achieve the goal of establishing a relationship with TikTok.
A sense of disinterest in TikTok prevailed in people’s responses to our preliminary
online questionnaire. Given its origins (as Musical.ly) and how it originally targeted
adolescents, many participants envisioned TikTok as an app made for young people, a
group of users they associated with notions of shallowness and irreflection. When asked
to define what TikTok meant for them when they signed up for the study, it was common
for participants to express such ideas as, “[It is] a platform for young people” or “for
kids,” or a “juvenile” app. In his questionnaire response, Federico also emphasized the
format of the content: “It’s an app for short [decontextualized] videos.” The combination
of young users and viral videos turned TikTok into what one participant defined as “the
incoherence of sanity and the debauchery of stupidity.” In this view, TikTok was an app
for people who were unable to understand for themselves the wider implications of the
content they watched, and who were mostly interested in vain, futile, and frivolous
things. These ideas reveal how depersonalized the relationship with TikTok was for par-
ticipants before they started using the app: TikTok was for others (specifically a vulner-
able group of people) but not for them.
This indifference toward TikTok can be conceptualized as desynchronized tempo-
ralities (Jordheim and Ytreberg, 2021). For our interlocutors, there did not seem to be
a place for TikTok’s fast pace in the rhythm of their own daily life. As some people
expressed in the preliminary questionnaire, TikTok seemed a “very time-consuming”
app with “accelerated content.” A few even mentioned the word “addictive,” thus sug-
gesting that TikTok prevented people from freely managing their time. Larissa thus
acknowledged in the questionnaire: “I have never used [TikTok], I think I would not
stop watching [videos].” Against the background of a cultural mandate to always be
productive and the futility attributed to the app, many participants did not hesitate to
call TikTok “a waste of time.”
Participants nuanced these accounts during the focus groups carried out before they
began completing the diaries. During these conversations, many admitted being curious
about TikTok, which most considered “fashionable” at the time of the study. For some,
this created pressure to learn more about the app. Most participants narrated instances
where they had seen close people in their social circles using TikTok (children, col-
leagues, friends), which had sparked some interest in it. The experiences of these early
adopters revealed to participants that the app could contain content of potential value to
them. Learning to communicate with other users through the app’s technological and
cultural codes was an important motivation for some participants. That was the case of
Sonia, a 66-year-old translator who said she wished to better understand her teenage
grandchildren.
On several occasions during focus groups, participants framed our study as a legiti-
mate opportunity for them to use an app they were curious about but had mostly discred-
ited. “I was looking for an excuse to use TikTok!” said Ricardo, a 63-year-old woodworker.
Siles et al. 7

Participants defined this “excuse” in utilitarian terms: TikTok could provide them with
valuable information for their professional activities. In exchange for their participation,
they hoped to obtain resources to fulfill a personal or professional purpose.
Except for one person, none of the participants mentioned the term “algorithm” in the
preliminary questionnaires or initial focus groups. This does not mean that participants
lacked awareness but rather suggests that algorithms were not necessarily central in
accounting for their perception of the app before they began using it. Although some
expressed ideas about how the app’s recommendations worked (so-called “folk theo-
ries”), they did not use the term “algorithm” to articulate those ideas nor centered their
explanations on the specific role of algorithms.

Training
Creating a rhythm between the desynchronized temporalities of TikTok’s time-consum-
ing pace and the productive daily life of diarists took work (cf. Jordheim and Ytreberg,
2021). This rhythm manifested primarily as “repetition” (Carmi, 2020): engaging in a set
of recurrent practices to produce a visible result. Participants understood this repetitive
work as (mutual) “training”: they had to strategically act to receive interesting recom-
mendations on TikTok, while TikTok had to work to adjust to their preferences. In a
focus group conversation, Larissa summarized the gist of the training activities: “I had to
learn how to make the algorithm work in my favor.” In this sense, training is comparable
to what other researchers have termed “domestication”: users “tame” algorithms as they
find value for them, develop routines around their use, and construct perceptions of the
world based on them (Siles, 2023; Simpson et al., 2022).
Most diarists indicated that the “For You” page was their default starting point when
they began using TikTok, thus situating algorithms at the core of their experience. Some
reported that their preferred strategy to begin training the app was to search for specific
profiles or keywords. Matilde (a 30-year-old saleswoman) captured the purpose of this
practice with precision in her first diary entry:

