Understanding User Beliefs About Algorithmic Curation in the Facebook News Feed

“More recently, scholars have been concerned with a new dimension of digital inequalities – whether people are aware of algorithms and what they know about them (Eslami et al., 2015, 2017; Hamilton et al., 2014; Rader & Gray, 2015).”

“There is an increasingly rich body of research on how users imagine the algorithm or understand themselves and their feeds in relation to it (Bucher, 2017; Rader & Gray, 2015).”

“This trend calls for new solutions to curate news feeds that show the right information to the right friends [35, 81].”

“For instance, a study (Rader and Gray 2015) on users beliefs and judgments about Facebook’s news feed algorithm shows a significant lack of awareness in users on how the system works and whether if the news feed algorithm is biased or not.”

“This links to what Siles, Espinoza-Rojas, Naranjo, and Tristán (2019) call mutual domestication (see also Rader & Gray, 2015; Eslami et al, 2015), where, as Gruber and Hargittai (2023) recently demonstrate, people's awareness of the collection and analysis of personal data shapes the ways in people…”

“algorithmic systems has focused, for example, on how Facebook’s algorithms are understood [12, 13, 33], how users may modify their behavior in response to what they know (or believe) about a service’s algorithms [12] and on trust in algorithmic management [24].”

“For instance, users have various beliefs about the Facebook news feed curation algorithm, including passive consumption, producer privacy, consumer preferences, missed posts, violation expectations, and speculation about the algorithm [73].”

“A related method in human–computer interaction scholarship has been to study folk theories — non-authoritative, intuitive explanatory theories developed by users that circulate informally despite their accuracy (Eslami et al., 2016; Rader and Gray, 2015).”

“Users show less surprise or shock about the algorithm now than in the earlier years of its implementation (e.g. Rader and Gray, 2015).”

“Furthermore, existing research on the interaction between algorithmic users and algorithmically curated content predominantly focuses on platforms such as Facebook (Papa and Pho-tiadis 2021; Rader and Gray 2015), Spotify (Bonini and Gandini 2019), YouTube (Ormen and Gregersen 2022) and so on.”

“Using an application (FeedVis), Eslami et al. (2015) visualise algorithmic outputs based on participants’ network and News Feed stories, presenting this information back to participants during interviews to demonstrate how algorithms curate their News Feed (see also Rader and Gray, 2015).”

“We thus receive tailored information from recommender systems (Zhou et al. 2012; King et al. 2010; Rader and Gray 2015), search engines (Haim et al.”

“Various concepts have been employed to describe how users make sense of algorithms, such as folk theories (DeVito et al., 2017; Siles et al., 2020, 2022), algorithmic imaginary (Bucher, 2017), and user beliefs (Alvarado, Heuer, et al., 2020; Rader & Gray, 2015).”

“The purpose of this endeavor is to highlight the importance of understanding algorithms at the social level (Seaver, 2017), with particular attention to users ’ algorithmic awareness and practices (e.g., Bishop, 2019; Bucher, 2017; Eslami et al., 2015; Rader & Gray, 2015).”

“We choose these three affordances because previous studies showed that different social computing systems make drastically different choices regarding each of the three affordances, which greatly influences user experience and user behaviors [10, 19, 26, 43, 55, 70, 72, 82, 83, 85, 89, 95].”

“Research has shown that many users are aware that algorithms influence what is shown to them, e.g. on Facebook [40].”

“Prior research indicates that general awareness of the algorithms’ shaping of everyday experiences, such as personalized search results, recommendations, and social media curation, is surprisingly low [26, 71].”

“, 2014; Lenhart, 2009), follow and comment on the news (Rader and Gray, 2015; Hille and Bakker, 2014), keep track of sporting events (Fernández et al.”

“Research has shown that many users are either largely unaware of the algorithmic filtering/reordering on the platforms they frequently use [3, 7] or may hold very different beliefs about how they work [8].”

“Facebook through its online socio-technical system employs curation which means that the information is organized, selected, and presented to users for consumption using automated algorithmic systems (Rader & Gray, 2015).”

“Apparently, users vary widely in the degree to which they perceive and understand the behavior of algorithms which affects how they interact with and experience the systems they are using [34].”

“This includes work on new interfaces for ideation and discovery [23, 47] to finding patterns in popular curation websites [12, 5] to the perception and consequences of algorithmic curation of social feeds [10, 30].”

“Such “secondary” (and tertiary, and beyond) uses of individuals’ social media data are often invisible to users [15, 18, 34, 35].”

“Users form their own beliefs about how algorithms work [40], sometimes referred to as “algorithmic literacy” [41].”

“The popularity of third-party apps can be partially attributed to the ActivityPub protocol, which allows users to access interoperable data across the federation network without installing separate apps for each platform.”

“It is constantly evolving due to algorithms being subject to change over time, driven by technological transformations and their mutual modeling with user interactions (Rader and Gray, 2015; Bucher, 2018).”

“A similar study was conducted by Rader and Gray [45], who classified a wide range of user beliefs about Facebook’s news feed algorithm.”

“Many studies have explored people’s knowledge of the algorithmic mechanisms operating—with commercial predictive purposes—at the basis of digital platforms, and have confirmed that only partial awareness does exist (see: Bucher, 2017; Eslami et al., 2015; Hamilton et al., 2014; Rader & Gray, 2015).”

“For example, Rader and Gray [17] found that a significant portion of survey respondents believed that AI prioritizes posts in their Facebook News Feeds based on their public personal knowledge.”

