Unpacking social scoring systems and their impact on fairness and digital dignity

Abstract

Governments worldwide are increasingly deploying algorithmic systems that influence citizens’ rights, benefits, and opportunities. These systems often resemble social scoring systems (SSSs) in functionality and societal impact. While extensively studied in China and the United States, SSS remain underexplored in other regions, and no systematization to compare SSS projects within and across regions exists. Moreover, the gap extends to the lack of a comprehensive overview of broader issues triggered by SSS, such as digital dignity, which go beyond the issues of accuracy and non-discrimination. This study addresses these gaps by focusing on existing systems in Switzerland and in developing a replicable codebook able to categorize their features, functionalities, and societal implications. Using a combination of document analysis and iterative coding, 51 systems were identified and analysed across different dimensions, such as dignity, privacy, and social equality. The findings reveal that public administration and law enforcement entities are the primary actors deploying these systems, often leveraging personal data, big data, and historical records. While these systems aim to enhance efficiency and public safety, they also raise ethical concerns, particularly regarding invasive data practices, privacy violations, and the erosion of digital dignity. A key contribution of this research is the identification of underexplored dimensions of digital dignity, including agency, compassionate decision-making, individual and cultural sense of identity, and the dynamic nature of human beings. Through the developed codebook and the provided visualization tool, this study equips stakeholders to better assess and mitigate the risks of SSS in terms of equity, accountability, and fundamental rights.

Introduction

Governments worldwide are increasingly deploying algorithms that affect citizens’ daily lives (Smuha, 2024; Tamò-Larrieux, Guitton and Mayer, 2025; Tamò-Larrieux et al., 2024). Multiple works by researchers and NGOs (e.g., AlgorithmWatch) have illustrated the risks posed by algorithms used by private companies and public authorities (Grimmelikhuijsen and Meijer, 2022; Levy et al., 2021; Lu, 2020; OECD.AI, 2025). Recent advances in Artificial intelligence (AI) ethics scholarship further highlight the tension between technological innovation and ethical oversight, emphasizing privacy-preserving methods and accountability frameworks (Radanliev, 2025; Radanliev et al., 2024). These works underline that responsible AI development requires not only technical safeguards – such as differential privacy or fairness-aware algorithms – but also socio-legal and policy mechanisms that promote transparency, fairness, and human dignity. Among these discussions on automated decision-making systems (ADMs), social scoring systems (SSSs) stand out as particularly controversial (Loefflad and Grossklags, 2024). Building on the growing field of Critical Data Studies and debates on data justice (Dencik, Hintz and Redden, 2019; Iliadis and Russo, 2016), SSS attract attention due to their function and their impact (Kayser-Bril, 2019) having as a primary goal to steer the behaviour of individuals, businesses, social organizations, and government agencies (Kostka and Antoine, 2020).

Such systems exemplify the entanglement between data governance, social control, and human rights (HRs), as they operationalize social norms through data-driven categorization. The most common definition of SSS describes systems that assign individuals a numerical score (e.g., SCHUFA scoring based on their behaviours, decisions, interactions, and contributions in various decision-making contexts, such as a social or community setting (Loefflad, Chen and Grossklags, 2023)). The score is used to measure, for instance, social status, adherence to societal norms and values, and trustworthiness (Liu and Rona-Tas, 2021; State Council, 2014), aiming to guide individuals towards behaviour deemed socially accepted (Kostka, 2019). In this sense, SSS do not merely automate decision-making; they also embed moral and political judgments within data infrastructures.

Not all SSS produce publicly available scores. Like ADMs, SSS share the notion of steering, typically in an automated manner, socially relevant interactions. Often, both ADMs and SSS aim to reduce personal data inputs into a single number or measure; this allows the algorithm to reach a decision (SCHUFA Holding AG, n.d.), ultimately regulating human behaviour (Loefflad and Grossklags, 2024). In this article, we use SSS to refer to automated decision-making (ADM) systems designed to regulate society-wide behaviour, trying to promote pre-defined ‘socially acceptable’ and law-abiding conduct.

While most discussions about SSS focus on China and the United States (de Bonth, n.d.; Van Nguyen, Lafrance and Vu, 2023; Yang, 2024) few reports have started documenting government-led projects in other regions of the world (e.g., Atlas of Automation by AlgorithmWatch). However, there is far less debate and awareness in Europe, with citizens lacking transparency and knowledge over their use, their impact, and far-reaching implications, particularly concerning digital dignity, social inequalities (de Bonth, n.d.; Park and Humphry, 2019; Loefflad, Chen and Grossklags, 2023), the fundamental right to privacy (Packin and Lev Aretz, 2019; Vinayak, 2019), issues of discrimination and bias (Castets-Renard, 2019; Cho, 2020; Packin, 2021; Raz and Minari, 2023; Zarsky, 2014). By foregrounding the concept of digital dignity – a lens that bridges ethics, law, and social theory – this study contributes to the broader conversation on how data governance reshapes the conditions of human agency and social worth.

The situation is further complicated by the fact that, while the EU prohibits SSSs that result in unfavourable treatment (art. 5 (1) (c) of the AI Act), many issues may still arise. Some SSS, still with significant impact on the individual, may fall outside the scope of the AI Act and, as a result, not be covered by this prohibition due to the simplicity of the algorithms (ARTICLE 19, 2025). This regulatory ambiguity underscores the need for empirical mapping and normative reflection that can inform more inclusive and dignity-centred governance frameworks.

Therefore, a more systematic investigation and typology are needed to raise more social awareness of their use across the world, highlight their far-reaching implications (e.g., for digital dignity), and ultimately shed light on safeguarding measures. Specifically, this study responds to calls within Critical Data Scholarship for grounded, empirically informed analyses of algorithmic systems that connect technical practices with their socio-political effects. In particular, to address social inequalities caused by SSS employed by public bodies, as well as to increase transparency over their use, we need to systematically map existing systems and create a codebook to enable comparisons on across different countries.