It seems to me that you can find good content on tik tok [sic] (IF YOU LOOK FOR IT) but I
get the impression that the algorithm is there to recommend content with the most views,
regardless if the content is good or bad. (Caps in original)

Matilde’s words (and explicit use of caps) reveal how diarists thought it was their obliga-
tion to engage in training activities to revert TikTok’s default tendency to recommend
popular videos (as opposed to personal content).
It was common for participants to identify patterns in the kinds of recommendations
they received in their initial interactions with TikTok. In his very first report, Federico
revealed his surprise when TikTok recommended a video made by someone in the same
place where he lives in the north of Costa Rica. Two days later, he reported how he
thought TikTok was showing him content based on the profiles of users he had previ-
ously watched, under the assumption that this revealed his preferences.
Finding ideal times and places to use the app was key in training TikTok’s algorithms.
Scholars in the domestication tradition have referred to this process as the “incorporation”
8 new media & society 00(0)

of algorithms into routines (Simpson et al., 2022). Violeta, a 45-year-old housewife,

explained her own routines during a focus group:

I used [TikTok] several times throughout the day. Maybe I was at work and wanted a mental
break, so I grabbed and checked TikTok, Instagram, and Facebook. At lunchtime, I always
checked it a couple of times. Before going to bed, I looked at [TikTok] a little bit.

These routines shaped awareness in different ways. On one hand, diarists used TikTok
for relatively short periods to pass the time while doing something else (a work break,
waiting in line at the bank). That was the case of Federico, who reported using TikTok
for periods of approximately 10 minutes at a time. This typically occurred in public
spaces, which made participants more attentive to content that could bother others around
them. On the other hand, participants devoted time to specifically using the app. The
most common example given by diarists was using the app “before going to bed.” Larissa
thus waited until the end of the day to use TikTok and indicated that she watched videos
for periods that often exceeded 1 hour. This routine often took place in private spaces,
which participants considered ideal to assess the app’s recommendations and identify the
types of content they liked.
Participants integrated TikTok into their daily social media routines and created fertile
grounds for comparisons across platforms. Diarists’ early understanding of TikTok
developed precisely as they identified similarities and differences with other apps that
they were familiar with, most notably Instagram. In her first report, Matilde noted, “My
first impression is that [TikTok is] about videos to entertain; the more ridiculous they are,
the more attention they attract and the more followers they get. It was very similar to the
Reels I see on Instagram.” Since she had used Instagram prior to encountering TikTok,
Matilde assumed the content she found on Reels was more characteristic of Instagram
than any other app.
Training activities were associated with specific emotions in participants. Federico
wrote in his sixth entry that the uncertainty of training activities made him feel “inse-
cure.” Many diarists found this activity frustrating because it required constantly invest-
ing time and effort without perceiving immediate results. Participants also used the
notions of “boredom” to describe this activity and “disinterest” to recount its results (cf.
Lupinacci, 2022). Many diarists also reacted emotionally against the persistence of what
they considered biases in their “For You” page. Recalling this process, Florencia
(41-year-old teacher) said during a focus group: “At first, it was horrible. There were
very sexualized videos, which made me feel uncomfortable. I didn’t like it.”
Compared to initial questionnaires and early focus groups, many more participants
mentioned the term “algorithm” in the diary reports to describe training activities. The
emergence of the word “algorithm” (often in the singular) in participants’ vocabulary
added some precision to what had been a generic description of TikTok. “The algorithm”
became an umbrella term to refer to what participants thought were TikTok’s specific
computational procedures to recommend content. It was also common for diarists to use
“TikTok” and “algorithm” almost interchangeably, consistent with a view of TikTok as
an assemblage, an inseparable tissue of relationships between app, algorithms, and users
(Siles and Meléndez-Moran, 2021).
Siles et al. 9

