The purpose of this paper is to present how social media contributes to the mental harm of its users. Three groups of mental health risks for which social media platforms are responsible due to their business model will be presented and discussed. The proposed argumentation will be directed at shedding light on the complex landscape of the role of social media in modern society, particularly in the context of the ambiguity of research findings. The paper draws attention to the need for countermeasures while providing clues on how to properly target them to increase their effectiveness.

Full extended abstract can be found in PDF file.

Social media have fundamentally transformed how people connect, fostering a new kind of connection where physical and synchronous presence are no longer necessary. While research has shown that social media can connect people in various ways, an overview and clear conceptual understanding of this social media-induced connection (SMIC) is lacking. To address this, a systematic review of 121 peer-reviewed studies across various disciplines was conducted. The review identifies two core components of SMIC: (1) the targets that people connect to, ranging from human (peer, group) to nonhuman entities (e.g., company, nature, society), and (2) the cues that people connect through, including digital-enabled (e.g., similarities) and digital-born cues (e.g., technical affordances). The review also finds that valence, depth, and duration of SMIC are dynamic, context-dependent characteristics. Taken together, the study conceptualizes SMIC as a gratification of using social media whereby one experiences relatedness that is directed toward human and/or nonhuman targets, formed through digital-enabled and/or digital-born cues, and characterized by valence, depth and duration. This conceptualization addresses inconsistencies in current research on social media usage and connection and offers a cohesive foundation for future research regarding this topic.

Social Media Eras: For Friends, For You, For Whom?

Social media is over 20 years old, and has changed drastically from platforms for connecting with friends to highly-customized media environments for obtaining any possible information (e.g., breaking news, comedy, recipes). We theorize three eras of social media: First, as a social space in the early World Wide Web, driven by users’ desires to connect with others. Second, as a mass information space, driven by content and dominated by algorithms. Finally, most recently, as a powerful but fragmented community space in search of its next direction (e.g., decentralization).

These eras parallel the stages of the Web, but focus on the attributes to the social media landscape specifically. We argue that Social Media 1.0 began as early as the Web itself, and led to Web 2.0. Social Media 2.0 began in about 2013 with the implementation of content recommendation algorithms across platforms. As we move to Web 3.0, we believe we are also on the cusp of Social Media 3.0, fueled the user exodus from the biggest platforms to decentralized networks. Next, we briefly describe each era, its timeline, characteristics, relevant platforms, and implications for communication. We focus on globally-available platforms, while noting that our analysis comes from a western perspective, and acknowledging that these eras may not apply to all cultures and their relevant media spaces.

Social Media 1.0: “People You May Know” (2003-2012)

Though other social networking sites existed earlier, when MySpace was launched in 2003, it arguably became the first widely-adopted and mainstream social media platform. Shortly after, in 2004, Facebook was launched for college students, and in 2006 it became publicly available to users. From there, Facebook heavily shaped the first social media era.

Learning from Friends

The focus of these platforms was friends: Finding them, making new ones, and keeping up with them. As a result, information shared was largely friends’ experiences. These sites became popular for self-presentation, social support, and advice. Features highlighting this usage are MySpace’s “Top 8,” and Facebook’s suggestions of “people you may know” through mutual friends. LinkedIn was also launched in this era (2003), and – though focused on professional networking – offered similar features suggesting contacts based on 1st-, 2nd-, and 3rd-degree connections. For users, these features allowed them to build their networks on the Web, combining people they knew offline with people they met online. For platforms, keeping people engaged with each other meant keeping them on their site, and enabled them to build user profiles for advertising.

Leading the Shift

Twitter, also launched during this time (2006), may serve as the bridge to the second era, focused on more public information sharing. It shares with Myspace, LinkedIn, and Facebook the focus on one’s network. However, its connections were directed rather than mutual, and its network was public by default. Therefore, it quickly became a space for rapidly-developing information (e.g., @jkrums famous 2009 tweeted photo breaking the news of Captain Sully’s landing on the Hudson).

