Source-linked AI summary
Available but Not Usable: Dark Patterns and Interaction Cost in Social Media Privacy and Safety Settings for Teens
Jingxin Dong, Lingyun Chen, Chen Ling, Colin M. Gray
arXiv:2609.28672v1cs.HC
TL;DR
Protective settings can exist without making their routes or outcomes clear to teenagers. The paper pairs expert reconstruction of privacy and safety task flows with usability testing, examining interaction cost and dark patterns. Dark-pattern strategies recurred on every examined platform, and the paper presents a wayfinding audit as an initial approach.
Problem
The presence of a protective setting reveals little about the effort required to find, interpret, and configure it or whether it remains effective.
Source: §1 · p. 2Method
The paper pairs expert reconstruction of six privacy and safety task flows across four platforms with usability testing and dark-pattern analysis.
Source: §1 · p. 2, §7 · p. 21Results
Dark-pattern strategies recurred on every examined platform, while some attempts cost less because participants stopped before completing the required configuration.
Source: §7 · p. 21Takeaways & Limitations
The paper offers a wayfinding audit that combines expert evaluation, usability testing, interaction-cost measurement, and dark-pattern analysis.
Source: §1 · p. 2, §7 · p. 21The procedure is scoped to the examined platforms and tasks; other contexts may require a different balance among its components.
Source: §5.2 · p. 20
Abstract
from arXiv · showhide
Social media platforms are central to teenagers' lives, and their designs can expose users to privacy, safety, and wellbeing harms. Platforms increasingly offer protective settings, though the presence of a control reveals little about whether teenagers can find, use, and benefit from it over time. We paired an expert evaluation of six privacy and safety tasks across TikTok, Instagram, Snapchat, and YouTube with moderated think aloud sessions in which 11 teenagers aged 14 to 17 attempted the tasks. Interaction cost and dark patterns analysis allowed us to compare the complexity designed into each task with the effort participants incurred as they located, configured, and interpreted controls. Recurring dark patterns appeared across tasks, and most participant attempts exceeded the expert baseline. Protective settings therefore risk being insufficiently usable or durable in practice, and we propose a wayfinding audit that integrates expert evaluation, usability testing, interaction cost, and dark pattern analysis.
1 Introduction
Protective settings may be difficult for teenagers to find, interpret, configure, and keep effective even when they exist. The study compares expert-documented routes with teenagers’ task attempts to assess interaction cost and dark patterns.
1 Introduction
- A setting’s presence reveals little about the effort needed to locate, interpret, configure, and verify it, while some safeguards require activation or can expire or be displaced.Research on digital wellbeing tools reports that limits are often snoozed, deleted, or worked around.Source: §1 · p. 2
- The study asks how much interaction cost and which dark patterns occur across six tasks on four platforms, and how teenagers’ costs diverge from an expert baseline.The platforms are TikTok, Instagram, Snapchat, and YouTube.Source: §1 · p. 2
- Researchers documented task routes, interface states, interaction costs, and dark patterns, then used moderated think-aloud sessions to observe teenagers completing the tasks.They reconstructed attempts from recordings and coded time, clicks, interface states, scrolling, completion, and observable struggle.Source: §1 · p. 2
- The paper operationalizes interaction cost to separate platform-designed route costs from costs accumulated through search, interpretation, backtracking, and uncertainty.The comparison puts the documented route beside the cost incurred when a participant must find, interpret, and complete it.Source: §1 · p. 2
- Most participant attempts cost more than the documented route, including costs accumulated before participants entered it.The comparison highlights a gap between route complexity as documented by experts and effort incurred during teenagers’ attempts.Source: §1 · p. 2
- The proposed wayfinding audit combines dark-pattern and temporal analysis, interaction-cost measurement, expert evaluation, and usability testing to assess whether intended users can find, use, and interpret a safeguard.It is offered as support for diagnosis before deployment and evaluation by researchers, auditors, and regulators.Source: §1 · pp. 2–3
2 Related Work
Prior work examines manipulative platform design and the usability of privacy controls, including teenagers’ responses to dark patterns. This paper connects those lines by measuring the cost of reaching controls across platforms.