Some of the ideas that diarists developed about algorithms came from the active use
of TikTok. In their reports, diarists described algorithms as computational mirrors of
their actions. They interpreted algorithms as reactive entities that responded to their
training practices in a precise manner. In her second diary report, Sonia noted, “As I
interact with the application, it adjusts to my tastes. [. . .] It has one formula to detect
trends and prioritizes this content, in addition to taking into account user interactions.” In
this account, it is the users who lead the process by shaping the behavior of algorithms
through their actions.
Awareness of algorithms during training activities also led to gradually personal-
izing the relationship with the TikTok assemblage. To describe their interactions with
TikTok, diarists employed terms that suggest a change in their rapport with the app
and its algorithms, compared to their initial expectations. After her third report,
Larissa began employing a vocabulary that suggested she was engaging in a dialogue
with TikTok through her training practices. She interpreted each recommendation as
a response to her interventions as part of this dialogue (enacted by such practices as
liking certain content, spending more time watching videos, or avoiding engaging
with specific content). Although diarists assigned algorithms a more reactive role,
they also emphasized how capable and precise they thought these were in responding
to training.
Participants also integrated “exogenous” sources of knowledge into their awareness
of TikTok’s algorithms. Some knowledge came from comparisons between TikTok and
other apps (cf. Espinoza-Rojas et al., 2022). Larissa said during a focus group that her
understanding of algorithms came from her previous experiences with Facebook. Diarists
also drew on articles in media outlets or cultural products to interpret the role of TikTok’s
algorithms. Agustina, a 51-year-old journalist, narrated how various Netflix documenta-
ries had shaped her understanding of algorithms:

What I learned is that apps track your behavior and compare it statistically with all the other
users. [The goal] is to attract [people’s] attention, to make them use the apps again, to make
room for more advertising, to make them watch other things. Personally, I don’t demonize
[algorithms]. Sometimes they’re useful.

Agustina’s premise was that the algorithms of different platforms (including TikTok)
perform the same function: to manipulate users. But she also relativized the importance
of this manipulation by suggesting she remained in control of algorithms and by high-
lighting the services they could provide to her.

Personalization
After a few reports that focused mostly on describing training activities, diaries evinced
a relatively different understanding of algorithms. Diarists began noting that training
seemed to have finally worked. For Larissa, this occurred on the third day of using
TikTok. She expressed this with enthusiasm in her diary entry: “This time I liked [TikTok]
better because the algorithm is already giving what I like.” Federico reported a similar
experience also on the third day of use: “I accepted it a little bit more. I got carried away.”
10 new media & society 00(0)

Four days later, he went further: “I [arrived to] moments when I laughed inside [and] the
joy overflew without thinking about it.”
Lupinacci’s concept of “harmony” aptly describes the “algorhythm” enacted in this
activity. For Lupinacci (2022), a harmonious rhythm takes place when “the content and
the people shown to you first are those attuned to your individual preferences and past
engagement” (p. 9). Harmony also manifests in the sensation that algorithms produce
“right-time” recommendations in ways that seem anticipatory (Bucher, 2020).
Whereas the notion of training situated people as those leading the relationship with
TikTok, participants now felt they could “sit back” and enjoy the rhythm of harmonious
algorithmic personalization. Roles and responsibilities shifted during this activity: algo-
rithms offered them content and participants reacted to these recommendations. Florencia
explained during a focus group: “The content kept changing until I suddenly realized that I
had spent an hour watching videos.” In the accounts of Florencia and Federico (cited above),
it was algorithms that made them lose control of time while they focused entirely on the app.
As Siles (2023) has noted, it would be misleading to describe changes in users’ under-
standing of their agency as a move to passive roles. Instead, Siles (2023) refers to this
state as “active passivity,” which Gomart and Hennion (1999) define as “a movement in
which loss of control is accepted and prepared for [. . .] the abandonment of forces to
objects and the suspension of the self” (p. 227). Reaching this state is thus experienced
as an accomplishment. Personalization is about letting algorithms act on users after users
have acted on algorithms (Siles and Meléndez-Moran, 2021).
Both in their reports and in posterior focus groups, many participants highlighted the role
of algorithms (using explicitly the term) in explaining how recommendations had improved
in such visible ways. In Violeta’s words, “The algorithm might have read my profile a little
better and what I’ve liked. It [also] read that I didn’t like certain content.” This comment
shows again the precision that diarists attributed to TikTok’s computational capacities.
TikTok confirmed expectations derived from their experiences both with this app and other
platforms, and ideas in public discourse about the operation of algorithms.
Many participants expressed the sense that they had achieved personalization rela-
tively fast. As with Larissa’s and Federico’s, most diary reports included statements of
this nature after the third or fourth day of use. Florencia noted in her second report that
TikTok was still in “default mode.” The following day, she wrote instead, “The algorithm
is starting to work. I don’t hate [TikTok] as much.” In her sixth report, she added, “The
algorithm is getting to know me 😏😅” (emphasis added). In this perspective, TikTok was
not a depersonalized entity with which they had no relationship any longer but a personal
platform that had something specifically for “me.” The emoji Florencia used conveyed
how she felt obligated to acknowledge that the app’s algorithms seemed to have worked
and the pleasure she derived from this result. Participants thus felt interpellated by
TikTok’s algorithmic recommendations. In this context, interpellation refers to “the work
embedded in algorithms to convince [people] that [platforms] are speaking directly to
them, ‘hailing’ them in particular ways” (Siles, 2023: 35). As Cohn (2019) notes, algo-
rithmic interpellation functions when users begin thinking in the terms offered through
the interface of a platform.
Diarists showed much more positive emotions to describe their awareness of algo-
rithmic personalization compared to other moments. Despite her initial contempt for
Siles et al. 11