“…across various digital platforms for (news) content, there are disparities in public awareness and in individuals’ ability to effectively evaluate algorithms (Eslami et al., 2015; Powers, 2017; Rader & Gray, 2015), which could be described as an emergent algorithmic divide (Gran et al., 2021).”

“For example, the lack of transparency surrounding Facebook’s News Feed filtering algorithm has been shown to leave many users unaware of its existence, leading to initial surprise and concern (Eslami et al., 2015; Rader & Gray, 2015).”

“…formed by users to make sense of their algorithmic experiences [16, 29], have been widely researched in the context of social media feeds [29, 78], content recommendations [51], and gig economy work experiences [41, 102] to understand how platform users make strategies to take advantage of…”

“For instance, studies reveal that most consumers do not recognize the use of intelligent algorithms such as Facebook News Feed Curation [58], ridesharing platforms [46], or Amazon and Netflix recommendations [70].”

“…media platforms, which is consistent with the results of studies surrounding media experiences and obtainment of literacy for content management algorithms employed by online platforms (Bishop 2019; Bucher 2017; Eslami et al. 2015; Klawitter and Hargittai 2018; Rader and Gray 2015; Swart 2021).”

“For low-quality information, Facebook users may experiment with unfollowing and hiding undesirable content to prevent it from showing up and thus reduce their exposure to it (Rader & Gray, 2015; Vraga & Tully, 2021).”

“However, the studies Rader & Gray, 2015).”

“This can be attributed to the highly realistic nature of AI face-swapping technology, which evokes user concerns regarding the security of their personal information (Rader and Gray 2015).”

“Nevertheless, most prior studies on recommender systems in social feeds have been conducted from the perspective of algorithmic experience and therefore focused on exploring user experiences in relation to opaque algorithms of AI [2, 6, 7, 15, 21].”

“Eslami et al interviewed people about their comprehension of the Facebook News Feed’s curation algorithm, asking how satisfied they are by it and their folk theories of how it works [24] [23] [59].”

“…operate, regardless of the reality, and other academic scholars use terms such as ‘ algorithmic imaginaries ’ (Bucher, 2017), ‘ folk theories ’ (DeVito, 2021; Ytre-Arne & Moe, 2021) or ‘ intuitive theories ’ (Rader & Gray, 2015) when online users do not have a technical understanding of algorithms.”

“The attribute of time decay was considered to be one of the aspects determining the visibility of posts on Facebook in the era of EdgeRank (Bucher, 2012) and supposedly also influences the new News Feed algori-thm (Rader & Gray, 2015).”

“Others have also utilized visualization approaches for awareness in social networks, to enable users to control the connections and topics that have been included by the algorithm in their filter bubble [4, 6].”

“…by algorithmic systems that preferentially expose users to ideologically aligned content while reducing exposure to opposing viewpoints [13, 22, 25, 50, 62, 86, 116, 121, 140, 189], potentially creating a reinforcing cycle between the linguistic patterns we identify and users’ information diets.”

“Several studies have addressed customers’ perception of automated curation [12, 40].”

“…that explanations are helpful and that algorithmic transparency is important [47,15,8], the amount of empirical research that investigates explanations of curation systems in experimental user studies is limited, with a few notable exceptions focused on Facebook [37,36] and YouTube [2,23].”

“Other HCI work has sought, similarly, to understand such folk theories in algorithmically managed contexts, with assumptions and self-beliefs also playing roles in this literature [2, 12, 13, 29].”

“So far, research on algorithm awareness is mainly confined to (social) media platforms, like Facebook, YouTube, or other online contexts [8, 9, 27, 31, 39], but neglected voice assistants and chatbots, even though the topic might be especially relevant here.”

“[10] In some cases, users think friends are actively hiding posts, showing that algorithmic curation on social media sites may have implications for relationships.”

“Meanwhile, Facebook employs the recommender system3 to shape people’s encounters (Rader & Gray, 2015; Seaver, 2019).”

“Researchers have used the folk theory framework to examine people’s understanding of various technologies, such as thermostats (Kempton, 1986) and social media feeds (Rader & Gray, 2015).”

“Э. Ридер и Р. Грей (Rader E., Gray R.) показали, что пользователи Facebook часто не осведомлены о том, что рекомендованная информация на домашних страницах сформирована алгоритмами и похожа на ту, которую они уже когда-то искали, и не противоречит их политическим убеждениям [33].”

“TM1 and TM2 promote algorithm awareness, an essential prerequisite for informed and effective use of technology [18, 56, 57, 60].”

“Algorithm literacy may strengthen perceived behavioral control by enhancing users’ sense of agency, efficacy, and interpretive confidence in dealing with algorithmic curation (Rader and Gray 2015; Shin and Park 2019).”

“Past work has tried to reason about "black-box" recommender systems that govern their social media feeds [18, 21, 22, 40].”

“…one emphasizes algorithmic curation and “filter bubbles,” arguing that recommender systems systematically narrow exposure to cross-cutting views [4,20,27,28]; the other stresses selective exposure and homophily, highlighting users’ tendency to seek attitude-consistent content and social ties…”

“This mechanism also incorporates users’ folk theories of algorithmic curation, often formed through algorithmic lore and gossip, that shape how repeated recommendations are interpreted as implicit validation (Bishop, 2019, 2020; Rader & Gray, 2015).”

“On TikTok, a creator’s content choices feed the recommendation algorithm (Kang & Lou, 2022), which structures what other users see, which shapes subsequent creative behaviour in a recursive loop (Esling & Devis, 2020; Rader & Gray, 2015).”