The systematic mapping involves selecting a specific region (in our case, Switzerland) and identifying the systems deployed by public authorities. To do so, we have conducted online searches and search of cantonal databases (e.g., parliamentary discussions and reports) and existing public databases (e.g., Atlas of Automation by AlgorithmWatch) that contain specific references to public administration or public bodies, and to automatic or scoring decision-making solutions. We specifically focus on systems that grant rights and benefits or impose sanctions and restrictions on citizens. Based on this, key systems were retrieved and manually coded for their features, purposes, operational mechanisms, and societal impacts. The development of our comprehensive codebook was iterative and partly data-driven. Such an approach fosters accountability among public authorities and empowers citizens with knowledge about how these systems operate and affect their lives. Ultimately, this effort aims to promote more transparent algorithmic governance.

This paper has four sections: A review of the literature identifying three research gaps; a description of the methodology used to study SSS in Switzerland; a presentation of results on system types, purposes, and implications for digital dignity, equality, and transparency; and a discussion situating the findings within broader debates on algorithmic governance, privacy, and social justice.

Background

SSSs and their impact on humans

Public services have evolved from early e-government to advanced decision-making systems (Ubaldi et al., 2019) across domains such as enforcement, regulation, monitoring, adjudication, public services, and engagement (Ahn and Chen, 2022; Misuraca and Van Noordt, 2020). One technology receiving growing attention is SSS, most notably China's social credit system, intended ‘to provide trustworthy people with benefits and discipline the untrustworthy’ (Alguliyev and Alakbarova, 2021; Liu and Rona-Tas, 2021). While limited information exists, some European examples illustrate their adoption or influence (Packin and Lev Aretz, 2019), including the UK's deployment of live facial recognition in London (Dodd, 2020), the EU's proposed Digital Green Certificate introduced during COVID-19 to enable safe travel via proof of vaccination, test results, or recovery (Dodd, 2020), and Switzerland's use of electronic health records (EHRs) to assess health-insurance eligibility (De Pietro and Francetic, 2018). Overall, discussion of SSS in Europe, including Switzerland, remains sparse, with the literature focusing mainly on China and the United States. While SSS promise to offer certain benefits, research shows that they raise concerns similar to those discussed in relation to other public AI systems, such as biometric surveillance and facial recognition (FRA, 2020), risk-assessment instruments (Pattison-Gordon, 2024), and algorithmic systems used to optimize and monitor workers’ performance, including empathy-assessment technologies (Dzieza, 2020). These systems not only pose technical challenges, but they also bring ethical, legal, and societal risks (Bespalko and Gaponenko, 2021; McWilliams, 2020; Orgad and Reijers, 2019). The literature on SSS highlights problems commonly discussed in AI and public administration: AI technology (e.g., data quality, AI safety), AI law and regulation (e.g, accountability, safety), AI ethics (e.g., discrimination, the compatibility of machine and human judgment), and AI and society (e.g, social acceptance) (Wirtz et al., 2019). Examples include the vast privacy violations (Loefflad, Chen and Grossklags, 2023; Packin and Lev Aretz, 2019; Vinayak, 2019), reliance on past data/previous behaviours that are unrelated to the context of scoring (Raz and Minari, 2023), data misuse (Chen and Cheung, 2017), and data bias (Castets-Renard, 2019; Cho, 2020; Packin, 2021; Raz and Minari, 2023; Zarsky, 2014). Legal scholars have likewise pointed to the lack of transparency (Falletti and Gallese, 2023) and, thus lack of adequate voluntary participation and consent when deploying SSS on a larger scale (Packin and Lev Aretz, 2019).

In addition to restricting rights to social security, privacy, and non-discrimination (Human Rights Watch, 2023), SSS may lead to even more far-reaching and harmful implications. To understand these risks, this study places SSS within wider debates on algorithmic accountability and the societal impact of quantifying people's behaviour (Van Brakel, 2021). A key concern in these debates is how to protect human dignity in an increasingly digitized and datafied society (Daly et al., 2021; Spulbar and Mitrache, 2024), where people risk being reduced to data points and separated from their lived, social contexts (Teo, 2023). Drawing on research in surveillance studies (Lyon, 2022) and datafication and governance (Dencik et al., 2019; Iliadis and Russo, 2016; Prodnik, 2021), we argue that the challenges of SSS go beyond traditional cyber risks. They threaten deeper human ‘frames’ that are especially vulnerable in automated environments – most notably human agency and human dignity. For instance, the deployment of SSS is criticized for the potential of violating the right to dignity (Amnesty International, 2024) and the values of equality and justice (Raz and Minari, 2023). These systems are also seen as infringing upon fundamental rights of citizens, including freedom of politics, personal freedom, personal dignity, and the inviolable right to the home of citizens (de Bonth, n.d.). Additional concerns voiced include manipulation, excessive market power, and social segregation (Packin and Lev Aretz, 2019). Additionally, critics argue that SSS treat individuals as numbers rather than as moral agents (Yeung, 2018). For example, China's Social Credit System has been criticized for undermining personal autonomy and moral growth (Pitsillides, 2024). Moreover, surveillance scholars have described the individual implications as well as societal effects of such systems, in particular when SSS is coupled with (law) enforcement powers (Van Brakel, 2020). Such systems can create chilling effects, raising concerns about deterrent effects and the erosion of individual rights (Büchi et al., 2020).

Legal scholars argue that algorithms are not merely technical tools (Prodnik, 2021) but socio-technical and political systems that reinforce existing power dynamics (Bergman et al., 2023). Their impact extends beyond technical issues, shaping rights, opportunities, and autonomy in public administration. Understanding these effects requires broader theoretical lenses rooted in inclusion, equity, and human dignity (Bonelli, 2013). Ethical and Responsible AI frameworks (Corrêa et al., 2023; Göllner et al., 2024) stress that systems must be assessed not only for accuracy but also for ethical, legal, and social implications (Radanliev, 2025; Radanliev et al., 2024).