the app, Florencia began feeling “more comfortable” with the content she received
after the third day of using TikTok. Similarly, Mónica, a 38-year-old babysitter, wrote
in her third diary entry: “I don’t find TikTok so boring anymore!” and mentioned she
felt “happiness!” when using it. For diarists, this change in their affective state justified
a sustained use of the app: “I did like it today. I will keep using it,” wrote Clara in her
second diary entry.
Participants described how they engaged in new use practices as a reaction to relevant
recommendations they received once they had reached personalization. A telling exam-
ple is how diarists began sharing and exchanging videos with others. In her sixth diary
entry, Larissa recounted, “[Today] I got things that I was able to share. It was very enter-
taining!” For Larissa, sharing content with others for the first time was only possible
because TikTok recommended her videos that were worth sharing. She thus understood
her own practice as a fitting response to the quality of algorithmic recommendations she
had begun to receive. Becoming aware of changes in algorithmic operations thus moti-
vated her to act differently compared to previous days.

Oscillations
Some participants noted that they permanently enjoyed such a state of harmonious algo-
rithmic personalization. In their reports, these diarists indicated they continuously
received video suggestions that met their interests. Alvaro, a 23-year-old college student,
used the words “joy” and “laughter” repeatedly in his reports until the end of the study.
But, for most diarists, recommendations oscillated between videos they liked and others
they did not, even after having reached personalization. Larissa thus noted in her seventh
diary entry, only a few days after her enthusiastic assessment of TikTok’s algorithms
reported in the previous section: “It’s very ‘choppy’ (revuelto), [there is] a lot of every-
thing.” She elaborated on her report the following day: “Whereas other days I received
very good suggestions and I took the time to watch them, I didn’t like yesterday’s sug-
gestions very much. I don’t know why!”
The “algorhythm” enacted in this activity could be described as “oversaturation,” that
is, the sensation that people had to deal with too many and very mixed recommendations
at an exceedingly fast pace. In other words, some participants felt the pace of TikTok’s
algorithms surpassed the rhythm of their lives. How participants experienced this rhythm
is nowhere clearer than in the words of Violeta, who wrote, “Too much movement and
everything is very fast. It is too dynamic for my taste.”
Becoming aware of changes in algorithms led our interlocutors to express another
fundamental expectation: stability. Most diarists assumed they would continue to receive
recommendations they found germane once they felt personalization had been achieved.
When this expectation was not met, they manifested feelings of surprise and confusion.
In response to oscillation in her recommendations, Violeta wrote in her 10th report: “I
feel there is a setback.” Unstable recommendations thus went against the premise that
algorithms should be stable and precise rather than variable and prone to errors that users
felt had been overcome.
This belief in algorithms as entities that always perform the same tasks in an unalter-
able manner confirms studies that have shown how users increase their trust in
12 new media & society 00(0)

algorithmic systems when they feel these systems do not “violate” their expectations
(Kizilcec, 2016). Participants did not feel the need to further decipher how TikTok’s
algorithms worked unless one of their expectations was not met. Whereas diarists elabo-
rated on how they thought that algorithms worked during training and personalization
activities, they reduced their observations on this issue after various days of sharing their
reports. Once they thought they had figured out how to reach personalization, partici-
pants did not expect algorithms to change and thus did not find it always necessary to
keep analyzing why algorithms were offering different recommendations. Thus, it was
common for diarists to use such expressions as “nothing new to report” or “just like what
I said yesterday” in their accounts of why they thought they were receiving certain rec-
ommendations on TikTok.
Unlike with training, in which participants thought of algorithms as entities that mir-
rored their actions with precision and immediacy, during oscillation, they interpreted
algorithms as responding primarily to “external” forces, such as viral trends or contex-
tual issues. Agustina explained how she thought that TikTok was integrating more viral
trends into her recommendations compared to the beginning:

I have the feeling that, if there is content that is going viral and is not part of my list of interests,
it must be based on information from many users who are watching it or who want to know
more about it. Or because of advertising issues. Those were two things that made me want to
be a little more distant [from TikTok].

