No standardized questionnaire currently incorporates an ethical dimension for assessing User Experience (UX). We explored how the ethicality of interface design is reflected in current UX metrics and how they could be extended. To this end, we adapted the User Experience Questionnaire (UEQ) and enriched it with supplementary items specifically designed to capture user responses on social media to unethical interface designs, commonly referred to as 'dark patterns'. Through an exploratory analysis of a survey involving 120 participants who evaluated a selection of 15 social media dark patterns, we found preliminary evidence that (1) an aggregated UX score using items from the User Experience Questionnaire does not efectively indicate unethical user interface design. Instead, (2) subscale measures from the questionnaire show a relationship with unethical design. Furthermore, extending the User Experience Questionnaire seems promising, as (3) users can identify interfaces with addictive and pressuring properties, and (4) evaluations demonstrate consistency within groups of unethical design strategies.
User experience (UX) designers are responsible for designing all aspects of user interfaces [
Meanwhile, there is an emerging consensus that social media companies engage in unethical
design practices [
This paper initiates a refocus of user experience metrics to evaluate all facets of users’ perceptions, explicitly including those with potential for harm, as delineated by ISO 9241210 [12].
Therefore, we made a first attempt to develop a quantitative method to assess users’ ethical perception of dark patterns in social media. In pursuit of this, we adapted the UEQ and enriched it with supplementary items specifically designed to capture user reactions to unethical interface designs.
In the subsequent sections of this workshop paper, we explore existing UX metrics and examine how dark patterns are assessed in current literature. Next, we describe the process of developing an adapted UEQ incorporating an ethical dimension, informed by emerging taxonomies classifying unethical interface design strategies in social media, known as dark patterns. Following that, we present key findings from the survey and discuss their implications in the last part.
The literature outlines various methods for assessing user experiences, highlighting diverse approaches to identifying software weaknesses. Methods range from IsoMetrics[13] and AttrakDif2 [14] to others, each with unique focuses. The Microsoft Desirability Toolkit simplifies capturing subjective aspects like "fun" and "enjoyment" through techniques like the Faces Questionnaire and Product Reaction Cards, yielding qualitative feedback[15]. In contrast, the System Usability Scale (SUS) ofers a quantitative perspective by scoring a system’s usability on a scale from 0 to 100 based on user responses to ten statements[16]. Though alternative methods exist, such as emotion recognition for evaluating dark patterns[17], surveys remain the predominant tool for gathering user experience data. This is evidenced by well-established methods like SUMI, which assesses usability through 50 statements[18], and the once-popular AttrakDif2, an inspiration for the widely used User Experience Questionnaire [19, 20].
The UEQ is designed to rapidly assess users’ subjective perceptions of product features and their immediate efects, using 26 items across six scales to capture a broad view of user experience [ 14]. These scales assess both hedonic qualities—Stimulation and Novelty, which focus on engagement and innovation—and pragmatic qualities—Eficiency , Perspicuity, and Dependability, which relate to task accomplishment. Attractiveness evaluates the overall product impression, synthesizing insights from both hedonic and pragmatic dimensions [21].
Users rate their agreement with paired adjectives on a scale from -3 to +3, indicating their experience. The UEQ-S, a concise version of the full questionnaire, contains eight items, four items from the hedonic scales and four from the pragmatic scales [21].
User Experience Design is about making business objectives and user goals work together [22].
Key Performance Indicators (KPIs) or Objectives and Key Results (OKRs) measure the move to
achieve those business goals, and UX measures are established, as such KPIs that can measure a
user interface designs UX performance holistically [
In the realm of technology design, ethical considerations present complex challenges [25]. The influence of dark patterns on user decision-making has been well-documented, illustrating how certain design choices can manipulate user consent, as seen in studies on cookie banner consent designs [26]. However, the assessment of tech ethics remains fragmented, with current methodologies often leading to qualitative insights.
Although instruments like the Ethical Climate Questionnaire exist to assess organizational ethics [27] and tools like the Facebook Addiction Questionnaire gauge individual platform dependency [28], these do only yield data on business organization and individual human long-term efects. More concretely, some researchers have proposed ’moral cards’, designed for designers to use as a reflective tool on adhering to ethics by design [ 29]. In the field of artificial intelligence, considerable eforts have been made to incorporate ethics, with many ethics frameworks emerging. Although checklists, audits, and fairness metrics have been proposed to encourage AI developers towards responsibility, a review indicates that these tools investigate ethics narrowly, focusing on a few dimensions rather than examining all aspects holistically [30]. This underscores a significant research void: the lack of tools specifically designed to evaluate the ethical aspects of user interfaces.