2.1 Social Media, Dark Patterns, and Teenagers
- Research distinguishes engaging strategies that prolong use from governing strategies that narrow users’ choices, and studies find that awareness alone does not ensure resistance to dark patterns.Teenagers may interpret manipulative designs through advice from parents and peers, and resistance depends in part on recognizing a design as manipulation.Source: §2.1 · p. 3
- Field-consolidation work has followed; Gray et al. harmonized ten regulatory and academic taxonomies into a three-level ontology providing a shared vocabulary for naming manipulative design.Source: §2.1 · p. 3
2.2 Usable Privacy and Security
- Usable-privacy research finds that controls may be inconsistently labeled or buried, and users may give up even when a site formally offers the option.The field therefore treats a control’s presence as weak evidence that people can reach it.Source: §2.2 · p. 4
- Where earlier evaluations often examined one platform or mechanism at a time, this study compares routes across four platforms and treats each control as having measurable cost.The comparison asks what a control demands of the person trying to reach it.Source: §2.2 · p. 4
3 Method
This section describes a method combining expert mapping of privacy and safety task flows with moderated think-aloud sessions in which 11 teenagers attempted tasks on platforms they regularly used. It explains how participant struggle was coded and aligned, and how participant paths were descriptively compared with expert flows without inferential claims about platform or participant differences.
3 Method
- The two-part design used expert task-flow evaluation as a structural baseline and think-aloud sessions to examine how 11 teenagers navigated the same controls in practice.Participants completed tasks on two platforms they reported using regularly.Source: §3 · p. 4
3.1.1 Platforms and Tasks.
- The expert evaluation covered TikTok, Instagram, Snapchat, and YouTube and six tasks, treating each available platform-task combination as a separate flow.Tasks were setting a screen-time limit, making an account private, managing notifications, reporting content, downloading account data, and deleting an account; not every task applied on every platform.Source: §3.1.1 · pp. 4–5
3.1.2 Expert Evaluation Procedure.
- Experts used under-18 research accounts, started each task at the platform home screen, and documented the route and its minimum clicks, interface states, scrolling, pages, and alternative prompts.Screenshots were organized sequentially to reconstruct each flow, which served as the baseline when participants’ attempts were compared.Source: §3.1.2 · p. 5
3.1.3 Operationalizing Interaction Cost.
- Interaction cost counted minimum clicks and interface states, while researchers documented scrolls and potentially obstructive flow elements using dark-pattern categories including obstruction, sneaking, and forced action.Two researchers reviewed the flows, resolved disagreements through discussion, and the last author audited the final mappings.Source: §3.1.3 · p. 5
3.2.1 Participants and Recruitment.
- Researchers recruited 11 teenagers aged 14–17 who actively used at least two of the four study platforms, reached them through their parents because they were minors, and assigned each two platforms they regularly used.They also used snowball sampling through participating parents’ own networks to reach additional teenagers.Source: §3.2.1 · p. 5
3.2.2 Think-Aloud Task Procedure.
- Each session lasted approximately 45 to 60 minutes; participants completed privacy and safety tasks on two platforms using provided accounts with birth dates under 18, and shared their phone screens while thinking aloud.Source: §3.2.2 · p. 6
3.3.1 Unit of Analysis.
- The unit of analysis was one participant’s attempt at one task on one platform, beginning with the first task-directed interaction and ending at completion, abandonment, cancellation, or transition to the next task.Because participants completed multiple tasks on a platform, an attempt could start from the interface state left by the preceding task.Source: §3.3.1 · p. 6
3.3.2 Interaction Cost Measures; 3.3.3 Completion and Struggle Coding.
- Researchers coded clicks, screens, popups, layovers, total interface states, scrolls, completion, and struggle, distinguishing independent completion, completion after guidance, and noncompletion.Typing, scroll swipes, and accidental touches without visible responses were not clicks; automatic transitions and system prompts counted as states but not clicks.Source: §3.3.2 · p. 6, §3.3.3 · p. 7
3.3.3 Completion and Struggle Coding.
- Struggle coding distinguished sustained searching and stuck, assisted, or incomplete attempts from interaction cost: a long route could be easy, while a short route could be difficult.High struggle overlapped by definition with assisted or incomplete outcomes.Source: §3.3.3 · p. 7
3.3.4 Coder Alignment and Visual Codebook.
- Two researchers coded sequences from P5, P6, and P7, covering 27 of 88 attempts, and resolved differences in counts and struggle ratings to consensus.Alignment addressed screen and overlay boundaries, attempt start and end points, scrolls versus taps, researcher assistance, and moderate versus high struggle.Source: §3.3.4 · p. 7
- Exemplar clips were turned into GIFs for a visual codebook, and one researcher coded remaining recordings using finalized definitions, with new edge cases discussed before affected attempts were finalized.The GIFs illustrated clicks, scrolls, screen transitions, popups, layovers, and struggle levels.Source: §3.3.4 · p. 7
