Source-linked AI summary
I am Definitely Manipulated, Even When I am Aware of it. It s Ridiculous! -- Dark Patterns from the End-User Perspective
Kerstin Bongard-Blanchy, Arianna Rossi, Salvador Rivas, Sophie Doublet, Vincent Koenig, Gabriele Lenzini
TL;DR
The paper examines whether transparency and awareness of dark patterns are sufficient for users to control or resist manipulative online influence. It relates awareness, detection, and susceptibility to manipulation, finding that recognition is possible but awareness alone does not significantly predict resistance, motivating varied interventions.
Problem
The paper asks whether users’ awareness and detection of dark patterns enable them to control or resist their influence.
Method
The study relates users’ awareness, ability to detect dark patterns, and likelihood of being influenced.
Results
Users are generally aware of and capable of recognising manipulative designs, but awareness does not significantly predict their ability to resist them.
Takeaways & Limitations
Effective responses should combine interventions that raise awareness, facilitate detection, strengthen resistance, or eliminate dark patterns across users and environments.
Takeaways & Limitations
Because participants came from an online platform and were probably more accustomed to online designs, the findings might overestimate awareness in less tech-savvy UK populations.
Abstract
from arXiv · showhide
Online services pervasively employ manipulative designs (i.e., dark patterns) to influence users to purchase goods and subscriptions, spend more time on-site, or mindlessly accept the harvesting of their personal data. To protect users from the lure of such designs, we asked: are users aware of the presence of dark patterns? If so, are they able to resist them? By surveying 406 individuals, we found that they are generally aware of the influence that manipulative designs can exert on their online behaviour. However, being aware does not equip users with the ability to oppose such influence. We further find that respondents, especially younger ones, often recognise the "darkness" of certain designs, but remain unsure of the actual harm they may suffer. Finally, we discuss a set of interventions (e.g., bright patterns, design frictions, training games, applications to expedite legal enforcement) in the light of our findings.
1 INTRODUCTION
Dark patterns are manipulative design choices that steer users toward decisions they would not make when fully informed, potentially harming individual and collective welfare. The study examines whether users lack awareness, fail to recognise these designs, or cannot resist them despite awareness, with findings intended to guide countermeasures.
- Online services deploy manipulative practices at low cost and large scale to influence purchases, increase time spent online, and encourage privacy-invasive consent.These practices operate in dynamic, interactive, intrusive, and adaptive environments.
- Dark patterns coerce, steer, or deceive users into choices they would reject if fully informed and able to select alternatives.They can provide convenience or immediate gratification while undermining privacy, finances, behaviour, competition, and consumer trust.
- The study tests whether dark patterns exploit users’ lack of concern, inability to recognise them, or inability to resist them despite awareness and recognition.The aim is to identify requirements for countermeasures targeting individuals or their surrounding environment.
- Potential countermeasures include warnings that increase risk salience, friction that disrupts automatic behaviour, and stronger environmental protections such as company fines.
- Users can recognise dark patterns but are only vaguely aware of the concrete harm they entail; recognition is also positively related to self-protection ability.People under 40 and those with education beyond high school are more likely to recognise them.
2 RELATED WORK
Related work documents dark patterns across online services and domains, explains their effectiveness through cognitive limitations, and studies user recognition, stakeholder attitudes, and possible interventions. It also highlights unresolved boundaries between manipulative and admissible designs.
- Researchers have catalogued dark patterns across online services and domains, including privacy, e-commerce, video games, automated systems, and home robots.Collections support awareness, alternative designs, and training corpora for detection algorithms.
- Clear boundaries remain unsettled between inadmissible manipulation and admissible digital nudges that support praiseworthy goals.
- Dark patterns exploit cognitive biases and bounded rationality, including status quo bias, bandwagon effects, hyperbolic discounting, and optimism bias.These limitations can direct users toward choices they may later regret or cause them to underestimate their vulnerability.
- Prior recognition studies find that informing participants about possible dark patterns improves detection, while subtle patterns are more easily missed than aggressive ones.
- Design practitioners, regulators, consumer organisations, and researchers have expressed concerns about dark patterns and examined related ethical conflicts.
- Proposed interventions include privacy-friendly bright patterns, procedural training, browser-based detection, and regulatory tools to prohibit or fine manipulative practices.Bright patterns modify the choice environment by making privacy-protective options more salient.
3 RESEARCH GAPS AND RESEARCH QUESTIONS
The paper asks whether users are aware of, can recognise, and can resist manipulative interface designs, while examining factors that may shape influence. It treats transparency as potentially insufficient because opposing manipulation can carry cognitive and practical costs.