Agustina’s “distance” toward TikTok came from a premise that, once a “list of interests”
had been defined, algorithms should not depart from it.
In a similar manner, diarists noted that, compared to the early days of the study, algo-
rithms were recommending more videos to them about specific events (such as the war
in Ukraine or the Costa Rican presidential runoff election.) These recommendations
made participants feel connected to others in the simultaneous experiences of “media
events” that took place in the hybrid formats of “real-time” (including television and
social media; Sumiala et al., 2018). Noticing how these events unfolded on TikTok also
shaped the awareness of diarists. During a focus group, Agustina accounted for the
repeated presence of videos about one presidential candidate on her “For You” page by
noting it could only be because of sponsored posts or because “the algorithm” was rec-
ommending what other people “with the same profile” were interested in.
For several participants, the expectation derived from assessing one video after the
next made the app worthwhile and the experience of using it exhilarating, even if uncom-
fortable. “I feel this is vice-inducing,” wrote Larissa. These participants developed, as it
were, a feel for the rhythm of “oversaturation” on TikTok.
Diarists offered two rationales for continuing to use TikTok despite algorithmic oscil-
lation. The first rationale emphasized utilitarian reasons. As noted earlier, previous
expectations provided the background against which participants assessed the merits of
algorithms. Accordingly, some participants argued they valued TikTok because it pro-
vided them with information that was useful in their personal or professional lives. In her
seventh report, Larissa marveled at the number of ideas she had received by watching
videos on TikTok: “If you want to start a business, there is a lot out there!” She
Siles et al. 13

mentioned having watched insightful tips about how to create soaps, dog biscuits, and
personalized gifts. This statement drastically differed from claims made by participants
before using TikTok. Rather than a frivolous app made for gullible others, TikTok was
now a tool for “finding information and useful content about things to do,” specially
made for each one of them, as Violeta put it.
In addition to utility, participants emphasized a logic of complementarity. In this per-
spective, TikTok was unique compared to other platforms, different enough to secure a
place in the social media ecologies of participants. That was Federico’s case, who indi-
cated in his ninth diary report that he was already “using TikTok more than Facebook and
Instagram” because he thought it was “better to watch a video in which the content has
not been altered so much.” By the eighth day of using TikTok, Larissa arrived at a similar
conclusion: TikTok offered her videos she noted, “could not find on Facebook” and won-
dered whether this difference could be explained by a difference in the quality of algo-
rithms: “Maybe Facebook’s algorithm isn’t as good as TikTok’s.”

Rejections
Thus far, we have discussed how the use of TikTok led to various forms of algorithm
awareness over time. But awareness was also tied to the non-use of the app. Throughout
the study, some diarists manifested their desire to stop using TikTok. The most common
reason they advocated was the frustration derived from training activities. For these par-
ticipants, the rhythm of repetition had become tiresome. Karppi et al. (2021) aptly
describe how awareness is tied to the search for disconnection as it leads to “an awaken-
ing of a counter-mood with a desire to act and to knowledge of what to do” (p. 1604).
That was precisely Federico’s case. In his final entry, exactly 1 month after he began
using TikTok, he wrote, “There hasn’t been any new content for days, nothing that
catches my attention.” This led him to a concrete desire for disconnection as a temporal
rhythm to relate to the app: “I have voluntarily used [TikTok] less,” he indicated in the
same report. Other participants noted they also had become “bored” of the app because
it kept recommending videos that were not of interest to them. Nothing they had done
seemed to have changed this pattern.
Other diarists who considered stopping using the app after the end of the study focused
on “oversaturation” and the oscillation of content between what they deemed interesting
and uninteresting video recommendations. For these participants, TikTok’s “choppy”
rhythm and deviation from their expectation of algorithmic stability proved difficult to
surpass. Framing it as an affective matter, Sonia, the 66-year-old translator, argued in a
focus group discussion that she had felt gradually more distant from the app: “I kept getting
cold. It got to a point where it got so cold I didn’t feel like using [TikTok] anymore.” For
Sonia, rejecting TikTok thus became a means to reassert her agency in relation to the app.
One final reason for rejecting TikTok was provided by those who felt the app was not
offering them anything sufficiently different from what they could find on other plat-
forms. Ricardo, the woodworker, offered a detailed explanation of this during a focus
group discussion:

