1 Introduction

We face new challenges for the didactic transposition of statistical and data-related knowledge and practices into educational systems, as data science and demands for data literacy (DL) and statistical literacy (SL) are rapidly expanding. Furthermore, scholarly knowledge in this regard is interdisciplinary, dynamic, and constantly evolving (Friedrich et al., 2024). Data can be used for social good, including work to improve sustainability and address climate change, but it can also promote the economic and political interests of a few while failing to serve the interests of most citizens (Zuboff, 2019). Data’s use has raised massive concerns about privacy, misuse, ethics, and surveillance of citizens, to name a few. Awareness of the non-objective nature of data is also raising concerns about potential gender or racial biases arising from how data are collected, selected, labeled, or used to train algorithms (D’Ignazio & Klein, 2020; O’Neil, 2016; Ridgway, 2016). Data science and data-driven artificial intelligence (AI) have led to breakthroughs in science and society, yet generative AI (GenAI) has only accelerated and exacerbated such concerns (Vassel et al., 2024).

Trends such as those sketched above are informing calls for including more data science and data literacy into the school curriculum, using varied conceptualizations of such broad constructs (e.g. Data Science Learning Progressions, 2025; IDSSP, 2019; NASEM, 2023; NCTM et al., 2024; NRC, 2026). However, related concepts and practices are only slowly making their way into school curricula (Sukol 2024, 2025). The multitude of issues mentioned provided a strong rationale for a survey commissioned by the International Program Committee of the 15th International Congress on Mathematics Education on the theme “Statistics and Data Science Education as a Vehicle for Empowering Citizens” (Biehler et al., 2025). The survey team came across several Special Issues in mathematics, statistics, computer education, and the learning sciences published on DL and data science education (Acker et al., 2024; Biehler et al., 2022; Knight et al., 2022; MacGillivray et al., 2021; Sanusi & Chiu, 2024; Wilkerson & Polman, 2020) showing the growing interest in statistics and data science education across multiple disciplines. Two key takeaways from that survey were that there is a plurality of conceptions of ‘data literacy’ and of ‘citizenship,’ but little work on a needed conceptualization of ‘data literacy for citizenship’ and related empirical research. In the current survey paper, we aim to take steps to further address the gap in conceptualization of ‘data literacy for citizenship’ (DataLitCit), as well as introduce this Special Issue “Enhancing Data Literacy for Citizenship: Innovative Approaches to Data Science and Statistics Education” of ZDM – Mathematics Education 58(6) that responds to the need for more empirical research in this area.

This survey paper builds not only from sources noted earlier and others below, but also from two prior Special Issues in ZDM, in particular vol. 55(5) “The connections between citizenship education and mathematics education” (Geiger et al., 2023) and vol. 57(1) “Developing students’ reasoning with data and data-ing” (Fielding et al., 2025). In the current paper, we push beyond Geiger et al. by considering conceptions of citizenship in relation to statistics and data science education. We expand from Fielding et al. (2025) by unpacking how conceptions of citizenship lead to particular views of SL and DL, which are our core focus. We also go beyond Civic Statistics (see Engel & Ridgway, 2022 and other chapters in Ridgway, 2022) by connecting the rapidly expanding literatures from data literacy and AI literacy. Specifically, this paper aims to contribute to extant knowledge in two primary ways:

  1. 1.

    Explicating the links between different conceptualizations of citizenship (or citizen) with views of SL and DL and related constructs that involve data, broadly viewed, and

  2. 2.

    Conceptualizing ‘data literacy for citizenship’ and its key areas (i.e. criticality, data argumentation, modeling, and building a bridge to AI literacy) that go beyond prior conceptualizations of SL and DL.

To start conceptualizing DataLitCit we have developed the following working definition:

DataLitCit: Citizens’ knowledge, skills, values, and dispositions that enable them to collect, locate, interrogate, critically evaluate, use, and discuss data, statistical claims or analyses or data-based messages, and engage with the outputs and practices of data-driven systems (e.g., models and AI systems), while understanding their significance and impact on citizens’ lives and society at large, as well as their limitations.

This definition encompasses a broad view of the demands of DL, such as statistical, mathematical, and computational knowledge (procedural and conceptual) and skills needed to: investigate data-based claims, produce and communicate evidence, engage critically with data-driven systems, and participate meaningfully in civic and societal decision-making. We also focus on preparing learners for roles as citizens, rather than on preparing professional data scientists or data analysts.