3.3.5 Comparison with Expert Task Flows.
- Participant paths were compared with same-task, same-platform expert paths for route variance and interaction cost, with deviations marked on task-flow figures.Notification management had no fixed endpoint, so its count reflected how far participants entered configuration rather than efficiency along a shared route; counts were descriptive, with no inferential claims about platforms or participants.Source: §3.3.5 · p. 7
4 Findings
The findings examine how protective settings and notification controls were organized across platforms, how interaction cost varied with the changes teenagers sought, and how navigation routes, fragmented controls, labels, and recurring dark patterns shaped those paths.
4.1.1 Expert Baseline.
- Basic daily limits required between four and eight clicks across three to seven screens, while customizing limits by day added steps and screens, particularly on TikTok and YouTube.The greater effort created an asymmetry between enabling a limit and tailoring it to different schedules.Source: §4.1.1 · p. 9
- Screen-time flows combined Obstruction through Adding Steps and Privacy Maze with Interface Interference; TikTok also showed False Hierarchy, and TikTok and YouTube used soft framing.The friendly illustrations and calm framing aligned with Emotional or Sensory Manipulation, while TikTok made entry into screen-time management clearer than routes to leave or disengage.Source: §4.1.1 · p. 9
4.1.2 Participant Performance.
- Screen-time route transfer sometimes worked, but differing labels and settings structures disrupted reuse and could lead teenagers through multiple settings areas to researcher guidance.Participants also encountered multiple routes to related controls within a platform, or struggled to distinguish adjacent controls with similar self-regulation goals.Source: §4.1.2 · pp. 9–11
4.2.1 Expert Baseline.
- Notification management was the most interaction-intensive task: TikTok required 40 clicks across 6 screens, Instagram 83 clicks across 22 screens, Snapchat 27 clicks across 6 screens, and YouTube 16 clicks across 3 screens.TikTok and YouTube kept more adjustments within a simple notification settings space.Source: §4.2.1 · p. 11
- Instagram and Snapchat distributed notification controls across categories, so full customization required repeated selections; granular options and enabled-by-default categories mapped to Choice Overload and Bad Defaults.Users seeking fewer alerts had to opt out repeatedly rather than use a consolidated control.Source: §4.2.1 · p. 11
- Long lists of visually similar controls without an overview made reviewed settings hard to identify, while socially meaningful cues around friends, messages, activity, and personalized content were coded as Social Engineering through Personalization.The evaluation described these cues as framing controls whose notification-disabling effort remained distributed across individual settings.Source: §4.2.1 · p. 11
4.2.2 Participant Performance.
- Participant paths ranged from leaving settings unchanged or briefly pausing notifications to selective or item-by-item configuration, making interaction cost conditional on configuration scope and persistence.P8 left Instagram’s existing configuration unchanged, while other routes entered category settings or used a temporary pause.Source: §4.2.2 · pp. 11–12
- Instagram’s temporary pause met P10’s stated need with little immediate effort, while persistent category-level customization required substantially more interaction.The pause bypassed the larger set of category-level decisions required for detailed configuration.Source: §4.2.2 · p. 12
- TikTok’s concentrated notification controls let participants make several changes without repeated category-screen movement, while detailed Instagram routes exposed fragmentation, Choice Overload, Adding Steps, and Hidden Information.Instagram’s item-by-item review required repeated movement between category screens; similar-looking screens made categories already reviewed or changed difficult to identify.Source: §4.2.2 · p. 12
4.3 Reporting Content
This section examines how users locate and navigate reporting and account-deletion controls, including the steps involved and how they interpret choices and post-submission outcomes. It describes variation in deletion routes and hesitation over deactivation versus permanent deletion, while noting that alternative prompts did not establish Trick Questions.