- The paper identifies a broad intervention challenge because dark patterns range from coercive restrictions to subtle visual nudges, preventing one universal solution.
- RQ1 asks whether users are aware of and concerned about manipulative interface designs’ influence on their behaviour.
- RQ2 asks whether users are able to recognise manipulative interface designs, in light of prior claims about dark-pattern blindness.
- Transparency may enable resistance but may not suffice because resignation, service benefits, and the cognitive costs of opposition can undermine resistance.
- RQ3 asks whether users are likely to be influenced despite being aware, concerned, and capable of recognising manipulative designs.
- The study also examines whether education, age, and online-service use frequency are associated with awareness, recognition, and influence.
4.1 Study design
The study uses an online survey that measures participants’ mindsets, behaviour, demographics, and recognition of redesigned interfaces containing dark patterns. Its three parts proceed from general attitudes to reported behaviour and then specific detection tasks.
- Participants completed an online LimeSurvey survey administered through Prolific, with demographic data collected alongside the three substantive parts.Questions were mandatory except for a final feedback field.
- The first part measured awareness and concern using paired general-versus-personal statements rated on a 5-point Likert scale.Affirmative or undecided responses prompted examples of perceived influence, harm, and worries.
- The second part measured online-service exposure and participants’ usual behaviour in eight situations involving common manipulative strategies.Item order was randomised and phrasing was kept neutral.
- The survey used both “influence” and “manipulation” terminology to reduce negative priming while keeping questions relevant to dark patterns.
- The third part evaluated recognition using ten randomly ordered, uniformly redesigned interfaces, including one control without a dark pattern.The other nine examples represented financial, privacy, and time-and-attention harms.
- Interfaces were shown for 10 to 40 seconds, after which participants reported noticed influence, perceived susceptibility, and acceptability.Time limits and advance notice that some examples lacked manipulative elements constrained excessive searching.
4.2 Participants
The study analyzed 406 participants after excluding seven unusable responses. The sample was recruited to represent the UK population by age, gender, and ethnic origin, with broad age and education coverage.
- 413 responses were collected, and seven gibberish responses were excluded, leaving 406 participants for analysis.
- Participants ranged from 18 to 81 years old, with a mean age of 45.2 years and standard deviation of 15.5.
- Education levels ranged from high school or lower to postgraduate study, with 106, 236, and 64 participants in these categories respectively.
- Sixteen pre-test participants helped improve question comprehensibility and reduce survey duration to a maximum of 30 minutes.
4.3 Ethical and Legal Considerations
The study followed institutional and professional research-ethics requirements, obtained prior ethics-board authorization, and collected responses anonymously without identifying information.
- The study adhered to the University of Luxembourg’s research-ethics guidelines and the European Federation of Psychologists’ Associations’ code of ethics.
- The University’s Ethics Review Board authorized the study before data collection began.
- Survey responses were anonymous, and questions did not request information that could identify participants.
4.4 Data analysis
The analysis combined descriptive and inferential statistics with coded evaluations of dark-pattern recognition. Regression models examined detection and influence while controlling for relevant participant characteristics.
- Awareness ratings were summarized using means, medians, and modes, while a two-sided sign test assessed personal-versus-general awareness differences.
- Open responses were deductively coded as 0, 0.5, or 1 according to whether participants failed, partly, or fully identified each manipulative design element.
- Dark-pattern detection scores were summed for each participant, producing totals ranging from 0 to 9.
- An OLS regression controlled detection outcomes for age, educational level, online-service use frequency, and disposition to be influenced by online designs.
- A second OLS regression estimated associations between influence likelihood and awareness, dark-pattern detection, acceptability, and demographic data.
5 RESULTS
Participants generally recognised that online designs can influence behaviour, yet remained uncertain about personal harm and resistance. Dark-pattern recognition varied by design and participant characteristics, while awareness alone was not associated with lower influence likelihood.
- People’s awareness of the influence of online designs on their choices and behaviour: Participants recognised that online designs can influence their choices and behaviour, especially content consumption and service selection.They associated influence mainly with personalised content, recommendations, and special offers.
- People’s awareness of the influence of online designs on their choices and behaviour: Participants were uncertain whether manipulative designs could harm them and were undecided about worrying, despite naming psychological, physical, and financial harms.They were more concerned about vulnerable people than themselves.
