PaperlayerView exact revision on arXiv

Source-linked AI summary

"A Necessary Evil": Teenagers' Sensemaking of Privacy and Safety Settings on Social Media

Jingxin Dong, Lingyun Chen, Chen Ling, Colin M. Gray

arXiv:2609.28685v1cs.HC

TL;DR

The paper asks whether social media privacy and safety controls give teenagers meaningful protection and agency. It analyzes think-aloud task sessions with teenagers and finds that they may accommodate weaker-than-expected protections while still treating platforms as authorities on what protection should look like.

  • Problem

    The paper asks whether social media privacy and safety controls give teenagers meaningful protection and agency.

    Source: §1 · p. 1

  • Method

    The authors analyzed think-aloud sessions with 11 teenagers aged 14–17 completing six privacy and safety tasks, developing two themes.

    Source: §1 · p. 2

  • Results

    Participants worked out settings’ meanings through routines and prior experiences, accommodated softer-than-expected protections, and still treated platforms as authorities on protection.

    Source: §7 · p. 19

  • Takeaways & Limitations

    Evaluating controls one feature at a time cannot establish whether teenagers are protected.

    Source: §7 · p. 19

    The study captures one task-based session.

    Source: §6 · p. 19

Abstract

from arXiv · show

Social media platforms are embedded in teenagers' daily lives, supporting friendship and identity while exposing teenagers to unwanted contact and privacy harms. Previous scholarship has documented how attention capture strategies and dark patterns shape social media use, and we extend this work to better understand platform settings that ostensibly provide privacy and safety protection. We report on think-aloud sessions with 11 teenagers aged 14 to 17 who completed six privacy and safety tasks on Instagram, TikTok, Snapchat, and YouTube. We show how participants worked out what a setting meant through their routines, boundaries, and prior experiences, how they accommodated protections softer and less predictable than expected, and how they treated the platform as the authority on what protection should look like. We argue that feature-by-feature evaluation cannot establish whether teenagers are protected, and that platforms should carry the obligation to show that a protective action took effect and is durable.

1 Introduction

The paper examines whether social media privacy and safety controls provide teenagers meaningful protection and agency, given that settings being available does not ensure they understand, trust, or experience them as protective. It studies teenagers’ interpretations in everyday context and finds they may accommodate weaker-than-expected controls while still treating platforms as authorities on protection.

1 Introduction

  • Social media supports teenagers’ friendship, communication, and identity while exposing them to unwanted contact, privacy violations, and difficult-to-regulate patterns of use.The introduction frames adolescence as a formative period for developing boundaries, autonomy, and self-regulation.Source: §1 · p. 2
  • The study examines how teenagers interpret privacy and safety controls and whether those controls support agency in everyday use.The introduction notes that controls being available does not ensure teenagers understand their purpose, trust their operation, or experience meaningful protection.Source: §1 · p. 2
  • The researchers used think-aloud sessions with 11 teenagers aged 14–17 completing tasks involving notifications, screen time, account privacy, reporting, data access, and deletion.They analyzed the sessions through reflexive thematic analysis and developed two themes.Source: §1 · p. 2
  • Participants interpreted controls through their routines, boundaries, and prior experiences, often accommodating protections softer or less predictable than expected.They could recognize controls as weak, partial, or confusing while continuing to rely on familiar platforms they found difficult to avoid.Source: §1 · p. 2
  • Dense or unclear settings were often left unchanged, while extra steps could signal safety, responsibility, or legitimate design and reinforce platform authority.Repeated use and limited alternatives could strengthen that reliance.Source: §1 · p. 2

relate these interpretations to their everyday practices, boundaries, and responsibilities?

The paper contributes an account of teenagers’ situated interpretations of privacy and safety controls and the authority they grant platforms. It argues that evaluating controls individually cannot establish whether teenagers are protected, placing responsibility on platforms to demonstrate durable protection.

1 Introduction

  • Across six tasks, the study documents what teenagers noticed in controls and how they judged their strength and purpose in relation to routines, boundaries, and prior experiences.The contribution focuses on situated use and participants’ responsibilities.Source: relate these interpretations to their… · p. 3
  • Teenagers could recognize a control as weak yet treat the platform as the authority on protection, and friction in a protective flow could itself prompt trust.This complicates design responses intended to make manipulative patterns more visible.Source: relate these interpretations to their… · p. 3
  • Feature-by-feature evaluation cannot establish whether teenagers are protected, so platforms should show that protective actions took effect and remain effective without continued management.The paper proposes an asymmetric allocation of responsibility to platforms.Source: relate these interpretations to their… · p. 3

2 Related Work

Prior work examines engagement and dark-pattern mechanisms, usable privacy and self-limiting controls, and how young people’s privacy and protection are shaped by social belonging and reliance on platforms. Together, this research motivates examining what teenagers believe a setting will do and whether it will hold when they choose it.