Our initial endeavor seeks to bridge this gap by proposing a method that integrates ethical considerations into user experience assessments. Despite a thorough literature search using comprehensive keywords, we found no existing instruments tailored for assessing the ethicality of user interfaces. This discovery confirms the need for innovative approaches to incorporate ethical evaluation into UX design practices, marking a pivotal direction for future research.
A few researchers have studied unethical design practices in social media [
In contrast, Bongard-Blanchy et al. [33] conducted a survey of whether users can recognize dark patterns in diferent domains. The participants were asked to elaborate in an open text ifeld on whether they had seen manipulative design elements presented to them for a duration from 10 to 40 seconds. One of the ten mockups served as a control condition and did not contain any dark pattern. The findings indicate that 59% of the participants successfully recognized five or more dark patterns among the nine interfaces. Furthermore, the results highlight that some dark patterns, such as high-demand / limited time message and confirmshaming, are easier to detect than others like hidden information[33].
Monge Rofarello and De Russis [
To study how unethical interface design materializes in user experience metrics and could be quantified through them, we identified 15 social media dark patterns, reported in the literature. These were then used to stimulate an extension to the User Experience Questionnaire, translated into mockups and evaluated using an online survey.
To advance standardized tools for measuring unethical user experience, we base our work on a
set of social media user interfaces identified as unethical. These interfaces were selected from
a literature review of academic and governmental dark pattern taxonomies. We integrated
ifndings from two peer-reviewed papers [
The dark patterns and their associated strategies are listed in Figure 1.
The UEQ2 was adapted to include considerations of ethical and unethical design. This adaptation was conducted through iterative sessions among the authors, focusing on how users might evaluate the identified unethical design patterns using the full UEQ and what additional associations might arise.
Initially, we reviewed the 26 adjective pairs of the UEQ to determine their applicability in evaluating the 15 dark pattern instances we had identified. Subsequently, we took the complete UEQ and assessed which items might elicit strong opinions from participants—specifically, 1https://osf.io/nw2tj/ 2https://www.ueq-online.org/ Nagging Forced Access Overcomplicated Process Hindering Account Deletion False Hierarchy Trick Wording Toying With Emotion Expectation Mismatch Sneaky Bad Defaults Gamification Social Pressure Social Connector Endlessness Pull To Refresh Content Customization
Unethical Design Strategy Nagging Forced Action Obstruction Obstruction Interface Interference Interface Interference Interface Interference Sneaking Sneaking Social Engineering Social Engineering Social Engineering Social Engineering Social Engineering
Social Engineering which items might be rated as either -3 or +3. These two steps resulted in two lists of UEQ items, which were then merged. In this process, we found the Novelty scale to be unnecessary for evaluating users’ perceptions of dark patterns. Given that the resulting list was too lengthy for practical use, we favored items from the short version of the UEQ (UEQ-S) and omitted additional items from the same UEQ categories.
Upon re-evaluating our questionnaire, we identified a lack of items addressing users’ perceived manipulation and the techniques implied by dark patterns. To explore these aspects without directly inquiring about manipulation and perceived ethics, we adhered to the UEQ’s structure of adjective pairs. Leveraging the Product Reaction Words from the Microsoft Desirability Toolkit, we revisited the 15 dark pattern mock-ups to pinpoint descriptive gaps. Consequently, we introduced four new items with both positive and negative adjectives: pressuring / suggesting, addictive / non-addictive, covert / transparent, and deceptive / benevolent. These introduced four items are treated as distinct scales in the calculations but are collectively labeled as Manipulation in the box plots. Thus, our final questionnaire comprises 14 items, with 10 of them combined into 5 two-item scales and 4 newly added stand-alone item scales. Four scales—Perspicuity, Eficiency , Dependability, and Stimulation—fully represent the UEQ-S. The Attractiveness scale is a two-item scale utilizing adjectives from the original UEQ’s Attractiveness scale.
Finally, the new items were translated into German, as the survey was conducted in both English and German to reach a larger audience. We included a "don’t know / not applicable" option and randomized the item order, arranging the pairs so that half start with the positive term and the other half with the negative term. The adapted UEQ is presented in Figure 2, reflecting the layout seen by participants.
