While habit formation has been extensively studied in domains such as health behavior, the unique characteristics of smartphones, including their permanent availability, may create distinct pathways for habit development. More specifically, social media apps may present a special case for smartphone habit formation due to their social nature and engagement-driven designs. In turn, scholars have theorized how users acquire social media habits (Anderson & Wood, 2020; Bayer et al., 2016, 2022). In particular, theories of social media habit formation emphasize the role of platform design features in triggering habitual use. A prominent example of engagement-driven design is algorithmic personalization. From the moment a user installs a social media app, their interactions “feed” the algorithm, resulting in an increasingly tailored and rewarding scrolling experience. TikTok’s recommendation algorithm is widely viewed as one of the most effective of its kind, with leaked internal documents revealing that TikTok’s own research showed that new users can form a habit in only 35 minutes of usage (Allyn et al., 2024).

However, surprisingly little research has empirically examined the initial stages of social media habit formation, and the role of algorithmic recommendations in this. The present controlled field experiment aims to address this gap by examining how new TikTok users develop habits in the first two weeks following app installation (N=105 students). Specifically, we compare two groups of new users: half of the sample used the default personalized TikTok feed, while the other half was instructed to disable personalization upon installing the app. We assessed habit formation processes through daily measures of perceived habit strength, perceived rewards, and objective usage (frequency; time spent). Linear Mixed Models in R were used for confirmatory analyses. The results of this study may provide initial evidence for the effectiveness of the TikTok algorithm in driving habit formation.

Research on traditional media has demonstrated that the surrounding media context plays a crucial role in shaping how advertisements are perceived and evaluated. However, especially in digital media, the impact of media context remains largely unexplored. This study examines how different emotionally charged social media posts affect the evaluation of a subsequent advertisement. Our experiment extends prior research, which has relied on a traditional valence-based comparison of media contexts, by utilizing an appraisal-based approach to provide more detailed insights into the effects of three distinct incidental emotions.

According to the appraisal-tendency framework, both happiness and anger are associated with high-certainty appraisals, leading to heuristic information processing. In this mode of processing, incidental emotions are more likely to be misattributed and incorporated into the evaluation of subsequent stimuli. Consequently, we assumed that happiness and anger would directly influence the evaluation of an advertisement. In contrast, sadness is associated with uncertainty and systematic processing, making it more likely to be excluded from the evaluation process and to result in a less emotion-driven judgment. Therefore, we predicted that, compared to a sadness-inducing media context, a happiness-inducing context would lead to a more positive advertisement evaluation, while an anger-inducing context would result in a more negative evaluation.

Based on a pilot study to validate the stimulus materials, we conducted an online experiment (N = 500) with three conditions. Participants were either presented a happiness-inducing, anger-inducing or sadness-inducing instagram feed. Each feed contained two emotional posts followed by one advertisement for a fictional brand. After swiping through the feed, participants reported their current emotional state as well as their level of certainty, product- and feed-involvement. Key dependent variables included attitude toward the advertisement, attitude toward the brand, electronic word-of-mouth and click intention.

Social media platforms featuring audiovisual content, such as TikTok, have gained relevance for news consumption (Reuters, 2024). The infotainment reporting style of news videos on these platforms may, however, foster a false feeling of being informed. This illusion of knowing (Ryffel & Wirth, 2020) can encourage avoiding further information-seeking or forming poorly justified opinions (Schäfer, 2020). In this study, we therefore investigate how common production features of TikTok videos influence subjective and objective knowledge about news issues and if they contribute to an illusion of knowing.

We focus on two production features that characterize TikTok news compared to traditional news (Wirz et al., 2023): thematic framing and emotionalization. Thematic framing is considered valuable because it broadens the context rather than focusing on specific events or examples (Gross, 2008). Emotionalization (Esser, 1999), on the one hand, is seen critically because it may distract from important aspects of an issue. On the other hand, however, it could also enhance memory about specific issues (Bas & Grabe, 2015). The combination of thematic framing and emotionalization is thus interesting; the focus is placed on the issue, but emotionalization may still attract attention.

A preregistered lab experiment (N = 198) was conducted in fall 2024 to 1) test how thematic framing and emotionalization influence subjective and objective knowledge, and 2) explore attention, emotional arousal, perceived complexity, processing fluency, and entertainment experiences as underlying mechanisms that may foster an illusion of knowing. In a 2×2×2 design, participants saw two TikTok videos on different topics (within-subject factor), which were manipulated along two dimensions (between-subject factors): thematic vs. episodic framing and high vs. low emotionalization. Supplementing self-reports, physiological measures for attention and arousal were collected. Until the conference, we will also have replicated the study with a larger online sample.

Social media algorithms shape users' experiences, influencing personal preferences and societal phenomena like public opinion (Pariser, 2011; Noble, 2018). As digital platforms become central to public discourse, they act as a barometer of public opinion, reflecting and amplifying societal trends (González-Bailón et al., 2015). Despite their influence, algorithmic literacy remains limited, reducing user agency and content diversity (Fletcher & Nielsen, 2018; Eg et al., 2023). Literature highlights the need for interventions that improve algorithm literacy, enabling users to navigate algorithmic systems and mitigate potential negative effects (Swart, 2021). While cognitive interventions raise awareness of algorithms, their impact on user behavior is underexplored (Swart, 2021). Additionally, little is known about how interventions shape user outcomes like agency, satisfaction, and folk theories of algorithms (Eg et al., 2023; Koenig, 2020). This study compares cognitive and behavioral interventions to improve algorithmic literacy and explore their effects on user agency, content satisfaction, social media as a barometer of opinion, and folk theories.

Research Questions

1. How do cognitive and behavioral interventions influence user agency, content satisfaction, social media as a barometer of opinion, and algorithmic literacy?

2. How does belief in different folk theories of algorithms change following cognitive and behavioral interventions?

Methodological Approach

This field experiment involves two groups: one receiving a cognitive-based intervention focusing on algorithmic knowledge, and the other a behavioral-based intervention using diary reflections on algorithmic influence. Data is collected at four measurement points, assessing algorithmic literacy, agency, content satisfaction, and folk theories. Participants, recruited from a university campus and personal networks, will complete surveys at each point. Statistical analyses will examine the effects of the interventions using regression and mediation models to explore relationships between algorithmic literacy, agency, satisfaction, and folk theories (Eg et al., 2023; Zarouali et al., 2021).