Having developed a preliminary rationale for the need for DataLitCit and a working definition, the remainder of this paper is organized to unpack our initial definition to further conceptualize DataLitCit. We start in Sect. 2 by considering and integrating conceptions of citizenship (and their related demands and constraints), and of SL and DL (of which there are many). Drawing from this integrated perspective and from recent literature and our related work (Biehler, 2026, accepted; Biehler & Wilkerson, 2025; Biehler et al., 2025; Burrill & Pfannkuch, 2024; Gal 2024, 2025; Pfannkuch et al., 2026) we present four key areas (i.e. criticality, data argumentation, models and modeling, and a bridge to AI literacy) in Sect. 3 for consideration in conceptualizing DataLitCit. Section 4 briefly introduces the other 17 papers of the Special Issue and discusses their relation to DataLitCit. Finally, Sect. 5 discusses our two main contributions and Sect. 6 ends with limitations and implications for future research and practice. Figure 1 presents a conceptual framing of our argument for DataLitCit and mirrors the structure of our paper.

Fig. 1
Fig. 1
Full size image

Conceptual framing of key ideas for DataLitCit

2 Citizenship Education and Data Literacy

As mentioned in Sect. 1, conceptualizing what DataLitCit encompasses first requires engaging with the multitude of views of a “good citizen” and the ever-evolving demands of society. It is common to see the needs of citizenship as a rationale for SL and DL education. However, what is rare to see is explicit discussion of how conceptualizations of citizenship drive what SL or DL students need. Furthermore, as society changes so do the demands of citizenship and in turn, what kind of DL people need, making it a moving target changing over time and based on social, political, spatial, or historical setting. In this section, we first try to bring some clarity regarding conceptualizations of citizenship (Sect. 2.1), and then discuss selected ideas related to SL and DL (Sect. 2.2) to lay a foundation for integrating conceptions of citizenship and SL and DL (Sect. 2.3).

2.1 Conceptions of Citizenship

There are various competing goals of education, such as preparing ‘productive’ citizens, work force development, or social mobility (Labaree, 1997). We firmly situate DataLitCit in the goal of preparing citizens where education is a public good. There are various conceptions of education for democratic citizenship commonly used in mathematics and statistics education, such as personally responsible, participatory, and justice-oriented citizens, each with somewhat different aims (Westheimer & Kahne, 2004, see Fig. 2 for descriptions). However, there are some shortcomings with this basic typology in that it can be considered locally and be enacted in isolationist ways furthering nationalistic trends such as “[Insert country name] First” that are more exclusive of who is considered a citizen and whose concerns should be considered. Global citizenship typologies provide a broader view of citizenship (Oxley & Morris, 2013) that can rebuff nationalism and isolationism.

Fig. 2
Fig. 2
Full size image

Typologies of citizenship (based on Westheimer & Kahne, 2004; Oxley & Morris, 2013, p. 306) that consider movement between levels of community that people are situated within

Oxley and Morris (2013) describe eight types of global citizenship perspectives in two main groups: cosmopolitan citizenship (roughly related to personally responsible and participatory) and advocacy citizenship (roughly related to participatory and justice oriented, see Fig. 2 for descriptions). Cosmopolitan types of citizenship focus on globalization, universalities, and fundamental values of individuals across cultures and states with particular foci (e.g. United Nations Convention on the Rights of the Child (1989)). However, such globalized perspectives are what have led to nationalistic swings in once liberal democracies pushing back against what some view as cultural erasure through migration, amongst other issues perpetuated by global elites. The alternative category is advocacy citizenship, which focuses on anti-hegemonic, anti-colonial, grassroots, and relativistic/communal perspectives. Advocacy citizenship types are less cohesive than cosmopolitan and generally coalesce around their contrast to mainstream cosmopolitan perspectives and advocacy focused on particular issues such as bias, environmentalism, or spirituality/religion and countering mainstream or top-down perspectives such as those described as cosmopolitan.

The conceptions of citizenship (Fig. 2) are distinct but related or even overlapping and people may move fluidly between them over time and in different contexts. In considering a pluralistic democracy these different conceptions can coexist but require constant work to balance. In that sense, we argue citizens, policy makers, and educational leaders need to be aware of various views of citizenship and be open to balancing between such views and commitments in making decisions and engaging in their communities from local to state to global scales. Hence, we do not advocate for one type. Instead, we highlight the importance of considering what type of citizenship is targeted by any research and/or educational effort and making it explicit and coherently integrated into their efforts rather than just a vague notion in their rationale.