Social Media 2.0: “For You” (2013-2022*)

As networks grew and information became shared more publicly, Facebook started prioritizing content in users’ feeds by “popularity,” and by 2013, it announced a proprietary content ranking algorithm. By 2016 Instagram and Twitter had followed suit with their own unique ranking algorithms. These algorithms have since evolved into complex machine learning methods that quickly react to user interactions with highly-targeted content. The most “aggressive” of these algorithms is on TikTok, whose platforms hinges on its super-customized “For You Page.”

Learning from Content

The focus of social media in this algorithmically-driven era shifted from “who” to “what,” as feeds began to fill with “suggested” posts, especially on Instagram and Twitter. In 2020, TikTok became the most-downloaded social media platform, where users spent hours scrolling short videos from unknown others. Users benefit from TikTok’s (and later, Instagram Reels’) sensitive algorithm leading them to hyper-relevant “Toks” (e.g., BookTok) which also connects them to identity communities of similar strangers. Platforms in this era benefit from highly-targeted advertising profits, built on rich user data profiles, and supplemented with real-time interaction data through algorithms. This era is further shaped by the resulting influencer marketing, playing on the self-expression of earlier social media, while benefiting from larger public networks. This combination has further amplified the influence of strangers through users’ parasocial relationships with them.

Predicting the Next Shift

For better or for worse, social media content has become truly endless, as suggested content keeps a feed full even once one has seen all their friends’ updates. There is no limit to information, but a limit to time and emotional bandwidth. “Doomscrolling” through overwhelming amounts of content became a common experience in this era. This behavior, certainly fed by stressors of the pandemic, has arguably launched us into the third era: One of contemplation about what we want out of social media.

Social Media 3.0: Who is it For? (2022-present)

The developing third era is marked by a social media landscape that feels out of control, while controlling its users. People are experiencing algorithmic fatigue, and are more concerned than excited about constantly-touted AI-generated content. These factors stirred dissent in 2022, when Musk purchased Twitter (X), came to a head when Trump was re-elected in 2024, and were further exacerbated by Meta’s political leanings in early 2025. Millions of X (then, Facebook, and Instagram) users have fled to Bluesky, and a smaller proportion to Mastodon, both decentralized platforms, not controlled by one “master algorithm.” While users have always migrated across platforms, the recent shift is a signal that they are serious about crafting a social media experience that offers agency over their experiences (e.g., the Friendster re-launch). Questions at this stage are: What will be the defining core of this era, as friends was for the first era, and content was for the second era? What will users seek and gain from platforms in this new era? Finally, how will platforms that promise a new model (e.g., algorithm -free) thrive at this stage?

Youth live in a world saturated with digital media such as video games or social media. Exposure to digital media may affect their social, cognitive, and identity development (Cone & Konijn, 2018). However, not all children and adolescents have equal opportunities to access and use digital media, which was especially visible during the COVID-19 pandemic (Rideout et al., 2021). Scheerder and colleagues (2017) distinguished several “inequality factors” that could be applied to youths’ digital media use. Further, the model by Valkenburg and Peter (2013) suggested that such factors may (1) predict the way how youth use digital media, and (2) moderate the relationship between digital media use and media outcomes. Based on that, we considered the following inequality factors related to youth’s (1) background, e.g., socio-economic status, (2) social environment, e.g., parental rules, and (3) individual factors, such as age and gender, personality traits, or diagnosis of developmental problems like ADHD. While some longitudinal studies have considered such factors (cf. Orben et al., 2022), until now, the possible effects of inequality factors on youth’s development were not systematically reviewed. Therefore, by taking a systematic review approach focused on longitudinal studies, we aimed to answer the following question:

What factors explain the effects of inequalities in media access and media use on children’s and adolescent’s cognitive, social, and identity development?

This would allow us to (1) map the possible factors contributing to the inequalities in youths’ digital media use and (2) unravel different patterns of susceptibility to digital media effects on youth’s social, cognitive, and identity development.