Human-rights-based approaches extend the focus beyond privacy and discrimination to equality, dignity, and empowerment (Niklas, 2021). Ensuring human-centric digital transformation requires integrating HRs throughout the technology lifecycle (OECD, 2025), as illustrated by proposals such as the right to reasonable inferences (Wachter and Mittelstadt, 2019). Such frameworks underscore the need to assess how technologies, including SSS, affect human agency and dignity. Because traditional rights frameworks emphasize individual harms, they often miss the invisible, collective, and systemic harms emerging from datafied systems. This has drawn attention to data politics (Ruppert et al., 2017), data activism (Milan and Velden, 2016), and data justice, which reveal how data infrastructures generate new forms of visibility, vulnerability, and exclusion (Taylor, 2017). Furthermore, reconceptualizing digital inequality within frameworks of social justice highlights the broader ethical concerns surrounding technology and HRs (Halford and Savage, 2010).

To be or not to be algorithm-aware remains a critical question in understanding the implications of SSS. The algorithm awareness divide emphasizes educational disparities in recognizing and contesting algorithmic influence. From a critical data perspective, such awareness also functions as a precondition for exercising digital agency and dignity (Gran et al., 2021). Research on this topic demonstrates that awareness can mitigate the chilling effects and promote engagement with technologies in ways that respect individual rights and collective fairness (Gran, Booth and Bucher, 2021). The exploration of inequalities and their impact on dignity further reinforces the need to theorize SSS through intersectional and inclusive frameworks (Park and Humphry, 2019; United Nations Development Programme, 2022). By recognizing horizontal inequalities across gender, race, ethnicity, sexual orientation, and other factors, these approaches provide valuable insights for evaluating and addressing the broader societal implications of SSS.

SSS raise not only concerns about discrimination and bias but also about the deprivation or empowerment of individuals’ rights, opportunities, and benefits, directly affecting human dignity and its dimensions of freedom, equality, and solidarity (Aizenberg and Van Den Hoven, 2020). These issues relate to broader debates on privacy, transparency, and non-discrimination in the ethical and legal literature on SSS (McWilliams, 2020; Orgad and Reijers, 2019). The literature remains heavily focused on China and the United States, with limited attention to Europe, including Switzerland, creating a geographic and epistemic bias that restricts understanding of SSS in democratic, decentralized governance contexts. Moreover, insufficient engagement with critical data scholarship leaves questions of power, agency, and digital dignity under-theorized.

SSSs and lack of awareness

In the EU and Switzerland, explicit SSS – systems assigning social scores to allocate benefits – do not exist, but algorithmic systems with similar objectives are increasingly used. The key difference lies in framing: Algorithmic systems focus on decision processes, while SSS carries a broader moral and societal framing akin to social credit. These systems shape access to services and benefits, yet public awareness of their operation and ethical implications remains low (Kayser-Bril, 2019). Studies highlight limited transparency and communication about their public use, including examples such as the EU Digital Green Certificate and Swiss EHRs used for insurance premiums, whose implications for privacy, discrimination, and inequality are poorly understood (Shahbazi et al., 2022). Such systems aggregate personal data across sources into simplified decision metrics, a process largely invisible to those affected (Kayser-Bril, 2019). The research gap lies in the limited understanding of how these systems influence social dynamics, digital dignity, and inequalities, coupled with inadequate public communication about their purpose and risks. While some scholarship explores the technical and ethical implications of SSS, such as their propensity to perpetuate bias (Castets-Renard, 2019) or invade privacy (Packin and Lev Aretz, 2019), there is a paucity of studies on how these systems reshape public perceptions of responsibility, solidarity, and citizenship. By focusing on individual scores, these systems risk shifting responsibility from institutional structures to individuals, reinforcing the perception that citizens alone are accountable for their societal position, while obscuring the systemic factors that contribute to inequalities (Aizenberg and Van Den Hoven, 2020). Without transparent communication about how these systems operate, citizens are often unaware of the extent to which SSS affect their rights and access to resources. This lack of awareness exacerbates the ‘chilling effect’, discouraging individuals from exercising legitimate rights due to fear of surveillance or scoring repercussions (Büchi et al., 2020; Van Brakel, 2020). For example, systems used in Sweden, Spain, or France to allocate social benefits, detect fraud, or monitor online behaviour have profound societal implications, yet they remain largely invisible to public discourse (Shahbazi et al., 2022)

The lack of communication and transparency about SSS has significant societal implications. It perpetuates inequalities by normalizing opaque systems that marginalize underrepresented groups, as documented in earlier studies on algorithmic bias (Shahbazi et al., 2022; Wachter et al., 2020). Furthermore, the absence of inclusive representation in datasets undermines the fairness of these systems, leading to unjust outcomes in employment, credit, housing, or social services (Castets-Renard, 2019; Raz and Minari, 2023). Ultimately, the findings aim to foster a more informed public discourse, empowering citizens to engage with and challenge the systems that govern their lives, while providing policymakers with actionable recommendations to ensure transparency, accountability, and equity in algorithmic decision-making.