I used [TikTok] but it didn’t give me the same feeling that Instagram does. It seemed to me that
it recommended very meaningless things. I think that TikTok comes at a time in my life when
14 new media & society 00(0)

I am already in love with Instagram and it didn’t make me break my relationship with Instagram,
an app that I’ve had since it started, that I’ve always liked, and where I think I can find more
friends with things in common.

After evaluating TikTok’s algorithms for a month, participants like Ricardo found
them wanting. Time was key in their explanations: TikTok came at a moment when their
affective ties to other platforms had already been formed. Ricardo’s words are a reminder
that algorithm awareness takes place as part of larger ecologies where platforms acquire
meaning and identity over time (Espinoza-Rojas et al., 2022).

The temporalities of algorithm awareness

In this article, we have argued for considering algorithm awareness not as a static posses-
sion that users either have or not but rather as a process that is variously manifested at
different points in time. We discussed five activities through which users expressed shift-
ing forms of cognitive, affective, and practical awareness of TikTok’s algorithms, and
enacted specific forms of temporal rhythms in their relationship with the app.
Many diarists began with sensations of apathy toward an app whose pace they felt
was incompatible with the rhythm of their daily life. Yet, most felt interpellated by
TikTok’s algorithms early on in the process and experienced a certain degree of harmoni-
ous algorithmic personalization after specific training practices. Participants were con-
fused and surprised when their expectation of stable algorithmic personalization was not
permanently met. How diarists conceived of their agency and that of algorithms shifted
throughout these activities. Thus, examining participants’ perceptions of algorithms
solely during one specific activity would have run the risk of assuming that certain mani-
festations of awareness were permanent rather than variable. It could have also limited
the possibility to identify changes in users’ cognitive, affective, and practical relation-
ships with algorithms.
Considering awareness as a process also helped to nuance the links between the active
use of platforms and user experiences of algorithms. We showed that the active use of
TikTok led diarists to naturalize the role of algorithms after some time and therefore to
think less about them (rather than the opposite) compared to the beginning of the study.
TikTok and its algorithms also came to mean something different to participants depend-
ing on whether they discovered them before or after using other apps (such as Instagram).
Given that our analysis drew on 43 participants and used a method that implies a per-
formative dimension, future work could help to diversify the sociodemographic charac-
teristics of participants by developing scales in surveys and experiments that account for
algorithm awareness as a process.
Research to date has emphasized how active use of platforms leads to more awareness
of algorithms. As a supplement, we showed that awareness of algorithms is also tied to
the non-use of platforms. Several participants disconnected from TikTok because they
found both the app and its algorithms wanting compared to other platforms they had
discovered previously in their lives or because they felt the app’s pace surpassed the
rhythm of their own daily life. Scholars have rightfully noted that awareness takes place
within larger digital ecologies. We demonstrated that the identity of platforms and the
relationships across them also shift over time, thus affecting awareness.
Siles et al. 15

Finally, our study helped broaden the understanding of the temporality of social media
platforms in two main ways. First, we demonstrated how a specific platform can orches-
trate multiple temporal rhythms. Most scholars have tended to assign one main kind of
pace to each specific platform (Weltevrede et al., 2014). As a supplement, we showed
how our interlocutors experienced various rhythms (desynchronicity, repetition, har-
mony, oversaturation, disconnection) in their relationship with the same app.
“Algorhythms” should thus be considered as temporal enactments in themselves rather
than a fixture of digital devices or given time regimes.
Second, we revealed how the experience of these rhythms was also tied to various
forms of awareness. Our analysis thus helped expand preliminary findings in the litera-
ture. In an examination of the “algorhythms” of media ecologies in London, Lupinacci
(2022) showed how people tended to be more aware of ephemerous content because of
the risk of not being able to watch it again. Our study provided further empirical evi-
dence of this process by examining how the experience of algorithmic rhythms shaped
forms of cognitive, affective, and practical awareness. Diarists developed different forms
of algorithm awareness based on whether they felt their relationship with them was
desynchronyzed, repetitive, harmonious, overly saturated, or if they considered they
needed to disconnect from the app.
Studying the experiences of people during 1 month made it possible to recognize very
specific activities that might have been difficult to identify otherwise. But it also might
have prevented us from examining a broader range of factors shaping awareness. Future
research could identify people’s awareness of algorithms over more extended periods.
These studies could also clarify whether there are temporal patterns (sequential, concur-
rent, and cyclical) in which people experience these five activities and others.