Method

This systematic review’s plan was pre-registered at PROSPERO (https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42025625378). The current work follows the recommended guidelines for systematic reviews (PRISMA 2020 Checklist; Page et al., 2021).

Inclusion and exclusion criteria

Papers had to match the following criteria to be included in our review: 1) be published in English in peer-reviewed sources, such as journal articles or book chapters, (2) include quantitative empirical studies in a longitudinal design, 3) test children (aged between 3-11) or adolescents (aged between 12-18) who may be enrolled in educational program from preschool education (kindergarten) to secondary education (high school). Papers involving parents or teachers reporting on their children’s or students’ outcomes within the given age group were also included. Finally, (4) we included only papers that tested at least one factor (background, social, or individual) explaining the effects of inequalities in media access or media use on changes in children’s and adolescents’ development (cognitive, or social, or identity) over time.

Search

The timeframe of our search within five databases (PubMed, ERIC, WoS, PsycArticles, SCOPUS) was from 2008 (after the launch of the iPhone) until inception to 5th of December 2024. The search included keywords and free text terms for (synonyms of): “digital media” (predictor) AND “inequality factors” (predictor/moderator) AND “longitudinal” (design type) AND “youth” (sample).

Screening procedure

Upon deduplication, papers were screened in Rayyan. All titles and abstracts were screened by one independent reviewer. Additionally, 30% of non-overlapping papers were screened by two other reviewers (~15% per second and third reviewer) to compute the inter-rater reliability score, aiming to reach an almost perfect agreement level (Cohen’s kappa 0.81-1.00; McHugh, 2012) with the first reviewer. Discrepancies were resolved through a discussion between the reviewers. In case a decision was not possible based on a title and an abstract only, a full text was reviewed. The same screening procedure was applied for the full-text screening and later – data extraction.

Data extraction

A standardized data extraction form was developed to extract the following data from each article: general information: title, author, publication year, doi, sample information: country, mean age of participants, age group (children, adolescents), sample size (percentage of male participants), education level, inequality-related variables: background details (e.g., family income, parent education level, home language, country of birth), individual details (age, gender, diagnosis: ADHD, autism, etc.); predictor-related variables: digital media exposure type (social media, video games, video or TV watching), digital media exposure frequency and time, digital media access device (smartphone, laptop, etc.); outcome-related measures: type of measurement (e.g., self-report), details of the measurement (e.g., scale name and reference), study design details: wave number, the time difference between the waves, waves dates. The data will be extracted independently by one reviewer. The extracted data will be recorded in an Excel spreadsheet.

Data synthesis

Following the recommended guidelines for systematic reviews (PRISMA 2020 Checklist; Page et al., 2021), a systematic review will be performed on the relationship between inequalities, digital media use, and developmental outcomes. Outcomes will be synthesized based on the developmental outcome type (social, cognitive, identity), media exposure type (e.g., social media, video games, video watching), inequality factor type (e.g., background, social, individual), and sample type (children and adolescents).

Risk of bias (quality) assessment

The quality of all included studies will be assessed by one reviewer. Similarly to the screening procedure, approx. 30% will be cross-checked by the second independent reviewer and any disagreements will be resolved through discussion with the third reviewer. The methodological quality and bias of the included studies will be evaluated using the Joanna Briggs Institute (JBI; Moola et al., 2020) checklist for Cohort Studies (11 items).

Preliminary Results

Search and Screening. The search of five databases resulted in k = 2077 records. Next, 462 duplicates were found and resolved in Rayyan, leaving 1908 titles and abstracts that are currently being screened. The outcomes of this review will be presented at the conference.

Discussion

We anticipate that this review will provide more nuanced insights into the possible effects of digital media on youth’s social, cognitive, and identity development. Moreover, they may contribute to scientific and public debates related to this topic and provide input to the current recommendations regarding digital media use among youth (APA, 2023). Further, based on the included studies’ quality assessment, practical implications for future research will be discussed, especially in terms of the importance of sample diversity in longitudinal studies with youth samples (Fakkel et al., 2020).