The questionnaire for investigating the user experience of unethical user interface design was implemented in an online survey. The survey involved additional questions about social media habits and problems. This paper focuses only on the evaluation of the dark patterns in social media and excludes all non-social media users. The survey is available as an appendix on OSF 3. We conducted three test runs, with the survey, before releasing it. As answering our questionnaire for 15 diferent dark patterns is time-consuming and overwhelming, only five randomly selected patterns were shown to each user. Users were recruited through online posts, on a survey exchange platform, and within our network. 4
We employed exploratory data analysis to determine whether and how (1) user experience questionnaires can indicate unethical design, and (2) our proposed extension is necessary for such identification. We refrained from conducting any confirmatory data analyses. First, data was formatted, cleaned and checked for any suspicious responses, such as flatlining. Using RStudio5, we created various data visualizations. Histograms were used to observe whether distributions appear normal. For space considerations and ease of analyzing higher order patterns, color coded box plots showing mean, median, quartiles, whiskers, and outliers were favored in the presentation of this paper. In these plots, data from the Likert scale is treated as an ordinal approximation of a continuous variable [34, 35].
We arranged the 15 dark pattern user experience visualizations on a digital whiteboard and visually clustered them to identify characteristics in the user perceptions.
Additionally, we computed averages for the new manipulation items (pressure, addiction, covert, deception) and UEQ scales (attractive, dependability, eficiency, perspicuity, stimulation), along with higher-order scores for hedonic and pragmatic qualities, as intended by the UEQ-S, and an overall UEQ-S score. A simple heuristic suggested by Schrepp et al. [21, Fig. 3] was used to interpret the quantified user experience from Likert scale data ranging from -3 to +3 as positive (≥ 1), neutral (between -1 and 1), or negative (≤ -1).
Lastly, we computed correlation matrices to assess the interrelationships among these UX measures. Given that visual clustering suggested potential relationships between patterns, we calculated correlation scores within and between the diferent strategies.
A total of 126 social media users filled out the survey completely and were included in this analysis. On average, each dark pattern was evaluated by 42 social media users. The results from the explorative data analysis will focus on the most relevant findings. The data will be made available upon request. 3https://osf.io/nw2tj/ 4The platform https://www.soscisurvey.de/ was used for creating and hosting the survey. The survey was posted in Facebook groups, as well as https://www.surveycircle.com/ 5https://posit.co/download/rstudio-desktop/
The overall UEQ-S score, obtained by combining the surveyed UEQ-S scales in our adapted questionnaire (dependability, eficiency, perspicuity and stimulation), ranged from a maximum of 0.58 to a minimum of -1.05 with an average of -0.16 (see Table 1). The mean scores across all scales, including the extra items from our adapted questionnaire, ranged from a maximum of 0.48 to a minimum of -0.93 with an average of -0.30. Applying the heuristics as described in subsection 3.4, only the dark pattern of Nagging resulted in an overall negative user experience, as defined by the UEQ. The meta-level concept of Hedonic Quality, influenced by Stimulation, and meta-level concept of Pragmatic Quality, influenced by Dependability, Eficiency, also did not yield a negative user experience, except for Nagging.
In our questionnaire consisting of 5 original scales and 4 adapted pairs of items, the scales received varied evaluations—some predominantly negative, many neutral, and a few positive. Notably, our newly proposed items ’pressuring - suggesting’ (Max -0.03, Min -2.25, Avg -1.08) and ’deceptive - benevolent’ (Max 0.07, Min -2.09, Avg -1.03) were ranked as the most negative. Patterns such as Forced Access, Trick Wording, Toying With Emotion, Gamification, Social Pressure, and Social Connector were negatively ranked in terms of pressuring. Additionally, False Hierarchy and Nagging were evaluated as very pressuring (≤ -2).
Further, Nagging, False Hierarchy, were evaluated as very ’deceptive’ and Forced Access, Expectation Result Mismatch, Trick Wording, Toying With Emotion as ’deceptive’. The ’Attractiveness’ scale, which included items from the long UEQ, matched our ’deception’ item in terms of negative evaluations, with Toying With Emotion being assessed as strongly negative.