2.1 Platforms, Attention Capture Strategies, and Dark Patterns

  • Interface features can weaken or support agency: on YouTube, autoplay and recommendations weakened participants’ sense of agency, while search and playlists supported more deliberate viewing [49]; Twitter feed filters and indicators that users had exhausted new content improved agency.Source: §2.1 · p. 3
  • Teenagers show partial awareness, while prior knowledge and social and technical circumstances shape resistance; an expert evaluation of teen accounts on four social media platforms found dark patterns in every privacy and safety task.Source: §2.1 · p. 4

2.2 Usable Privacy, Sensemaking, and Self-Limiting Tools

  • Usable-privacy criteria extend beyond locating a choice to understanding its implications, yet an easy-to-operate control can leave users unable to tell what it accomplished.Studies found that functional and randomly implemented feed controls both increased perceived control, though a later replication found a smaller satisfaction effect.Source: §2.2 · pp. 4–5
  • Sensemaking is ongoing, revisable, and frequently mistaken; users interpret platform cues, revise theories as platforms change, and an alternative-feed probe helped participants develop explanations informing subsequent plans.Source: §2.2 · p. 5
  • Research on feeds and disclosure is well developed, but evidence remains thin on how users read controls themselves.Prior work shows that people can hold conflicting expectations about a control’s effects and struggle to attribute later changes to a particular action.Source: §2.2 · p. 5
  • Self-limiting tools are valuable when their enforcement matches users’ own definitions of unwanted use, which can depend on situation and intention [49].Users seek controls strong enough to interrupt unwanted use yet flexible enough to fit personal definitions of distraction.Source: §2.2 · p. 5
  • Prior work largely evaluates privacy actions and self-limiting tools by usability or behavioral effects among adults, while this study asks what teenagers believe the controls do for them.Usable privacy research asks whether someone can complete a privacy action, and self-limiting research asks whether a tool changes behavior.Source: §2.2 · p. 5

2.3 Youth Expectations, Belonging, Identity, and Trust

  • Privacy and self-limiting controls can affect teenagers’ inclusion, self-presentation, and participation in friends’ routines.Visibility, sharing, and use limits can change opportunities for inclusion, while identity work and peer responses raise the social stakes.Source: §2.3 · p. 6
  • Trust in connected people and trust in platforms are distinct: ambiguous audiences can discourage disclosure [44], while many young people distrust companies’ ability to resolve harassment fairly.Distrust does not necessarily mean leaving a platform; users can trust some aspects of a system while doubting others.Source: §2.3 · p. 6
  • Teenagers may accept platform terms and permissive privacy practices to preserve social relationships, granting platforms practical authority through their defaults and rules.Workshop participants recognized personal-data reuse while avoiding practices that could disrupt ordinary interaction.Source: §2.3 · p. 6
  • The section asks how to understand teenagers who expect protection from platforms they doubt, since prior youth trust accounts measure attitudes away from the moment of a setting decision.The authors contrast resignation and doubts that available actions will protect users with the question of whether a particular setting will hold.Source: §2.3 · p. 6

3 Method

The study used one-on-one think-aloud interviews to examine how teenagers interpreted privacy and safety settings during tasks on platforms they already used. It recruited 11 teenagers aged 14–17, used task-based interviews to explore situated interpretations, and followed consent and confidentiality procedures.

3 Method

  • One-on-one think-aloud interviews captured what participants thought settings were for, how they expected them to work, and whether they would use them.The approach examined interpretations during common tasks on platforms participants already used.Source: §3 · p. 7

3.1 Participants and Recruitment

  • The interpretive, task-based study interviewed 11 teenagers aged 14–17 (M = 15.6, SD = 0.8) in one-on-one Zoom think-aloud sessions, assigning each two platforms they already used.Tasks included screen time, notifications, and setting an account to private.Source: §3.1 · p. 7

3.2 Research Ethics

  • The study received IRB approval and obtained digital consent and assent before interviews, explaining confidentiality, voluntary participation, and the right to skip questions or withdraw.Researchers also asked whether a parent was present so participants could speak freely.Source: §3.2 · p. 7

3.3 Teen Interview Study

The interview protocol elicited participants’ real-time interpretations of privacy and safety controls. Iterative analysis developed themes linking platform design with participants’ lived contexts and their deference to platforms.