4.3.1 Expert Baseline.
- Across platforms, reporting required selecting an option, choosing a reason, and confirming; the multi-step flow was coded as Obstruction through Adding Steps, and entry controls were hidden behind gestures or menus.TikTok and Snapchat used long-press interactions, while Instagram and YouTube used a three-dot overflow menu.Source: §4.3.1 · pp. 12–13
- Instagram’s reporting flow added a post-submission choice about whether to keep seeing reported content, potentially leaving it visible and making the relationship between reporting and removal unclear.The expert evaluation coded this choice as Interface Interference through Hidden Information.Source: §4.3.1 · p. 13
4.3.2 Participant Performance.
- Finding the report control was a major hurdle: P4 and P5 searched without recognizing TikTok’s long-press entry, while access depended on a gesture or overflow menu rather than a persistent control.P4 proceeded only after interviewer guidance; other participants used different entry gestures, including Snapchat’s three-dot menu and YouTube’s long-press or overflow options.Source: §4.3.2 · pp. 13–14
- After exposing the control, participants generally followed reason-selection and confirmation screens more easily, although Instagram and YouTube required longer sequences with additional choices and information.YouTube’s route could include reason categories, sub-options, a timestamp, and a free-text description.Source: §4.3.2 · p. 14
- Participants encountered disappearance, continued visibility, or uncertainty after reporting, and the observations captured immediate interface responses rather than whether platforms acted on reports.P6 and P3 saw reported Snapchat content remain visible, while P2 on YouTube and P9 on TikTok saw content disappear; P11 was unsure whether a YouTube report had worked.Source: §4.3.2 · p. 14
4.4.1 Expert Baseline.
- Instagram’s expert-evaluated deletion route required 13 clicks across 11 screens, while TikTok’s path was short and Snapchat’s settings route was more direct but included identity confirmation.The three platform flows also required account verification before deletion could proceed.Source: §4.4.1 · p. 15
- Instagram’s deletion flow foregrounded less-permanent alternatives and included prompts and a 30-day waiting period that could be canceled by logging back in, requiring users to restart the process.The expert evaluation coded the delay as Forced Action and the alternative presentation as Bait and Switch, among other dark patterns.Source: §4.4.1 · p. 15
4.4.2 Participant Performance.
- In the supplied Instagram deletion case, P5 reached the delete option without a navigational hint but still encountered alternatives and reason-selection screens, describing the process as hard because of the many options.The attempt remained close to the expert route while retaining its Adding Steps.Source: §4.4.2 · p. 15
4.4.2 Participant Performance
- P2’s Instagram deletion route involved the highest click count among the Instagram deletion traces, and P2 hesitated between deactivation and permanent deletion at the point flagged for False Hierarchy and Visual Prominence.The less permanent alternative was presented prominently alongside deletion.Source: §4.4.2 · p. 15
- P7’s Instagram deletion attempt diverged most from the expert path, and P7 could not find the route from memory, needed a hint, and found that the TikTok route did not map directly to Instagram.Repeated searching and backtracking through plausible privacy and account settings extended the Obstruction identified in the expert baseline.Source: §4.4.2 · pp. 16–17
- P8 could not independently locate the deletion path and required step-by-step guidance; P11 searched settings and privacy-related pages before assistance, then had difficulty evaluating alternatives in the deletion sequence.Source: §4.4.2 · p. 17
- During otherwise successful Instagram deletion attempts, prompts for deactivation, data download, and reasons for leaving lengthened the path and corresponded with Sneaking and Bait and Switch.Reason-selection traces alone did not meet the threshold for Trick Questions, which required consistent participant misunderstanding or redirection caused by the wording.Source: §4.4.2 · p. 17
- TikTok and Snapchat attempts generally stayed closer to their shorter expert baselines, with participants finding deletion controls through relatively compact routes.TikTok participants made fewer exploratory detours; Snapchat attempts were similarly compact, though the expert flow still included identity confirmation.Source: §4.4.2 · p. 17
4.5 What the Expert Baseline Does and Does Not Predict
The expert baseline models a route for an evaluator who already knows the control’s name and entry point, but participant cost also reflects recognition. Across tasks, costs often rose above that reference, while low counts could mark incomplete configuration rather than efficient completion.
4.5 What the Expert Baseline Does and Does Not Predict
- The expert baseline measures distance from the home screen for an evaluator who already knows the control’s name, goal, entry point, and route; it does not measure route recognizability.Source: §4.5 · p. 17
4.5.1 Participant Interaction Is Often Higher, But Not Always.
- Screen-time and deletion tasks exceeded the expert baseline on every platform; TikTok deletion averaged 10.6 clicks versus 5, Instagram deletion 19.7 versus 13, and YouTube screen time 13.0 versus 4.Source: §4.5.1 · p. 17, Table 3 · p. 18
- Notification attempts fell below baseline on all four platforms, but Instagram participants averaged 4.6 clicks against 83 required to fully disable notifications, reflecting attempts that could stop short of the required configuration.Source: §4.5.1 · pp. 17–18