Methodological approaches to classify the impact of SSSs

To the best of our knowledge, no studies provide a comprehensive overview of existing SSS. Some reports, such as those by Algorithm Watch, offer insights into the use and implications of ADM (AlgorithmWatch, 2020), and initiatives like the TA-SWISS project ‘Systems of Social Scoring’ investigate and define SSS in Switzerland. Prior research approaches the impact of SSS from different perspectives: A human-rights analysis based on the Universal Declaration of Human Rights and the European Convention on Human Rights (de Bonth, n.d.); a design-oriented translation of HRs into context-specific requirements (Aizenberg and Van Den Hoven, 2020); and studies of data injustices using illustrative cases such as biometric databases and migrant-visualization systems (Taylor, 2017). Taylor's Data Justice framework (2017) emphasizes visibility, engagement with technology, and anti-discrimination, but does not fully address broader humanistic dimensions such as dignity, empathy, or the dynamic experiences of individuals (Aizenberg and Van Den Hoven, 2020). Similarly, Ersoy (2024) calls for stronger transparency and reporting but provides no tools for systematic mapping or cross-regional comparison. To fill these gaps, this study has employed thematic analysis, a valuable technique for coding meaningful themes, categorizing common codes, and conceptualizing their inner meaning (Clarke and Braun, 2013; Jnanathapaswi, 2021), which might ultimately might facilitate a nuanced understanding of diverse sets (Hecker and Kalpokas, 2025). By linking thematic coding to a normative framework grounded in digital dignity and data justice, we have adopted this comprehensive and replicable methodology to classify and analyze the actors employing these tools, the target audience involved, the interests at stake, and the impact of SSS. This methodology also addresses critical research gaps in the field. First, it provides a standardized approach for comparing systems across regions, addressing the lack of cross-regional analyses highlighted in the literature (AlgorithmWatch, 2020; Smuha, 2024). Second, it bridges the gap between technical functionalities and societal impacts by linking operational mechanisms to broader ethical and social concerns, such as discrimination, bias, and inequalities (de Bonth, n.d.; Castets-Renard, 2019). Third, it operationalizes the theoretical dimensions of fairness, transparency, and dignity (Radanliev, 2025; Taylor, 2017) into an empirical coding framework, thereby translating abstract ethical principles into measurable analytical categories.

Methodology

Gathering of data

To identify the relevant systems in Switzerland, we have conducted online searches and searches of cantonal databases (e.g., parliamentary discussions and reports) and existing public databases (e.g., Atlas of Automation by AlgorithmWatch) having the following characteristics ¹ : (‘government’ OR ‘administration’) AND (‘algorithmic decision-making’ OR ‘automated systems’ OR ‘social scoring’ OR ‘digital surveillance’ OR ‘AI governance’ OR ‘algorithms’ OR ‘machine learning’ OR ‘Big Data’). We specifically focus on systems that determine which grant rights and benefits, or impose sanctions and restrictions, on citizens (thus excluding chatbots or information retrieval tools). We also include systems that have not (yet) become applicable in Switzerland. For instance, the concept ‘jedem sein CO2-Budget’ (everyone gets their CO2 budget) in 2022 involved allocating personal CO2 budgets to individuals as a way to encourage sustainable living, with the potential for carbon emissions to become a measurable and tradable resource, similar to money, and to potentially impact behaviour in promoting sustainability. We acknowledge that our dataset includes only publicly documented systems, thereby excluding proprietary or classified initiatives that may nonetheless have significant societal impacts. This limitation, however, aligns with the emphasis on open and accountable data sources. In total, 51 systems were identified and coded until August 2025. The inclusion criteria required that (i) the system is implemented or proposed by a Swiss public authority; (ii) it supports ADM affecting individuals’ rights, benefits, or obligations; (iii) sufficient public documentation exists (e.g., official reports, parliamentary debates, media, or AlgorithmWatch's Atlas of Automation); and (iv) it involves data-driven scoring, classification, or risk assessment, even without a public ‘score’. Systems limited to information retrieval, chatbots, or internal administrative optimization were excluded.

Development of codebook

Thematic analysis is a valuable technique for coding meaningful themes, categorizing common codes, and conceptualizing their inner meaning (Clarke and Braun, 2013; Jnanathapaswi, 2021), which ultimately might facilitate a nuanced understanding of diverse sets (Hecker and Kalpokas, 2025). This analytical approach was selected because of its capacity to bridge technical coding practices with interpretive, theoretically informed insights – a central requirement for Critical Data Scholarship (Iliadis and Russo, 2016). In addition, thematic analysis has proven effective in illustrating challenges that emerge from different tools. For instance, it has been used in a study on user preferences and concerts on trigger-action platforms (Hecker and Kalpokas, 2025), as well as in research identifying the challenges and solutions for integrating security tools (Romare et al., 2023). We have adopted this comprehensive and replicable methodology to classify and analyze the actors employing these tools, the target audience involved, the interests at stake, and the impact of SSS etc. The approach combines systematic data collection with an iterative coding process to identify and categorize the features, functionalities, and societal impacts of these systems. The development of a structured codebook was a key innovation, enabling the categorization of systems based on dimensions such as actors involved, inputs, methods, outputs, legal frameworks, and impacts (both positive and negative). Building on digital dignity and data justice frameworks (Aizenberg and Van Den Hoven, 2020; Taylor, 2017), we extended standard ethical impact categories – fairness, privacy, bias – to include relational dimensions such as agency, empathy, and deshumanization, thus operationalizing theoretical constructs into empirical indicators. Furthermore, existing databases, such as Atlas of Automation from AlgorithmWatch and COHUBICOL ² , provided valuable inspiration for developing our codes (e.g., functionality and affected groups). To ensure coherence with emerging AI ethics frameworks, the codebook also integrated cross-references to transparency and accountability principles.

The coding process followed a systematic approach. The collected systems were manually based on a series of deductive dimensions to ensure a comprehensive evaluation. However, the categories within each dimension were partly inductive and partly data-driven, based on the analysis of documents (e.g., newspaper articles, reports) related to each system. Furthermore, the development of codes was guided by the need to understand who is involved, how these systems function, and their impact. This resulted in key dimensions such as actors involved, actor interests, functionality, target audience, input, method, output, legal framework, and impact on dignity and privacy (see Table A.1 of Appendix A.1). Regarding the impact, the goal was to explore far-reaching implications of SSS. Therefore, the primary focus was to identify different dimensions of dignity, rather than addressing commonly discussed issues like accuracy and non-discrimination. The coding process was iterative, especially for the impact category. The first stage involved reviewing diverse specific and general sources and identifying quotes and expressions of broader human implications, such as dignity. In the second stage, a student assistant and a PhD student worked together, discussing points of agreement and disagreement, integrating specific and general sources, refining existing categories, and adding new ones to improve the conceptual framework.