2.2 Statistical and Data Literacies

This subsection moves from conceptions of citizenship to examine SL and DL, before later integrating them. SL predates DL and has gone through many iterations beginning with Wallman’s (1993) definition “Statistical literacy is the ability to understand and critically evaluate statistical results that permeate our daily lives–coupled with the ability to appreciate the contributions that statistical thinking can make in public and private, professional and personal decisions” (p. 1). Following that, Gal (2002) defined SL as:

“people’s ability to interpret and critically evaluate statistical information, data-related arguments, or stochastic phenomena, which they may encounter in diverse contexts, and when relevant (b) their ability to discuss or communicate their reactions to such statistical information, such as their understanding of the meaning of the information, their opinions about the implications of this information, or their concerns regarding the acceptability of given conclusions” (pp. 2-3).

In parallel, Watson and Callingham (2003) conceptualized SL as a hierarchical construct consisting of six levels. These definitions have proven foundational and are commonly found in DL literature that relates to STEM education (Friedrich et al., 2024).

The use of the term DL is relatively new, beginning in the early 2000s and exploding after 2015 (Ghodoosi et al., 2023). DL comes out of a diverse set of disciplines, hence not widely agreed upon in educational goals and contexts (Ghodoosi et al., 2023; Risdale et al., 2015). For example, DL is often related with other literacies, such as digital humanities literacy, media, news, information literacy, digital literacy, AI, and machine learning literacy. Moreover, the datafication of disciplines has led to subject-specific DLs shaped by the social and natural sciences. DL work is published in a plethora of journals by scholars from a sweeping variety of disciplines (Vermeire et al., 2025). Moreover, concepts such as critical datafication literacy, personal data literacy, and data awareness, acumen, conscience, ethics, activism, and feminism are put forward in the discourse (Biehler et al., 2025).

One commonly used definition for DL is “the ability to collect, manage, evaluate, and apply data, in a critical manner” (Risdale et al., 2015, p. 2). We will not go into detail reviewing various definitions and frameworks as those discussions can be found elsewhere (Biehler et al., 2025; Ghodoosi et al., 2023; Vermeire et al., 2025). Despite the prevalence of theoretical definitions and frameworks, Friedrich et al. (2024) found that most scholars “only partially align their research with or incorporate these frameworks” (p. 10). Furthermore, in a systematic review of DL, Ghodoosi et al. (2023) found no clear disciplinary connection in research discussing DL. As Wilkerson (2025) points out from her bibliometric analysis, “the emerging field of data science education is relatively fragmented, and we are all still working to find one another” (p. 22). We note that these reviews do not discuss AI literacy and its connection to DL.

A recent Delphi study (Castañeda et al., 2024) polled experts in information and digital literacies about DL for critical citizenry that centers on the definition that “data literacy is about how people might understand, use, and engage with the data encountered in everyday life for the citizenry” (p. 299). Their framework includes four main elements: understanding data, acting on data, and engaging through data, with an ethics and privacy element crossing them all. This work highlights aspects that we do not see as prominent in SL and go beyond it, such as “data as persistent and potentially stored; data security; data surveillance; opportunities with data (weather, maps, etc.); my data” (ibid, p. 301) and “self-regulate own data footprint” (ibid, p. 302), which we have also seen studied as data awareness (Höper & Schulte, 2024). The notion of needing to be able to deal with dis-, mis- and mal-information (Carmi et al., 2020) is also an addition that comes out of relevant DL work outside of STEM education.

Overall, the terms SL and DL are used inconsistently in the literature drawing from a plethora of frameworks and definitions, which are often used interchangeably (Friedrich et al., 2024; Gould, 2017). According to Gould (2017), DL can be understood as an extension of SL that reflects the evolving role of data in society. It includes understanding how and why data are collected, how to analyze data from both random and non-random samples, and how to interpret and represent data to address real-world questions. It also encompasses awareness of data privacy and ownership, the provenance and storage of data, variations in data representation within computer systems, and basic aspects of predictive modeling (Gould, 2017, p. 23). Friedrich et al. (2024, p.5) found the main difference to be that DL had considerations in addition to those commonly found in SL including: data protection and ethics, story-telling from data, formulating hypotheses and identifying sources of data, collecting data, generating visualizations, presenting data, and applying data. The nuances in perspectives of SL and DL are important to consider especially in relation to varying views of citizenship, to which we turn next.