Patterns Endlessness and Pull To Refresh were perceived by users as addictive. Trick Wording, Forced Access and Nagging were evaluated with negative dependability.
The scale Eficiency was neutral for all patterns, and Stimulation neutral for all, but Nagging.
Conversely, some patterns received positive evaluations. Social Pressure, Sneaky Bad Defaults, and Pull to Refresh were rated positively in terms of ’Perspicuity’. Sneaky Bad Defaults and Pull To Refresh as revealed. Moreover, Trick Wording and Sneaky Bad Defaults were perceived as non-addictive.
Visual clustering suggested that the user experience evaluations of patterns within the social engineering strategy were strikingly similar. This was substantiated by a correlation matrix, indicating that these patterns are indeed correlated, many of them strongly so, with an average correlation coeficient of 0.69. However, they did not show a correlation with patterns from other strategies. Similarly, the three patterns associated with the strategy of interface interference were correlated, yielding an average correlation coeficient of 0.85 (see Figure 4).
Furthermore, notable correlations were observed among various scales. The ’deception benevolent’ scale was strongly correlated with ’attractiveness’ (coeficient 0.93), as also observed in subsection 4.2. Additionally, ’addiction’ showed a moderate negative correlation with ’attractiveness’, ’perspicuity’, and the overall UEQ score.
Concluding, from the explorative data analysis, we find that (1) an aggregated UX score using items from the User Experience Questionnaire does not efectively indicate unethical user interface design. Instead, (2) subscale measures from the questionnaire show relationships with unethical design strategies. However, even considering subscales, it appears to be necessary to extend User Experience Questionnaires focusing only on Pragmatic and Hedonic Qualities for an ethical dimension as (3) users negatively identify interfaces with addictive and pressuring properties. Finally, (4) evaluations demonstrate consistency within groups of unethical design strategies, indicating that these may remain stable.
Currently, User Experience Questionnaires are employed to evaluate designs and in industry
often to measure the success of designers. However, aggregated measurements alone cannot
identify deceptive design practices. The reason why aggregated UEQ scores do not consistently
indicate unethical design is because dark patterns do not always represent poor design; rather,
they often incorporate sophisticated design elements. These user experience qualities can
outweigh other overlooked factors, such as in the evaluation of gamification, or with the pattern
pull to refresh, which was evaluated as ’addictive’ but also with high perspicuity. This challenge
of reflecting multidimensional UX in a few key performance indicators (KPIs) is also highlighted
in a brief paper by the developers of the UEQ [
We advocate for a nuanced approach that considers the complexity and multidimensionality of user experience, beyond simplifying it into single, aggregated metrics, including characteristics particularly indicative of unethical design.
Particularly notable is attractiveness, which, although only part of the long version of the UEQ, was identified six times as a potential indicator of dark pattern design. This suggests that individual subcategories of the UEQ can indeed reveal unethical design practices when examined separately. Furthermore, dependability appears relevant in identifying unethical design elements.
The literature on dark patterns in social media particularly highlights the risk of addictiveness in interface designs. While prior research found mixed results in users’ capability to identify dark patterns, our evaluation demonstrates that users are capable of identifying non-obvious design features that contribute to addiction, such as the "pull to refresh", and "endless scroll" mechanisms. This ability suggests there’s potential to incorporate such a dimension into an adapted User Experience Questionnaire. Users were also identifying various dark patterns as ’deceptive’. Interestingly, these were the same patterns that were classified as unattractive, implying a relationship. This exploration shows that users can identify unethical interface designs that impact aspects of the user experience if sensible constructs are surveyed. We conclude that an extension of user experience questionnaires with an ethical dimension is feasible and overdue.
Designers are continually crafting new dark patterns, always staying a step ahead of regulations. As soon as one pattern is prohibited, another emerges in its place. It is encouraging to find that users in our survey evaluated the underlying strategies of diferent patterns consistently. This implies that it is possible to construct a User Experience Questionnaire capable of evaluating unethical design patterns over the long term and yielding consistent results. This consistency suggests a deeper, more stable basis for identifying unethical designs beyond the surface level of ever-evolving patterns, pointing toward the feasibility of developing tools that remain efective as the landscape of dark patterns shifts.
Our exploratory survey design and data analysis present a first step towards creating a standardized assessment tool for ethical user experience. Our extension of the UEQ is grounded in an analysis of dark patterns academics identified in social media, exemplifying unethical user interface design. Incorporating a broader taxonomy of dark patterns, including other domains, is expected to reveal additional items for the scale.