These dimensions (see Figure 1) form our conceptual framework, which focuses on the interplay between three key components: Public actors (i.e., the entities implementing the tools), target audiences (i.e., those affected by the tools), and the impacts (the consequences of the tools’ use on the target audience). Public actors deploy the system to pursue specific interests. The system functions by processing certain data inputs through established methods to generate outputs. These outputs provide technical information to inform diverse decisions in the public sector (e.g., refugee integration decisions). The decisions taken by public actors, in turn, then lead to broader societal impacts that may influence individual dignity, privacy, etc.

Figure 1.

Conceptual framework of the study with the coded dimensions.

Coding and inter-annotator agreement

Two coders, one student assistant, and an expert PhD student, manually coded the systems, systematically discussing points and disagreements, while also complementing each other in terms of specific and generic sources about the systems. The resulting coding of the systems appears solid because it reflects complete agreement between the coders. This is notable considering that a good measure of inter-annotator agreement is typically 80%. This iterative process not only ensured thorough examination but also enabled the addition of new categories as needed for each dimension. This dynamic approach allowed the framework to evolve, accommodating unique or unforeseen aspects of the systems under evaluation, thus enhancing the comprehensiveness and relevance of the analysis. Periodic validation sessions were conducted with the team members to triangulate interpretations and ensure conceptual alignment with critical data perspectives. Overall, the mixed deductive–inductive strategy aligns with the emphasis on methodological transparency, interdisciplinarity, and critical engagement with data-driven power structures.

Results

Descriptive results

The data presented provides descriptive statistics about the coded dimensions based on the code book. Overall, the mapping reveals that SSSs in Switzerland are primarily embedded in public governance infrastructures rather than the private sector, confirming the growing integration of data-driven decision-making into welfare, security, and regulatory functions.

The analysis begins with the policy areas, showing that the majority (51 cases) focus on the public sector, followed by a much smaller presence in the public sphere (4 cases) and workplace (2 cases). Regarding the actors involved, the most commonly implicated entities are public administration (24 cases) and police (19 cases), followed by state/state departments (18 cases) and regulatory agencies (10 cases). Private sector actors, NGOs, legal bodies, and judges are less frequently involved. This concentration of systems in law enforcement and administration underscores how algorithmic tools are used to mediate eligibility, surveillance, and compliance – functions that have direct implications for human dignity and autonomy (Aizenberg and Van den Hoven, 2020; Yeung, 2018). When considering actor interests, recurring priorities include security (31 mentions), profit maximization/creditworthiness/market efficiency (15 mentions), law enforcement (9 mentions), fairness and equity (6 mentions), and transparency/accountability (4 mentions). Some categories, such as public acceptance and surveillance, are mentioned minimally. These findings align with a systematic review exploring the links and reciprocal relationships between AI and HRs, which shows that the right to security (19.4%) is one of the most frequently discussed rights in relation to AI (Mpinga et al., 2022). The dominance of ‘security’ and ‘efficiency’ reflects the utilitarian logic of these systems, often at the expense of participatory or rights-based values emphasized in responsible AI frameworks (Radanliev et al., 2024).

The functionality of the systems reveals that behavioural profiling (15 mentions) is a significant component, along with alerts and notifications (10 mentions) and surveillance/tracking (15 mentions combined across categories). Efficiency of the process is moderately cited (10 mentions). This prevalence of behavioural monitoring indicates a tendency toward predictive governance – anticipating risks or non-compliance through quantification – which may amplify concerns about social stigmatization and unequal treatment (Taylor, 2017).

For target audiences, the general public is the most frequently impacted group (25 cases), followed by victims/persons in legal proceedings (9 cases) and criminals (9 cases). Other groups, like employees, political actors, and students/teachers, are less represented. This distribution suggests that algorithmic governance is not confined to high-security contexts but extends to everyday public services, thereby normalizing low-visibility forms of social scoring.

The input data includes personal characteristics (25 mentions), big data and survey data (16 mentions each), and historical records (11 mentions). Legal documents, criminal records, and utility bills are among the less frequently used inputs. The methods most commonly applied involve machine learning and AI (30 mentions), rule-based scoring (26 mentions), and API connectivity (19 mentions). Systems using AI methods often scored higher on ethical risk categories, reinforcing links between technological opacity and reduced explainability. Data aggregation (8 mentions) and sensor integration (9 mentions) are also highlighted. The technical outputs are predominantly ‘no score assigned’ (41 cases), while score augmentation is relatively rare (9 cases).

For the broader effects, security or personal safety is the most cited impact (20 mentions), followed by detection of irregularities (8 mentions) and employment effects (4 mentions). The legal framework is strongly influenced by human-centric system guidelines (26 mentions) and trustworthy AI certification (26 mentions). Responsibility of developers and users (20 mentions) and ethical AI certifications (15 mentions) are also prominent.