4.5.2 Range Is Shaped By Recognition.
- TikTok deletion averaged 10.6 clicks against a baseline of 5, with every attempt between 9 and 12; Snapchat reporting averaged 11.3 across a range of 5 to 24 as route recognition varied.Source: §4.5.2 · p. 18
- A narrow range marks cost imposed on every participant, while a wide range signals variation in how legible the route was.Source: §4.5.2 · p. 18
5 Discussion
The discussion argues that settings’ presence does not ensure usable or durable protection for teenagers. It frames a preliminary wayfinding audit as a way to examine whether users can find controls, configure them protectively, and interpret outcomes.
5 Discussion
- The authors conclude that settings teenagers are expected to rely on fail basic usability grounds amid dark patterns, insufficient protective defaults, and limited understanding of settings.Source: §5.1 · p. 19, §5 · p. 19
5.1 Moving from Feature Availability to a Protected Experience
- A setting offers usable protection only when teenagers can recognize and locate it, understand its action, and complete a protective configuration.Source: §5.1 · p. 19
- Unfamiliar terms, hidden entry points, confusing endpoints, and engagement-preserving options that are easier to reach can impede protective use.Source: §5.1 · p. 19
- A short interaction may reflect retaining a default, choosing a temporary alternative, or stopping early, so evaluation must examine the endpoint and resulting state, not just action counts.Source: §5.1 · p. 19
5.2 Beyond Usability: A Preliminary Wayfinding Audit
- The preliminary wayfinding audit combines usability sessions, expert evaluation, interaction-cost comparison, and dark-pattern analysis to assess effort, route barriers, and outcomes together.Source: §5.2 · p. 20
- The audit is demonstrated only on the examined platforms and tasks; other contexts may need a different method balance, and future work includes testing versions and refining task-flow alignment with TADP.Source: §5.2 · pp. 20–21
6 Implications and Future Work
The sessions suggest that teenagers can learn about privacy controls when they encounter them, while the controls are often not encountered in ordinary use. The authors propose testing how controls are surfaced and whether the wayfinding audit transfers beyond the studied teenagers and platforms.
6 Implications and Future Work
- The evaluation approach could extend to operating systems, games, and other applications, including users with different prior knowledge and expectations about where controls are found.The authors note that measures and comparison are likely to transfer beyond the studied population and platforms; reconstructed routes can be assessed wherever protective actions sit behind nested navigation.Source: §6 · p. 21
- The study involved one session per participant, and the usability tests were not designed as interventions.The authors describe this as a boundary on what the sessions establish.Source: §6 · p. 21
- Participants encountered unfamiliar features, interpreted settings they had ignored, and sometimes named features they intended to adopt after sessions that functioned as incidental privacy education.One participant discovered “download my data,” accurately described its purpose within a minute, and another called the session “educative.”Source: §6 · p. 21
- Future work could test whether surfacing controls before a moment of need, or where they are relevant, reduces the searching identified as costly.Situated encounters could be placed within platforms, schools, or family guidance tools.Source: §6 · p. 21
7 Conclusion
Across four platforms, most of the 11 teenagers’ attempts cost more than the documented routes, with excess effort accruing before route entry; notification attempts could cost less when participants stopped short of the required configuration. The conclusion proposes a wayfinding audit that checks whether controls can be found and their resulting state verified.
7 Conclusion
- Most attempts exceeded the documented route cost, with excess effort accruing before participants entered the route, while notification attempts cost less when participants stopped short of the required configuration.The study paired expert reconstruction of six task flows across TikTok, Instagram, Snapchat, and YouTube with think-aloud sessions involving 11 teenagers.Source: §7 · p. 21
- Dark pattern strategies recurred on every examined platform, but routes to protective controls shared only the goal, so knowledge of how to reach a control did not transfer across services.The conclusion distinguishes transferable knowledge of what raises interaction cost from platform-specific route knowledge.Source: §7 · p. 21
- Evaluating a protective control requires checking whether its route can be found and whether the resulting state can be verified.The wayfinding audit is presented as an initial articulation and demonstration of a method for evaluating a protected experience.Source: §7 · p. 21
- The audit integrates interface design, dark patterns, interaction cost, observed use, and the durability of a protected experience with established HCI methods.The conclusion presents this integration as the contribution of the wayfinding audit.Source: §7 · pp. 21–22
A. Interaction Metrics
The appendix defines interaction metrics for coding teenagers’ task attempts, including interface states, clicks, and scrolling. The definitions distinguish interface components and actions for reporting or counting.