3.3 Teen Interview Study

  • The protocol covered platform familiarity, a think-aloud task session, and follow-up questions.Source: §3.3.1 · p. 8

3.3.1 Interview Procedure.

  • Participants completed privacy and safety tasks while narrating their actions and interpretations, then discussed what felt easy, difficult, useful, or trustworthy.Tasks included managing notifications, setting screen-time limits, making accounts private, reporting content, downloading data, and deleting accounts; participants shared phone screens over Zoom.Source: §3.3.1 · p. 8

3.3.2 Preparing the Corpus.

  • Researchers screened for interpretations, expectations, evaluations, and links to routines or prior experience, then tagged excerpts by platform and task for cross-platform comparison.Source: §3.3.2 · pp. 8–9

3.3.3 Organizing Sensemaking Excerpts.

  • Nonexclusive analytic codes captured what participants interpreted, while platform and task tags preserved encounter context; Organic and Prompted marked whether interpretations arose without or after interviewer intervention.Source: §3.3.3 · p. 9

3.3.4 Developing Themes.

  • Themes arose from relationships among excerpts, not tag frequencies, and the researchers did not assess intercoder reliability under reflexive thematic analysis.Source: §3.3.4 · p. 9
  • Comparison merged interface functionality and participants’ lived contexts into the theme Living With and Around Platform Design.Participants’ interpretations moved between interface features and their routines, relationships, expectations, prior experiences, and concerns.Source: §3.3.4 · p. 9
  • When controls were confusing or unexpected, participants sometimes left defaults, read extra steps as care, trusted familiarity, or deferred uncertainty to platforms; this pattern was not uniform.Source: §3.3.4 · p. 10

3.4 Researcher Positionality

The authors situate the study in an interdisciplinary team with prior familiarity with teen social-media privacy and safety settings.

3.4 Researcher Positionality

  • The four-member team brought expertise across education, law, youth studies, security, design, and HCI, alongside prior familiarity with teen privacy and safety settings.The authors say these perspectives helped them examine how interface design and dark patterns shape settings and teenagers’ understanding and use of social media.Source: §3.4 · p. 10

4 Findings

The findings describe how participants interpreted protective settings through lived experience and how they continued to treat platforms as authorities when settings were confusing or unexpected.

4 Findings

  • Living With and Around Platform Design captures how participants used routines, boundaries, and prior experiences to judge a setting when its interface did not explain its meaning or usefulness.Source: §4 · p. 10
  • Teens Defer to the Platform as Authority captures how participants treated the platform as the reference point for protection when a setting was confusing or behaved unexpectedly.Source: §4 · p. 10

4.1 Living With and Around Platform Design

This section examines how participants made sense of protective and unfamiliar platform settings through routines, personal boundaries, prior experience, and clear problems, and how they judged control by whether settings produced expected outcomes while continuing to treat platforms as authorities on their meaning.

4.1.1 The name of a setting did not predict its strength or its operation.

  • Settings’ names did not predict their operation: reminders could be snoozed, controls could be temporary or reversible, and platform differences made actions and outcomes hard to anticipate.Participants sometimes had to work out each platform’s settings independently because similar controls behaved differently and their effects were difficult to verify.Source: §4.1.1 · p. 11

4.1.2 Features are understood through everyday routines.

  • Participants understood settings through existing routines: limits were shaped by attention and study, while notification choices reflected availability to friends and concern about missing important messages.Some participants used sleep mode to signal unavailability, while others avoided pausing notifications because they feared losing contact or missing something important.Source: §4.1.2 · pp. 11–12

4.1.3 Privacy and safety are interpreted through personal boundaries around contact, visibility, and (dis)comfort.

  • Privacy and safety reflected personal boundaries over who could contact participants, how visible they wanted to be, and what information platforms collected or observed.Familiarity shaped whom participants accepted as a contact, while platform surveillance concerns could affect use without leading participants to leave.Source: §4.1.3 · p. 12