Regarding negative impacts on privacy, preserving confidentiality and anonymity is a significant concern (20 mentions). The issue of privacy is followed by issues of informed consent (7 mentions) and illegal sharing of sensitive information (3 mentions). This pattern mirrors broader debates on the limits of informed consent in large-scale data ecosystems (Dencik et al., 2019). Negative impacts on dignity include harm to agency and autonomy (5 mentions) and self-sufficiency concerns, particularly regarding the right to safety, freedom, healthcare, and employment (9 mentions). These effects reflect findings from a recent comprehensive study showing that AI can affect a wide range of fundamental rights, including civil, political, economic, social, and cultural rights (Mpinga et al., 2022). A major concern is solidarity-related issues, including group fairness, equality, exclusion, stigmatization, and discrimination (26 mentions). These concerns show that datafication is harmful not only because it restricts individual autonomy, but also because algorithmic scoring does more than reflect existing inequalities (Viljoen, 2021). It actively reproduces structural injustices, consistent with the concept of ‘data justice’ (Taylor, 2017). Addressing these issues requires a vision for socially beneficial data use that emphasizes democratized data governance. This means shifting from individual privacy toward broader concerns such as social legibility and group privacy (Lyon, 2022) – focusing less on individual rights alone and more on collective and institutional mechanisms that manage data fairly, reduce inequality, and ensure that technology supports broader social justice goals (Viljoen, 2021). Dehumanization manifests in several ways: The feeling of not being judged by a human or someone capable of moral reasoning, ethical consideration, and accountability (6 mentions), not treating individuals as complex beings with rich experiences and preferences, leading to objectification and a lack of respect (2 mentions), and a failure to acknowledge the dynamic and changing nature of human beings (3 mentions). This adds to the literature on AI's impact on human dignity, showing dehumanization through the disembodiment of empiric self-representation and contextual sense-making; the choice architectures for the exercise of cognitive autonomy; and the experiential context of lived experiences (Teo, 2023). Additionally, harm to social standing and reputation (2 mentions) is highlighted as a significant concern. A lack of compassionate decision-making (2 mentions) further contributes to dehumanization in automated processes. Negative impacts in other domains lack transparency (4 mentions), and other challenges related to community building, citizen confidence, and moral conformism (4 mentions). To concretize these dynamics, three short examples illustrate typical dignity and privacy implications:

Asylum seekers and migrants: A machine-learning system allocates asylum seekers to cantons for efficiency, but NGOs highlight fairness, privacy risks (non-informed consent), and dignity harms such as dehumanization and stigmatization.

Automated recruiting: Employers use AI scoring under Trustworthy AI certifications. Despite regulatory focus on fairness and privacy, candidates experience dehumanization and exclusion, showing how certified tools can still undermine autonomy.

Anonymization of judgments: The Federal Supreme Court uses machine learning to anonymize decisions, improving confidentiality but reducing contextual sensitivity and compassion, creating dignity concerns.

On the positive side, privacy benefits centre on preserving confidentiality (7 mentions), control over the data (1 mention), and informed consent to processing (8 mentions). Positive dignity impacts include agency and autonomy (5 mentions) and self-sufficiency (13 mentions). Solidarity, encompassing group fairness, equality, exclusion, stigmatization, and discrimination, is noted (3 mentions). Additionally, the importance of treating individuals as individuals with respect and recognition (2 mentions) is emphasized. The feeling of being judged by a human rather than an automated system (1 mention) is also highlighted as a positive aspect. Positive other impacts highlight efficiency (10 mentions) as a key benefit. Redistributive justice (2 mentions) is also recognized as a positive outcome. Additionally, community building and harmonized togetherness (2 mentions), access to resources, including technology (2 mentions), and confidence among citizens (3 mentions) are noted. Transparency is mentioned once (1 mention), potentially indicating improvements in openness in certain contexts. However, positive dignity impacts remain the exception rather than the rule, suggesting that current governance frameworks insufficiently operationalize the ‘human-centric’ ideals outlined in EU AI ethics guidelines.

Online application

The Shiny app in Figure 2 is designed to visualize the analyzed systems in Switzerland. It provides a clear and structured layout, allowing users to filter data based on functionality (such as alerts and notifications, behavioural profiling, process efficiency, and surveillance), as well as target groups (like asylum seekers, migrants, employees, and others). This interactive tool functions as a visualization device, translating complex data into a format accessible to citizens, journalists, and policymakers. The filtering interface is user-friendly with options to select or deselect all criteria. Moreover, the system allows the user to select all pertinent categories, rather than limiting them to just one category, such as choosing just one actor.

Figure 2.

Current version of the shiny app, which can be found at https://mdrv.shinyapps.io/SSS_typology.

The filtered data section displays detailed information about specific systems, such as the ‘IPL Algorithm for Refugee Integration’. This includes the purpose of the system, which in this case is the allocation of asylum seekers to cantons, and the involvement of actors such as state departments, public administration, and NGOs. It outlines the input, method, and output of the system, highlighting the use of machine learning and AI, and details the legal framework as a human-centric system.

The app also highlights potential impacts, both positive and negative, including privacy concerns, issues of dignity, dehumanization, and the effects on anonymity. In this way, the application operates as a public engagement interface that re-materializes abstract ethical principles – such as fairness, accountability, and transparency – into navigable, concrete representations (Radanliev, 2025). The layout seems informative, with a focus on transparency and engagement for users interested in understanding the implications of such systems. The app further includes sources (e.g., news articles, official reports, etc.) about the specific systems and provides examples of quotes/citations that we used to analyze the impacts.

Discussion

General discussion

This study's systematic approach and the development of a comprehensive codebook add to the literature by offering a replicable methodology for cross-regional comparisons. The descriptive statistics reveal the central role of public administration and law enforcement entities in implementing SSS, aligning with prior findings that governments often use algorithmic systems to enhance operational efficiency and public safety (Ahn and Chen, 2022; Ubaldi et al., 2019). By situating these findings within Critical Data Studies and data justice debates (Dencik et al., 2019; Iliadis and Russo, 2016; Taylor, 2017), this research underscores that algorithmic infrastructures are not neutral tools but social institutions that mediate power, visibility, and participation.