Our exploratory evaluation of the dark patterns is sensitive to the chosen heuristic cutof. With the chosen boundary (≤ -1), 4 patterns were counted as not being evaluated negatively on any evaluative scale. A more narrowly defined neutral ≤ -0.7 would have resulted in all patterns being evaluated negatively on at least one evaluative scale. Setting these cutofs is subjective to some degree; future benchmarking studies could inform sensible values for setting them. The survey was conducted in English and German, and therefore might have a western bias.
For this workshop paper, we did not perform tests of significance or factor analyses due to the preliminary nature of the study. Future research with a larger sample size is necessary to evaluate the validity and reliability of the newly proposed scale items.
This survey evaluated unethical design through a visual analysis on the pattern level. A limitation of the test method was that participants evaluated the designs in a non-interactive survey. If users were to click through a full user flow themselves, scales, particularly eficiency, might be evaluated worse. We suggest follow-up studies to explore how interactive prototypes and prolonged usage influence responses to an ethically adapted user experience questionnaire. Moreover, while dark patterns represent the most extensively studied unethical interface design strategies, other ethical considerations that afect user experience and potentially lead to longterm harms should also be considered.
Using known dark patterns in social media as examples of unethical user interface design, and combining an exploratory UEQ extension with data analysis, we have demonstrated the potential and limitations of using user experience metrics to evaluate the ethicality of design. We hope this workshop paper serves as a first step towards the creation of a rigorous tool to quantify ethical user experiences through questionnaires.
This work has been made possible through self-funding and family support. We express special gratitude to University Assistant Janis Lena Meißner, as well as the Digital Life Initiative and Prof. Helen Nissenbaum for enabling a research visit. [12] International Organization for Standardization, Ergonomics of human-system interaction — Part 210: Human-centred design for interactive systems, Standard, International Organization for Standardization, Geneva, CH, 2010. [13] G. Gediga, K.-C. Hamborg, Duntsch, The IsoMetrics Usability Inventory. An operationalisation of ISO 9241-10 supporting summative and formative evaluation of software systems, Behaviour and Information Technology 18 (1999) 151–164. doi:10.1080/ 014492999119057. [14] B. Laugwitz, T. Held, M. Schrepp, Construction and Evaluation of a User Experience Questionnaire, in: A. Holzinger (Ed.), HCI and Usability for Education and Work, Lecture Notes in Computer Science, Springer, Berlin, Heidelberg, 2008, pp. 63–76. doi:10.1007/ 978-3-540-89350-9_6. [15] J. Benedek, T. Miner, Measuring Desirability: New methods for evaluating desirability in a usability lab setting (2002). [16] J. Brooke, SUS: A quick and dirty usability scale, Usability Eval. Ind. 189 (1995). [17] S. Avolicino, M. Di Gregorio, F. Palomba, M. Romano, M. Sebillo, G. Vitiello, Ai-based emotion recognition to study users’ perception of dark patterns, in: HCI International 2022 - Late Breaking Papers. Design, User Experience and Interaction, Lecture notes in computer science, Springer International Publishing, Cham, 2022, pp. 185–203. doi:10. 