A. Interaction Metrics
- The interaction metric definitions specify what to count and exclude, including scrolling, typing, or accidental touches without a visible response.Table A1 presents the operational definitions used to code teen task attempts.Source: Table A1 · p. 25, A. Interaction Metrics · p. 26
- A click is a discrete tap that triggers an interface response; a tap opening a popup or layover and a later selection each count separately.Scrolling counts continuous swipe gestures that visibly move content, with repeated swipes counted separately.Source: A. Interaction Metrics · p. 26
- Total interface states are the sum of coded full screens, popups, and layovers, with component counts reported as S, P, and L where space permits.Layovers remain a separate state type rather than being collapsed into popups.Source: A. Interaction Metrics · p. 26
B. Completion
The appendix identifies a separate set of completion-coding definitions and describes how hesitation is recorded during task attempts.
B. Completion
- A touch that does not move content or a tap used to select an item is distinguished from a swipe that moves content.The coding guidance treats each continuous content-moving swipe as one scroll.Source: B. Completion · p. 26
- Completion coding is addressed separately from interaction metrics, while hesitation is coded through struggle ratings and, where available, notes on repeated scrolling, reversal, rereading, or pauses.Table A2 is identified as the completion-coding definitions used to assess teen task performance.Source: Table A2 · p. 26, B. Completion · p. 26
C. Struggle Level
Table A3 characterizes task difficulty through low, moderate, and high struggle levels tied to observable participant behaviors.
C. Struggle Level
- Low struggle involves minor hesitation followed by smooth, independent progress, with brief pauses or a few extra actions followed by quick recovery.Source: C. Struggle Level · p. 27
- Moderate struggle involves sustained searching, repeated scrolling or backtracking, and wrong turns, but the participant independently recovers and reaches the endpoint.Source: C. Struggle Level · p. 27
- High struggle applies when participants become stuck, need navigational assistance, abandon the attempt, or fail to reach the endpoint.Observable indicators include requests for help, researcher redirection, prolonged unsuccessful search, or task cancellation.Source: C. Struggle Level · p. 27
D. Detailed Task Performance by Participant, Platform, and Task
This section documents participant task performance across platforms and tasks using completion, struggle, time, clicks, scrolls, and interface-state records. It summarizes varied outcomes, including independent completions, unchanged controls, and attempts involving researcher assistance or unavailable paths.
D. Detailed Task Performance by Participant, Platform, and Task
- The table distinguishes full-screen transitions, popups, and layovers, and defines high struggle as needing navigational assistance or failing to reach the endpoint.Interface state totals equal S + P + L; completion is categorized as independent, assisted, or not completed.Source: D. Detailed Task Performance by… · p. 27, Table A4 · p. 27
- A TikTok screen-time trace located the control but paused to distinguish the daily limit from screen-time breaks before setting it.The task was completed independently in 1 min 10 sec.Source: D. Detailed Task Performance by… · p. 29
- A TikTok account-deletion trace was completed independently after the participant chose an alternative reason to avoid additional steps.The recorded completion time was 1 min 52 sec.Source: D. Detailed Task Performance by… · p. 29
- Screen-time tasks ranged from direct or transferred navigation to extended settings searches that required researcher guidance.One participant carried the route from an Instagram task, while another expected YouTube’s structure before locating the control.Source: D. Detailed Task Performance by… · pp. 29–30
- Participants independently toggled private-account controls in some attempts, while notification attempts included category changes, retaining defaults, or making no changes.One Snapchat data-download attempt found the control by scrolling within settings.Source: D. Detailed Task Performance by… · pp. 29–30
- Instagram account deletion repeatedly required researcher assistance after participants could not locate the deletion path.Recorded attempts included step-by-step guidance and completion with assistance.Source: D. Detailed Task Performance by… · pp. 29–30
- Some TikTok deletion attempts were completed independently, whereas another required guidance after an extended settings search.A Snapchat attempt also located account deletion independently.Source: D. Detailed Task Performance by… · p. 29
- Reporting outcomes varied: one Instagram attempt could not find a report option and was not completed, while other records show completed overflow-menu or long-press flows.One completed report flow still left the participant uncertain about its outcome.Source: D. Detailed Task Performance by… · pp. 29–30