The study's emphasis on behavioural profiling, surveillance, and tracking underscores the risks of invasive data practices, which resonate with documented concerns about privacy violations and data misuse (Chen and Cheung, 2017; McWilliams, 2020). These tendencies reflect what Yeung (2018) describes as ‘algorithmic regulation’ (governance through data-driven norms), which often privileges control and prediction over deliberation and empathy. The heavy reliance on inputs such as personal characteristics, big data, and historical records reinforces earlier critiques about bias and reliance on irrelevant or outdated data (Raz and Minari, 2023). A key novel finding is the nuanced analysis of both negative and positive impacts of SSS on privacy and dignity. For instance, while previous studies have broadly discussed privacy concerns (Loefflad, Chen and Grossklags, 2023; Packin and Lev Aretz, 2019), this research identifies specific issues such as inadequate informed consent and the illegal sharing of sensitive information. Similarly, the study's attention to dignity introduces underexplored dimensions, such as agency and the dynamic nature of human beings, thereby contributing to emerging discussions on the erosion of humanistic values in digital governance (Aizenberg and Van Den Hoven, 2020). In doing so, it complements recent AI ethics frameworks advocating for responsible deployment, fairness, and privacy (Radanliev, 2025), while extending these discussions toward a more human-centred conception of digital dignity.

Another significant contribution lies in highlighting how these systems disproportionately impact specific target audiences, such as the general public and victims in legal proceedings. This supports concerns about the exacerbation of social inequalities raised by (de Bonth, n.d.) and (Human Rights Watch, 2023), while adding empirical evidence of these systems’ operational focus. Moreover, the study documents the technical outputs and broader effects of SSS, such as their role in security and detection of irregularities. While these functions might enhance efficiency, they also raise ethical questions about fairness, transparency, and the chilling effects of surveillance (Büchi et al., 2020; Van Brakel, 2020). Our findings confirm that the pursuit of efficiency and security often displaces values of solidarity and inclusion – an imbalance that future governance frameworks must explicitly correct. This underscores the importance of critically examining how such systems are deployed and ensuring they do not entrench structural inequities further. The reliance on biased input data can further reinforce these inequalities (Castets-Renard, 2019; Raz and Minari, 2023). These trends reflect broader concerns in the literature about the ‘algorithm awareness divide’ and its role in perpetuating unequal outcomes (Gran, Booth and Bucher, 2021).

Methodological contribution of the project

The iterative development of the codebook is another novel contribution. Unlike prior work that has critiqued the lack of standardization in comparing SSS across regions (Pedreshi, Ruggieri and Turini, 2008; Taylor, 2017), this study offers a replicable framework for analyzing such systems. By categorizing features, functionalities, and societal impacts, the codebook provides a tool for systematically mapping SSS and highlighting their broader implications. The framework operationalizes theoretical concepts from digital dignity and data justice into measurable analytical categories, bridging the gap between critical social theory and empirical assessment. This approach not only addresses gaps in documenting government-led algorithmic systems (e.g., Atlas of Automation) but also empowers researchers, policymakers, and civil society to better assess their use. Furthermore, it supports the development of a more comprehensive framework for analyzing SSS, one that incorporates data justice principles (Taylor, 2017) and addresses the broader societal harms identified in the literature (Castets-Renard, 2019; Pitsillides, 2024). This methodological innovation directly responds to the call for theoretical integration and demonstrates how empirical coding can serve normative and policy-relevant objectives. This research is a critical step toward fostering accountability among public authorities and promoting equitable digital governance.

Overcoming awareness issues through science communication

Existing literature underscores that the opacity of algorithmic systems exacerbates public mistrust and disengagement (Büchi et al., 2020; Van Brakel, 2020). However, the public awareness of SSS and their implications remains critically low, particularly in regions where these systems are not explicitly labelled as ‘social scoring’ but have comparable functionalities and impacts (Orgad and Reijers, 2019; Rose et al., 2020). Communication to society of scientific developments with major implications on society, such as in our case, is crucial (Jucan and Jucan, 2014). It might promote science literacy and educate the public and policymakers (Lopes et al., 2024; Tuttle et al., 2023), ultimately empowering individuals to make more informed decisions, as demonstrated in the environmental domain (Multazam, Pujowati and Hartati, 2024). One method of science communication is through applications. This approach has proven efficient, which is why it was selected for our study. By way of illustration, a study has shown that mobile health apps are good tools for disseminating health information and empowering individuals to take informed actions about their health (Yusuf et al., 2022), while another study has revealed that the BrAware mobile app successfully raised women's awareness about breast cancer (The Global Legal Post, 2024). Our project bridges the gap between technical developments and public understanding by using an accessible, transparent communication tool – The Shiny app. This design embodies the ethos of ‘critical transparency’, translating research findings into a participatory format that allows citizens, journalists, and policymakers to interrogate algorithmic systems. Public-facing resources like this can empower citizens, fostering greater engagement with and oversight of these systems.

Developing digital dignity guidelines for public authorities

Our findings reveal significant impacts of SSS on digital dignity, including issues related to agency, autonomy, the dynamic nature of human beings, and humanization, etc. Building on the intersection of AI ethics (Radanliev 2025; Radanliev et al., 2024) and critical data justice (Taylor, 2017), we argue that regulatory and ethical frameworks must extend beyond technical safeguards to address relational and humanistic harms. A key question is whether the current legal framework adequately addresses these issues. While our focus rested on Switzerland, the development of guidelines to assess SSS should be understood more broadly, within the European context. In fact, Switzerland currently lacks specific laws, statutory rules, or regulations that directly govern AI (Lighthouse Report, 2023).

As Switzerland prepares its AI regulatory framework, alignment with international standards, particularly the EU AI Act, is recommended to prevent fragmentation (Lighthouse Report, 2023). Article 5(1)(c) of the EU AI Act prohibits AI-based social scoring that leads to unfavourable treatment unrelated to the original data context or disproportionate to individuals’ social behaviour. However, as this threshold requires the use of AI and demonstrable unfavourable treatment, not all impactful SSS are covered, allowing simpler systems to fall outside the prohibition. The Dutch SyRI scandal illustrates how low-complexity systems can nonetheless severely harm individual rights (Kompetenzzentrum Öffentliche IT, n.d.). Accordingly, forthcoming Commission guidelines on AI implementation should clarify when simple SSS should be prohibited and ensure that existing systems used in welfare and migration contexts are also covered (ARTICLE 19, 2025).