2.3 Integrating Conceptions of Citizenship and Data Literacy

Though the demands of citizenship are often cited as a reason why DL is needed, few frameworks directly link conceptions of citizenship, which have varying demands, and DL. A well-developed recent example related to SL that explicitly considers varying views of citizenship is Civic Statistics (J. Ridgway, 2022). Civic Statistics consists of a large body of empirical research supporting a conceptual framework that points to a diverse collection of skills, dispositions, and knowledge bases needed for citizens to engage with the statistics they encounter in their daily lives and make data-based decisions. Civic Statistics illustrates a general view of a citizen that provides a rationale for learning civic statistics, though the list of areas subsumed under Civic Statistics is broad and could fit multiple types of citizenship. For example, in conceptualizing Civic Statistics J. Ridgway et al. (2022) considered spectator, referee, and player roles, where the spectator aligns well with the notion of a personally responsible citizen and the description of a player leaves space for a justice-oriented citizenship perspective. We see Civic Statistics as a good starting point for considering DataLitCit, but it does not incorporate the emerging and rapidly expanding literature from DL, data science education, and AI literacy. Friedrich et al. (2024) point out in their bibliometric analysis of DL literature that there is a cluster of literature on citizen empowerment, but it comes from outside of STEM education often drawing upon media and information literacy for foundational texts and expertise (e.g. Castañeda et al., 2024; Carmi et al., 2020; Sander, 2024).

The demands of citizenship for SL have often been framed through the roles of consumer vs. producer. Consumer roles are often described through Gal’s (2002) description of the reading context of SL quoted earlier. Producer roles are often described as statistical enquiry (Wild & Pfannkuch, 1999). However, J. Ridgway (2022) distinguishes a wider range of roles such as spectator, referees and players, related to the level of engagement in social or political engagement, which is somewhat orthogonal to the producer/consumer distinction. In many contexts of life, the demands for DL can require citizens to take up a variety of the roles described to search for data-based information, critically consume information, and interpret texts with embedded mathematical, statistical, or data-based elements or claims. It is an open question how much and what types of knowledge are needed to be data literate though there is a wide range of literature that discusses this topic (i.e. Data Science Learning Progressions, 2025; Gal, 2002; NASEM, 2023; NCTM et al., 2024; NRC, 2026; J. Ridgway, 2022; Watson & Callingham, 2003), but as the demands of DL evolve so too will the knowledge needed.

In an increasing range of life contexts, citizens are not only producers or consumers, but additionally they are subjects of data practices (Höper & Schulte, 2026). People while interacting with a variety of digital and technology-based platforms (e.g. internet websites, applications, security cameras), whether knowingly or not (see Zuboff, 2019), often either provide personal data, or data are collected about them (e.g. regarding surfing habits, preferences, presence and behavior in physical or digital spaces, etc.). In these and related contexts, citizens are subjects of data-practices in the sense that various organizations collect, produce, process and analyze, possibly share or trade, and make decisions or act upon the collected data in various ways that may not necessarily be known or made explicit to the individuals concerned. Related literacy conceptions are data awareness (Höper & Schulte, 2024), personal data literacy (Pangrazio & Selwyn, 2019), and critical datafication literacy (Sander, 2024). However, these literacies can also be related to levels of citizen engagement as specified by Westheimer and Kahne (2004) or Oxley and Morris (2013). For example, consider Sander (2024) where a first level of critical datafication literacy is fostering systemic understanding of datafication, followed by encouraging critical thinking, which aligns with personally responsible citizenship moving into participatory. Sander’s (2024) third level is enabling learners to take different forms of action, which move into participatory and justice-oriented views of citizenship.

In summary, taking an integrated view of conceptions of citizenship, SL and DL is crucial for the coherent development of DataLitCit beyond the current often disjoint efforts. Furthermore, using the typology presented in Fig. 2 allows us to point to differences in DL aims based on views of citizenship and provides a tool for future work to situate itself explicitly in conceptions of citizenship. From the integrated perspective we present here we move to further conceptualizing DataLitCit in the next section.