- People’s ability to detect dark patterns: 59% of participants identified five or more of nine dark patterns, while one fourth recognised seven, eight, or all nine interfaces.Recognition varied by pattern: trick questions, pre-selection, loss-gain framing, hidden information, and bundled+forced consent were recognised by half or fewer, whereas high-demand or limited-time messages and confirmshaming were recognised by a majority.
- People’s ability to detect dark patterns: Younger participants and those with education above high school were more likely to recognise dark patterns.Regression results showed higher detection among Millennials/Gen Y and Zoomers/Gen Z than the older Baby Boomer+ reference group, while education above Bachelor was not additionally associated with detection.
- People’s ability to detect dark patterns: Recognition was positively correlated with awareness of manipulative influence, although participants reported surprise when they missed designs they later found obvious.Participant accounts described feeling manipulable even when they considered themselves aware.
- People’s likelihood to be influenced by dark patterns: People who detected dark patterns more easily reported a slightly lower likelihood of being influenced, whereas awareness itself was not a significant predictor.Participants who considered manipulative designs more acceptable reported being slightly more influenced, and disposition to manipulation was positively associated with influence likelihood.
6 DISCUSSION
The discussion interprets dark-pattern responses through intervention goals: raising awareness, facilitating detection, strengthening resistance, and removing manipulative designs. Findings motivate user- and environment-level measures, while highlighting unequal recognition across designs and generations.
- 6.1 Raising awareness: Users generally recognise that digital services can influence them, but often cannot identify the concrete harms caused by manipulative designs.The paper links this gap to insufficient motivation for protective actions and recommends making risks more salient.
- 6.2 Facilitating detection: Recognition varies substantially by design: confirmshaming and scarcity messages were often recognised, whereas deception, pre-selection, and forced consent were rarely recognised.The authors caution that these results concern specific implementations rather than entire dark-pattern categories.
- 6.4 Eliminating dark patterns: Gamified training could strengthen motivation to recognise dark patterns in real settings, while automated tools could help watchdogs identify and classify potentially illegal practices at scale.Automated approaches require a large pool of reliable data because some manipulative attempts are challenging even for humans.
- 6.3 Bolstering resistance: Higher recognition was associated with lower reported likelihood of influence, suggesting that detection and self-protection are related capacities.The survey disclaimer also encouraged counterfactual thinking and more reflective information processing.
- 6.3 Bolstering resistance: Resistance costs depend on whether a design is coercive, nudging, or deceptive, with forced consent potentially leaving users a take-it-or-leave-it choice.Accordingly, interventions may need to combine bright patterns, friction designs, education, and stronger environmental protections.
- 6.5 Targeting interventions - older vs younger generations: Older adults were less able to recognise manipulation and less aware of their own susceptibility, complicating the design of targeted safeguards.The paper identifies age below 40 and education above high school as thresholds associated with recognition, but reports no significant age correlation with influence likelihood.
7 LIMITATIONS
The study’s generalisability is constrained by its online UK sample and by self-reported influence likelihood. The detection task also differed from ordinary real-world use, and respondents sometimes conflated manipulative design with manipulative content.
- Sample and generalisability: Because Prolific participants may be more accustomed to online designs than the average UK population, the findings might overestimate awareness among less tech-savvy users.The authors also call for research in other countries.
- Measurement: Self-reported likelihood of influence is only an approximation and may not reflect participants’ actual behaviour.The authors therefore invite further research using measures closer to observed behaviour.
- Task validity: Participants sometimes cited manipulative content rather than manipulative design, so reported awareness and concern may be lower than the results indicate.The detection activity also explicitly asked users to search for manipulation, which does not entirely match real-world use.
8 CONCLUSIONS
The study finds that users recognize manipulative designs and their potential influence, but awareness alone does not sufficiently protect them. It therefore points toward coordinated design, technical, educational, and regulatory interventions, alongside further user-perspective research.
- Individuals are aware of manipulative designs’ potential influence and are relatively capable of recognising them, but awareness alone does not shield them from dark patterns.
- The discussion proposes design, technical, educational, and regulatory measures to heighten awareness, ease detection, strengthen resistance, or eliminate manipulative practices.
- Design frictions, bright patterns, automated detection applications, assessment tools, economic incentives, and regulatory solutions are identified as interventions warranting further investigation.
- Future work should combine established dark-pattern attributes with user perspectives to identify designs perceived as unrecognisable, irresistible, or unacceptable.
- Studying user perceptions may help establish which manipulative practices end-users consider legitimate or illegitimate.