We propose contributing to Commission guidelines to strengthen protections around SSS, which could also inform Swiss policy. These guidelines should build on existing regulatory efforts, including data justice principles (Taylor, 2017) and human-centric AI frameworks and certification schemes such as HLEG (Rajapakse et al., 2021) and OECD (European Commission, 2020), while expanding to incorporate respect for individual and cultural identity, human dynamism, and solidarity. Tailored digital dignity guidelines should explicitly address the social inequalities reinforced by SSS, including by mitigating systemic bias through mandatory fairness audits and stakeholder engagement (Raz and Minari, 2023; Taylor, 2017), and enabling marginalized groups to challenge adverse or erroneous decisions (Gran et al., 2021; Van Brakel, 2020). Given the control-oriented focus of many SSS on security and surveillance rather than equity or solidarity, guidelines should prioritize fairness and inclusivity as core design principles (Büchi et al., 2020; Park and Humphry, 2019). Integrating frameworks such as data justice and inclusive algorithmic design can help prevent disproportionate harms to vulnerable populations and promote equitable algorithmic governance (Taylor, 2017; de Bonth, n.d.). Some of the leading questions that can be used to identify and address the far-reaching implications are listed in Table A.2 of Appendix A.2. By combining insights from responsible AI and data justice frameworks, our proposed guidelines contribute to a broader vision of ‘dignity-centered AI governance’ that prioritizes inclusion, empathy, and accountability in public-sector algorithmic systems.

Future research and limitations

While this study provides valuable insights, it also highlights areas for further exploration. First, our analysis is limited to Switzerland, a context characterized by decentralized governance and a lack of formalized SSS. Expanding this research to other regions, particularly those with centralized governance structures or explicit SSS, could offer richer comparative insights. Comparative analysis across democratic and authoritarian contexts would also test the transferability of the digital dignity framework and its applicability to different socio-legal regimes. Second, the reliance on publicly available data may have constrained the scope of system identification and impact assessment. Future research could incorporate stakeholder interviews or ethnographic methods to uncover less visible aspects of SSS deployment. Participatory and co-design approaches could also help integrate citizen perspectives into the development of dignity metrics and audit tools, reinforcing the dialogic character of responsible AI governance. Lastly, while the codebook provides a robust framework for analysis, its application to diverse contexts may reveal the need for additional categories or refinements. Iterative validation in collaboration with interdisciplinary teams and affected communities would enhance its utility and adaptability.

Conclusion

The rapid institutionalization of government-led algorithmic scoring systems in Switzerland – particularly those with functionalities analogous to SSS – is no longer a peripheral policy issue but a matter of constitutional and democratic significance. In the Swiss context, where direct democracy and federal governance are foundational principles, the deployment of predictive analytics, risk-based profiling, and automated decision-support tools in public administration raises pressing questions about transparency, accountability, and the preservation of human dignity. At the European level, parallel debates within the EU and the Council of Europe reflect growing concern about the cumulative societal effects of such systems, particularly where they shape access to social services, migration control, taxation, or law enforcement. For instance, the national (Switzerland) and international (Council of Europe) authorities started addressing these challenges through concepts like a right to digital integrity, which encompasses the right to be forgotten, to security in digital space (EDRi, 2021; Rochel, 2021). The type of conversations and discussions is expected to continue, especially with the emergence of different organizations/coalitions that advocate for upholding rights in the digital space, such as the Digital Dignity Coalition (Marjanovic et al., 2022). This study shed light on the increasing deployment of government-led algorithmic systems with functionalities akin to SSS in Switzerland and their far-reaching implications. By systematically mapping these systems and developing a novel, replicable codebook, this research contributes to addressing critical gaps in the documentation and comparative analysis of SSS, particularly in regions outside China and the United States. Our findings reaffirm that even in democratic settings, ostensibly neutral scoring mechanisms can reproduce inequality and erode human dignity, thereby challenging complacent assumptions about ‘trustworthy’ AI. Our findings underscored the multifaceted societal implications of these systems, including their impact on privacy, dignity, and social equality, as well as their potential to exacerbate existing inequalities (Loefflad, Chen and Grossklags, 2023; de Bonth, n.d.). Key contributions of this work include the identification of underexplored dimensions of digital dignity, such as agency, autonomy, and humanization, and the development of a framework that links technical functionalities to broader societal impacts. These insights not only advance academic understanding but also provide actionable tools for policymakers, civil society, and researchers to evaluate and improve the governance of algorithmic systems (Aizenberg and Van Den Hoven, 2020; Taylor, 2017). Moreover, this study highlights the importance of public awareness and engagement in ensuring accountability and transparency in the use of SSS. Despite its contributions, this research acknowledges several limitations, including its focus on Switzerland and reliance on publicly available data. Future work should expand the scope of analysis to other regions and incorporate diverse methodological approaches, such as stakeholder interviews and participatory frameworks, to provide a more comprehensive understanding of SSS (AlgorithmWatch, 2020). Ultimately, this study underscores the urgent need for ethical and regulatory frameworks that prioritize digital dignity, transparency, and fairness in the design and deployment of algorithmic systems. Building on the principles of data justice and responsible AI, we advocate for a shift from efficiency-centred to dignity-centred algorithmic governance. By fostering greater accountability and promoting equitable digital governance, we aim to contribute to a future where such technologies enhance, rather than undermine, human and societal well-being (Castets-Renard, 2019; McWilliams, 2020).

Supplemental Material

sj-docx-1-bds-10.1177_20539517261434338 - Supplemental material for Unpacking social scoring systems and their impact on fairness and digital dignity

Supplemental material, sj-docx-1-bds-10.1177_20539517261434338 for Unpacking social scoring systems and their impact on fairness and digital dignity by Maud Reveilhac, Vlada Druta, Aurelia Tamò-Larrieux and Marlène Sapin in Big Data & Society