3 Conceptualizing Data Literacy for Citizenship

Our goal in this section is to take an initial step in advancing conceptions of the key areas of DataLitCit, depicted in Fig. 1, which were identified from our reading of the literature as essential to add to extant views of SL and DL (Sect. 2). All the areas are intertwined, but they are not well integrated in educational research and SL and DL literature, hence bringing them together is an essential step.

The first key area we focus on is criticality. We aim to conceptualize this term explicitly in relation to types of citizenship in Fig. 2. The second key area is data argumentation (including evidentiary practice and misinformation), which is well discussed in media and digital literacies and information sciences, but has not been widely translated to STEM education settings. The third key area is understanding data-based models and modeling, which is one that has been discussed for many decades in STEM education, but has evolved with machine learning and large language models that work in predictive rather than descriptive or inferential ways. The fourth key area is a bridge to AI literacy, which is not synonymous with DL, but is related with no clear boundaries as GenAI models are trained using data including personal data and are predictive in nature drawing upon stochastics and mathematics. We do not see these four key areas as exhaustive for DataLitCit, but as starting points to what is a dynamic and evolving concept.

3.1 Criticality

The term critical is often used in citizenship education (e.g. Oxley & Morris, 2013; Westheimer & Kahne, 2004), education broadly (e.g. Freire, 1970), critical mathematics education (e.g. Geiger et al., 2023; Rubel et al., 2021; Skovsmose, 1994), modeling (e.g. Barbosa, 2006; Geiger, 2024) and SL and DL (e.g. Gal, 2002; Wallman, 1993; Watson & Callingham, 2003; Weiland, 2017; Wild & Pfannkuch, 1999) in a variety of ways. The term critical is also frequently in discussions of SL such as critical stance (Gal, 2002) and critical statistical or data literacy (Louie, 2022; Tygel & Kirsch, 2016; Weiland, 2017; Weiland et al., 2026), which draw upon the educational philosophy of Freire (1970) and critical mathematics education (Skovsmose, 1994) to consider reading and writing the world with data. Criticality spans SL and DL but varies drastically in its consideration.

Sometimes critical is merely used as an adjective to emphasize the importance of something, or more commonly to signify critical thinking or critical pedagogy/literacy, which overlap, but have distinct differences. Scholars do not always make clear how they are using the term critical, leaving it to the reader. However, different views of citizenship align with different views of criticality. For example, participatory or cosmopolitan global citizenship are focused on using fundamental values or common norms to make sense of and critically evaluate situations, whereas in justice-oriented or advocacy global citizenship criticality is viewed more in terms of questioning the status quo and looking at root causes of environmental degradation or injustice. Our goal here is to bring these ideas together to highlight the overall importance of criticality for DataLitCit and to describe different aspects of criticality for DataLitCit.

Critical thinking surfaces in prior SL work such as Wallman’s (1993) definition (Sect. 2.2) and Watson and Callingham’s (2003, p. 14) hierarchical framework for levels of SL. We also see it in dispositions such as a critical stance or healthy skepticism (Gal, 2002; Wild & Pfannkuch, 1999). Gal introduces the idea of worry questions as one of the critical skills people need to critically evaluate messages. The focus with critical questions is on evaluating messages of others. Wild and Pfannkuch (1999) came up with the notion of trigger questions that they dubbed the ‘creative cousin” of critical questions which they described as “questions which elicit pertinent context information, ideas and explanatory inferences from clients” (p. 242) as part of the enquiry process (i.e. production). There is also the notion of searching for and evaluating the provenance of data (Büscher, 2024) and searching for additional information on the subject (Gal et al., 2022; Pfannkuch et al., 2026).

Critical literacy has also developed a place in the work of DL and SL scholarship; growing significantly in recent years (Fernández-Pascual et al., 2025). Early on Gal (2002) discussed the notion of critical evaluation drawing from Freire (1970) to consider potential bias and misrepresentation that could be occurring in messaging whether intentional or not. Weiland (2017) builds from Gal (2002) and Wild and Pfannkuch (1999) to consider practices at the intersection of statistical and critical literacies such as identifying and interrogating social structures that shape and are reinforced by data-based arguments. Rubel et al. (2021) discuss critical mathematics education perspectives on data visualization including questioning what story an author is trying to tell and how they have chosen what patterns to highlight and which to background or obscure. Bailey and McCulloch (2023, 2026) consider critical statistical habits of mind such as integrating social, cultural, political, and economic understandings and one’s own positionality to make sense of injustice and engaging in active citizenry. Sander (2024) points to critical reflection and action based on how people’s behaviors are datafied and pushing back against such datafication or companies’ uses of it. Overall, critical literacy work considers subjectivity and questions power dynamics, status quo, and taking-for-granted norms in an effort to fight for liberation for historically marginalized groups and the transformation of structures of injustice (D’Ignazio & Klein, 2020; Louie, 2022; Tygel & Kirsch, 2016; Weiland, 2017).