1007/978-3-031-17615-9\_13. [18] J. Kirakowski, M. Corbett, SUMI: the Software Usability Measurement Inventory, British Journal of Educational Technology 24 (2006) 210–212. doi:10.1111/j.1467-8535.1993. tb00076.x. [19] M. Hassenzahl, M. Burmester, F. Koller, AttrakDif: Ein Fragebogen zur Messung wahrgenommener hedonischer und pragmatischer Qualität, in: G. Szwillus, J. Ziegler (Eds.), Mensch & Computer 2003, volume 57, Vieweg+Teubner Verlag, Wiesbaden, 2003, pp. 187–196. URL: http://link.springer.com/10.1007/978-3-322-80058-9_19. doi:10.1007/ 978-3-322-80058-9_19, series Title: Berichte des German Chapter of the ACM. [20] L. Sousa, L. Pedro, C. Santos, An Educational Design Research approach for developing Open Educational Resources to doctoral education training, 2022, pp. 1–4. doi:10.23919/ CISTI54924.2022.9820097. [21] M. Schrepp, A. Hinderks, J. Thomaschewski, Design and Evaluation of a Short Version of the User Experience Questionnaire (UEQ-S), International Journal of Interactive Multimedia and Artificial Intelligence 4 (2017) 103. doi: 10.9781/ijimai.2017.09.001. [22] W. Knight, Business objectives vs. user goals, in: W. Knight (Ed.), UX for Developers: How to Integrate User-Centered Design Principles Into Your Day-to-Day Development Work, Apress, Berkeley, CA, 2019, pp. 29–36. doi:10.1007/978-1-4842-4227-8\_3. [23] R. L. Thomas, D. Uminsky, Reliance on metrics is a fundamental challenge for ai, Patterns (New York, N.Y.) 3 (2022) 100476. doi:10.1016/j.patter.2022.100476. [24] I. C. Kerssens-van Drongelen, O. A. M. Fisscher, Ethical dilemmas in performance measurement, Journal of business ethics: JBE 45 (2003) 51–63. doi:10.1023/A:1024120428490. [25] L. Sánchez Chamorro, K. Bongard-Blanchy, V. Koenig, Ethical tensions in ux design practice: Exploring the fine line between persuasion and manipulation in online interfaces, in: Proceedings of the 2023 ACM Designing Interactive Systems Conference, ACM, New York, NY, USA, 2023. doi:10.1145/3563657.3596013. [26] N. Bielova, L. Litvine, A. Nguyen, M. Chammat, V. Toubiana, E. Harry, The efect of design patterns on (present and future) cookie consent decisions, USENIX Security Symposium.
USENIX Association. Accepted for publication (2024). [27] D. K. Peterson, The relationship between unethical behavior and the dimensions of the ethical climate questionnaire, Journal of business ethics 41 (2002) 313–326. doi:10.1023/a: 1021243117958. [28] R. A. Elphinston, M. J. Gullo, J. P. Connor, Validation of the facebook addiction questionnaire, Personality and individual diferences 195 (2022) 111619. doi: 10.1016/j.paid. 2022.111619. [29] L. D. Urquhart, P. J. Craigon, The moral-it deck: a tool for ethics by design, Journal of responsible innovation 8 (2021) 94–126. doi:10.1080/23299460.2021.1880112. [30] E. Prem, From ethical ai frameworks to tools: a review of approaches, AI and ethics 3 (2023) 699–716. doi:10.1007/s43681-023-00258-9. [31] T. Mildner, M. Freye, G.-L. Savino, P. R. Doyle, B. R. Cowan, R. Malaka, Defending Against the Dark Arts: Recognising Dark Patterns in Social Media, 2023. URL: http://arxiv.org/abs/ 2305.13154. doi:10.1145/3563657.3595964, arXiv:2305.13154 [cs]. [32] A. Mathur, G. Acar, M. J. Friedman, E. Lucherini, J. Mayer, M. Chetty, A. Narayanan, Dark Patterns at Scale: Findings from a Crawl of 11K Shopping Websites, Proc. ACM Hum.-Comput. Interact. 3 (2019) 1–32. URL: https://dl.acm.org/doi/10.1145/3359183. doi:10. 1145/3359183. [33] K. Bongard-Blanchy, A. Rossi, S. Rivas, S. Doublet, V. Koenig, G. Lenzini, ”i am definitely manipulated, even when i am aware of it. it’s ridiculous!” - dark patterns from the end-user perspective, in: Designing Interactive Systems Conference 2021, ACM, New York, NY, USA, 2021. doi:10.1145/3461778.3462086. [34] A. Robitzsch, Why ordinal variables can (almost) always be treated as continuous variables: Clarifying assumptions of robust continuous and ordinal factor analysis estimation methods, Frontiers in education 5 (2020) 589965. doi:10.3389/feduc.2020.589965. [35] J. Sauro, J. R. Lewis, Chapter 9: Six enduring controversies in measurement and statistics, in: Quantifying the user experience: Practical statistics for user research, Morgan Kaufmann, Oxford, England, 2012. doi:10.1016/c2010-0-65192-3. [36] H. Sandhaus, Promoting Bright Patterns, CHI ’23 Workshop: Designing Technology and
Policy Simultaneously, 2023. URL: http://arxiv.org/abs/2304.01157, arXiv:2304.01157 [cs].
All presented social media dark pattern mockups and plots are available at https://osf.io/nw2tj/. (a) Mockup (b) Result