4.1.4 Features feel usable when their steps, risks, and outcomes are clear.

  • Prior experience made controls easier, but did not guarantee action; knowing where to find a feature did not make it feel safe, and fear of losing the account made leaving difficult.Source: §4.1.4 · pp. 12–13
  • PS_10 saw data download as a backup if an account was disabled, while PS_02 worried they might not recover an account after deletion.Neither participant had used the feature before.Source: §4.1.4 · p. 13
  • Participants found features useful when they could connect them to a problem they could already imagine, rather than simply because the features existed.Their accounts tied usefulness to anticipated problems, not to what a feature produced.Source: §4.1.4 · p. 13
  • Control depended on whether the feed matched expectations: unwanted content undermined PS_09’s sense of choice, and the platform’s next display was the only available evidence.PS_09 could not verify control beyond what the platform showed next.Source: §4.1.4 · p. 13
  • Participants continued to treat the platform as the authority on what a setting meant, even when their own expectations had not been met.Source: §4.1.4 · p. 13

4.2 Teens Defer to the Platform as Authority

Participants treated platform design as the main guide to what settings meant, often leaving confusing defaults unchanged and interpreting added steps as signs of care. Trust also reflected familiarity, perceived competence, and limited alternatives.

4.2 Teens Defer to the Platform as Authority

  • Participants treated platform design as the main guide to setting meaning, left confusing defaults in place, and looked to platforms to fix problems.Extra steps in reporting and deletion flows could signal safety, organization, listening, or concern.Source: §4.2 · p. 13, §4.2.2 · p. 14

4.2.1 Defaults and confusing design are often accepted rather than questioned.

  • Unclear settings and lengthy explanations reduced engagement, while uncertainty could leave participants stuck, preserve defaults, or end their attempts to act.Some stopped trying when a setting did not respond as expected.Source: §4.2.1 · pp. 13–14

4.2.2 Friction is often interpreted as care, organization, or responsible design.

  • Participants read extra steps as potentially protective or considerate, interpreting them as safety checks, organized reporting, attentive listening, or help before account deletion.The same added effort could make deletion longer while seeming to serve a helpful purpose.Source: §4.2.2 · p. 14

4.2.3 “Necessary Evil”: Trust and deference are shaped by familiarity, perceived competence, and constrained choice.

  • Trust drew on ease of use, connected services, and perceived competence, but constrained alternatives also kept participants attached despite doubts.Participants continued to look to platforms for fixes, even when they were unsure how to contact them.Source: §4.2.3 · p. 15

5 Discussion

The discussion considers how teenagers interpret privacy and safety controls through their routines and prior experiences, and why evaluating whether they can find and complete controls misses what protection they understand them to offer and what happens after they act. It argues that feature-by-feature evaluation is insufficient and that platforms should show that protective actions work as represented and last.

5.1 “They Want You to Stay”: Between Recognition and Deference

  • Teenagers recognized that protective controls could fail and serve platform interests, but this awareness rarely led them to adjust their behavior or leave.Participants snoozed or ignored screen-time reminders, and a participant who distrusted Instagram remained on it.Source: §5.1 · p. 16
  • Participants often read the effort required by privacy and safety flows as evidence of care, allowing visible complexity to signal platform competence rather than obstruction.Multi-step authentication appeared “safe in their own way,” while deletion prompts, reporting categories, and requests for reasons could seem protective.Source: §5.1 · p. 16
  • Even when participants recognized platform failures or distrusted a service, uncertainty led them to rely on its defaults and explanations as the safest available judgment.Participants left confusing settings untouched, and one distrusted Instagram while still treating its programmed settings as safest.Source: §5.1 · pp. 16–17
  • Participants lacked independent standards and recourse, leaving them dependent on platforms they questioned as the source of problems, their interpreters, and their imagined venue for repair.The discussion describes this dependence as making platforms a “necessary evil.”Source: §5.1 · p. 17
  • Evaluation should assess whether teenagers can verify that protection took effect, persists without continued management, and can be challenged through an independent authority.The authors argue that protective design should be judged by the power it gives teenagers, including support beyond the platform’s own judgment.Source: §5.1 · p. 17