Acknowledgements
The authors wish to thank Kelley Cotter, Andreas Schellewald, and Brita Ytre-Arne for their inval-
uable suggestions on earlier drafts of this work. They also thank the anonymous reviewers for their
excellent comments about previous versions of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/
or publication of this article: This work was supported by the Vicerrectoría de Investigación,
Universidad de Costa Rica, Project no. C0451.

ORCID iD
Ignacio Siles https://orcid.org/0000-0002-9725-8694

References
Siles I, Espinoza-Rojas J, Naranjo A, et al. (2019) The mutual domestication of users and algo-
rithmic recommendations on Netflix. Communication, Culture & Critique 12(4): 499–518.
Siles I, Segura-Castillo A, Solís-Quesada R. et al. (2020) Folk theories of algorithmic recommenda-
tions on Spotify: Enacting data assemblages in the global south. Big Data & Society 7(1): 1–15.
Siles I and Meléndez-Moran A (2021) “The most aggressive of algorithms”: User awareness
of and attachment to TikTok’s content personalization. Paper presented at the 71st Annual
Conference of the International Communication Association.
16 new media & society 00(0)

Siles I (2023) Living with algorithms: Agency and user culture in Costa Rica. Cambridge, MA:
MIT Press.
Boffone T (ed.) (2022) TikTok Cultures in the United States. New York: Routledge.
Bucher T (2020) The right-time web: theorizing the kairologic of algorithmic media. New Media
& Society 22(9): 1699–1714.
Carmi E (2020) Rhythmedia: a study of Facebook immune system. Theory, Culture & Society
37(5): 119–138.
Cohn J (2019) The Burden of Choice: Recommendations, Subversion and Algorithmic Culture.
New Brunswick, NJ: Rutgers University Press.
Cotter K (2022) Practical knowledge of algorithms: the case of BreadTube. New Media & Society
0(0): 1–20. DOI: 10.1177/14614448221081802.
Cotter K and Reisdorf BC (2020) Algorithmic knowledge gaps: a new dimension of (digital) ine-
quality. International Journal of Communication 14: 745–765.
DeVito MA, Birnholtz J, Hancock JT, et al. (2018) How people form folk theories of social media
feeds and what it means for how we study self-presentation. In: Proceedings of the 2018 CHI
conference on human factors in computing systems, Montreal, QC, Canada, 21–26 April, pp.
1–12. New York: ACM.
Dogruel L, Masur P and Joeckel S (2021) Development and validation of an algorithm literacy
scale for Internet users. Communication Methods and Measures 16(2): 115–133.
Eslami M, Rickman A, Vaccaro K, et al. (2015) “I always assumed that I wasn’t really that close
to [her]”: reasoning about invisible algorithms in the news feed. In: 33rd annual SIGCHI
conference on human factors in computing systems, Seoul, Republic of Korea, 18–23 April,
pp. 153–162. New York: ACM.
Espinoza-Rojas J, Siles I and Castelain T (2022) How using various platforms shapes awareness
of algorithms. Behaviour & Information Technology. Epub ahead of print 17 May. DOI:
10.1080/0144929X.2022.2078224.
Gomart E and Hennion A (1999) A sociology of attachment: music amateurs, drug users. The
Sociological Review 47(1): 220–247.
Gran AB, Booth P and Bucher T (2021) To be or not to be algorithm aware: a question of a new
digital divide? Information, Communication & Society 24(12): 1779–1796.
Gruber J, Hargittai E, Karaoglu G, et al. (2021) Algorithm awareness as an important internet skill:
the case of voice assistants. International Journal of Communication 15: 1770–1788.
Hargittai E, Gruber J, Djukaric T, et al. (2020) Black box measures? How to study people’s algo-
rithm skills. Information, Communication & Society 23(5): 764–775.
Jordheim H and Ytreberg E (2021) After supersynchronisation: how media synchronise the social.
Time & Society 30(3): 402–422.
Kang H and Lou C (2022) AI agency vs. human agency: understanding human—AI interac-
tions on TikTok and their implications for user engagement. Journal of Computer-Mediated
Communication 27(5): 1–13.
Karppi T, Chia A and Jorge A (2021) In the mood for disconnection. Convergence 27(6): 1599–
1614.
Kizilcec RF (2016) How much information? Effects of transparency on trust in an algorithmic
interface. In: Proceedings of the 2016 CHI conference on human factors in computing sys-
tems, 2016, pp. 2390–2395.
Kruger HA and Kearney WD (2006) A prototype for assessing information security awareness.
Computers & Security 25(4): 289–296.
LABS (2022) TikTok soars in Latin America: app hits the 100 million user mark. Available at:
https://labsnews.com/en/news/technology/tiktok-soars-in-latin-america-app-hits-the-
100-million-user-mark/ (accessed 5 May 2022).
Siles et al. 17