5.2 The Limits of Feature-Specific Evaluation of Privacy and Safety Controls in a Platform Ecology

  • Protective defaults can reduce teenagers’ burden, but offering a control alone does not establish meaningful agency or complete platform protection.Platforms still determine available options, their framing, applicable defaults, and what follows a selection, while users may be unable to verify the resulting practice or response.Source: §5.2 · p. 17
  • Across platforms, engagement-oriented strategies recurred, but the work of managing protection did not transfer, limiting what feature-by-feature comparisons can establish.Participants wanted consistent settings, carried expectations between deletion flows, lost their path, and linked trust across connected services.Source: §5.2 · p. 17
  • Control qualities such as awareness, comprehension, effort, reversibility, and exposure to nudging depend on interactions among defaults, explanations, system behavior, routines, and wider environments.Source: §5.2 · p. 18
  • Privacy controls depend on explanations and faithful implementation to verify unseen data practices, while safety controls initiate platform processes for classification, review, remedy, and enforcement.The discussion proposes examining the point where a visible user action passes into an organizational process the user cannot independently inspect or complete.Source: §5.2 · p. 18
  • Evaluating whether teenagers can locate and complete controls misses their interpretive work to assess what protection a setting offers, what safeguards operate behind it, and who can change the conditions producing risk.Participants also had to infer which options the platform had already judged protective.Source: §5.2 · p. 18
  • Platforms and participants assigned responsibility differently: platforms treated a control’s availability as fulfilling their obligation, while participants took it as evidence the option had been evaluated and would work.Source: §5.2 · p. 18
  • Because teenagers contribute contextual knowledge, preferences, and harm reports while platforms control defaults, explanations, implementation, and moderation, platforms should test, monitor, correct, and respond to failures.Source: §5.2 · p. 18

6 Implications and Future Work

The paper proposes evaluating protection by whether teenagers can form an intention, find and complete a protective change, and understand and trust its outcome. It also identifies limits of task-based evidence and open questions about whether design interventions close gaps between expectations and protections.

6 Implications and Future Work

  • A setting’s existence is insufficient when a control can remain hard to reach or easy to reinterpret in ways that serve platform goals.Source: §6 · p. 18
  • Current platforms fall short on each proposed evaluation dimension for at least some settings: completion, teen-reported confidence, and durability.The authors propose assessing whether teenagers can form a clear protective intention, locate a setting without expert guidance, complete a change without undue friction or reversal, and understand the outcome well enough to trust it.Source: §6 · pp. 18–19
  • The study captures one task-based session, leaving unresolved whether difficulty reaching a setting translates into avoidance over time.Source: §6 · p. 19
  • Participants often had inaccurate expectations, such as expecting screen-time limits to lock an app or reports to remove content immediately.These mismatches follow predictably from soft, reversible, and ambiguously labeled design.Source: §6 · p. 19
  • Whether honest framing or visible confirmation can close the gap between expected and actual protection remains a question for direct testing.Source: §6 · p. 19
  • Task-based interviews may not show how teenagers interpret settings during their own use, especially when study accounts lack followers, message histories, or accumulated activity.Some interpretations also reflected feature explanations given during sessions, and the boundary between spontaneous and supported understanding was not always sharp.Source: §6 · p. 19

7 Conclusion

Think-aloud sessions with 11 teenagers examined privacy and safety tasks on platforms they already used. Participants interpreted controls through their routines and prior experiences, accommodated protections softer than expected, and still treated platforms as authorities on protection; the authors call for platform accountability beyond feature-by-feature evaluation.

7 Conclusion

  • Participants worked out what controls meant through their routines and prior experiences, accommodated softer-than-expected protections, and still treated platforms as authorities on protection.Source: §7 · p. 19
  • The authors argue that evaluating controls one feature at a time cannot establish whether teenagers are protected, and that responsibility belongs to platforms that design options and control their consequences.Source: §7 · p. 19
  • Assessment should ask whether a protective action can be confirmed, holds without continued management, and allows teenagers to contest a platform’s decision elsewhere.Source: §7 · p. 19

Sensemaking

The sensemaking analysis distinguishes participants’ interpretations during tasks from their later reflections and contextual talk. Its codes capture how features connect to platform use, privacy concerns, lived experience, expectations, and comparisons across platforms.

Sensemaking

  • During tasks, the analysis captured moments when participants worked out what a feature does or how it works.Source: Sensemaking · p. 24
  • Contextual codes captured platform use, privacy and safety concerns, links to lived experience, and comparisons across platforms.Source: Sensemaking · p. 24
  • Reflection codes covered follow-up talk about the experience, while platform-expectation codes recorded what participants thought a setting ought to do.Source: Sensemaking · p. 24
  • Contextual markers recorded how an interpretation entered the interview, not task completion outcomes; analytic codes were nonexclusive and counts represented code applications, not unique excerpts.Source: Sensemaking · p. 24, Table 3 · p. 24
Loading 2609.28685v1…