Lupinacci L (2022) Phenomenal algorhythms: the sensorial orchestration of “real-time” in the

social media manifold. New Media & Society 0(0): 1–21.
Markham T and Couldry N (2007) Tracking the reflexivity of the (dis) engaged citizen: some
methodological reflections. Qualitative Inquiry 13(5): 675–695.
Moe H and Ytre-Arne B (2021) The democratic significance of everyday news use: using diaries
to understand public connection over time and beyond journalism. Digital Journalism 10(1):
43–61.
Oeldorf-Hirsch A and Neubaum G (2021) What do we know about algorithmic literacy? The
status quo and a research agenda for a growing field. Available at: https://doi.org/10.31235/
osf.io/2fd4j
Risi E, Bonini T and Pronzato R (2020) Algorithmic media in everyday life. An experience with
auto-ethnographic student diaries. Etnografia e ricerca qualitativa 13(3): 407–422.
Schellewald A (2021) On getting carried away by the TikTok algorithm. AoIR Selected
Papers of Internet Research. Available at: https://journals.uic.edu/ojs/index.php/spir/
article/view/12039
Schellewald A (2022) Theorizing “stories about algorithms” as a mechanism in the formation and
maintenance of algorithmic imaginaries. Social Media + Society 8(1): 1–20.
Shin D, Kee KF and Shin EY (2022) Algorithm awareness: why user awareness is critical for per-
sonal privacy in the adoption of algorithmic platforms? International Journal of Information
Management 65: 1–13.
Simpson E, Hamann A and Semaan B (2022) How to tame “your” algorithm: LGBTQ+ users’
domestication of TikTok. Proceedings of the ACM on Human-Computer Interaction
6(GROUP): 1–27.
Sumiala J, Valaskivi K, Tikka M, et al. (2018) Hybrid Media Events: The Charlie Hebdo Attacks
and the Global Circulation of Terrorist Violence. Bingley: Emerald Publishing.
Swart J (2021) Experiencing algorithms: how young people understand, feel about, and engage
with algorithmic news selection on social media. Social Media + Society 7(2): 1–11.
TikTok (2020) How TikTok recommends videos #ForYou. Available at: https://newsroom.tiktok.
com/en-us/how-tiktok-recommends-videos-for-you
Timmermans S and Tavory I (2012) Theory construction in qualitative research: from grounded
theory to abductive analysis. Sociological Theory 30(3): 167–186.
Weltevrede E, Helmond A and Gerlitz C (2014) The politics of real-time: a device perspective on
social media platforms and search engines. Theory, Culture & Society 31(6): 125–150.
Yeomans M, Shah A, Mullainathan S, et al. (2019) Making sense of recommendations. Journal of
Behavioral Decision Making 32(4): 403–414.
Ytre-Arne B and Moe H (2021) Folk theories of algorithms: understanding digital irritation.
Media, Culture & Society 43(5): 807–824.

Author biographies
Ignacio Siles is a professor of media and technology studies in the School of Communication and a
researcher in the Centro de Investigación en Comunicación (CICOM) at Universidad de Costa
Rica.
Luciana Valerio-Alfaro is a student in the School of Communication at Universidad de Costa Rica.
Ariana Meléndez-Moran is a graduate of the School of Communication at Universidad de Costa
Rica.