We argue a productive approach is to view criticality as a system of lenses that provide different views coming from different epistemologies and conceptions of citizenship (Fig. 2). Literacy requires using multiple lenses moving between these different perspectives at different times and based on one’s view of citizenship. For example, only following statistical practices leads to blind spots in terms of taken-for-granted norms that can perpetuate injustices (D’Ignazio & Klein, 2020), while questioning the status quo all the time does not lead to building knowledge for meaningful progress or decision making. Towards a more holistic view of criticality, we propose an expanded notion of Gal’s (2002) worry questions to consider critical questions that come from different epistemologies and practices through the metaphor of a system of lens depicted in Fig. 3. The lenses we depict in Fig. 3 are not meant to be exhaustive, but examples drawing from some of the literature we reviewed earlier. The system of lenses can be used as a cross-cutting method for DataLitCit broadly.

Fig. 3
Fig. 3
Full size image

Example of a possible system of lenses to interrogate data

In summary, considering criticality as a dynamic system of analytical lenses directly underpins the core definition of DataLitCit. Rather than treating criticality as a monolithic skill, this multi-perspective model equips citizens with the flexible knowledge, skills, values, and dispositions required to navigate an increasingly datafied society. By seamlessly shifting between disciplinary critical thinking (e.g., applying “worry questions” to evaluate data provenance and statistical claims) and critical literacy (e.g., interrogating power dynamics, surveillance, and systemic bias), citizens are able to not only to rigorously interrogate, evaluate, and discuss data-driven systems and their outputs, but also recognize their profound impacts, limitations, and ethical dimensions—thereby fostering active, informed, and transformative democratic participation.

3.2 Data Argumentation, Evidentiary Practices, and Misinformation

From a citizenship perspective, data argumentation involves the recognition of the power of data to uncover new knowledge and information about real-world phenomena to learn more in the context sphere (Wild & Pfannkuch, 1999) and to improve societal situations. It also involves challenging others’ findings through systematic analyses using worry questions and multiple lenses (Fig. 3) to interrogate explicit and implicit evidentiary practices underlying data stories and claims. Data argumentation contributes to DataLitCit through three interrelated dimensions: understanding, engaging with, and enacting data argumentation (Castañeda et al., 2024; Gal et al., 2022). Citizens require the knowledge and competencies to understand arguments, the dispositions and habits of mind to engage with them, and the capacity to use data argumentation for advocacy.

Understanding data argumentation means drawing on many domains of knowledge such as contextual, statistical, and argumentation knowledge (Fielding, 2025), and in today’s world computational and GenAI knowledge (Sect. 3.4). From a scientific perspective, citizens need to acquire knowledge about evidentiary practices (Duncan et al., 2018) employed in discipline domains (e.g. health, environment). Evidentiary practice knowledge includes explanations and interrogation of evidence to support claims; evidence that provides information transparently and comprehensibly from rationale for investigation, data provenance and analytical methods employed, to visualizations and language used to communicate findings (Büscher, 2024). Citizens, however, need to be aware of the non-neutrality of data argumentation that can arise from representations, human subjectivities, and politics in data argumentation (Poirier, 2021). For example, inter-country and inter-state comparisons using official datasets are fraught with problems including different definitions for variables (e.g. child maltreatment deaths, homelessness), jurisdictional practices for data collection and storage, and definitions changing over time, sometimes due to political lobbying, and suppression about what counts, does not count, and is allowed to be collected. Poirier (2021) argued that citizens need to know that a dataset carries a genealogy of its cultural and political history, with the caveat that even though there may be powerful institutional structures behind how data are represented, there are always trade-offs for data producers seeking to standardize data for comparative purposes.

Critiquing others’ claims means knowing what constitutes quality evidentiary practices. For example, good risk evidentiary practices include communicating absolute risks alongside relative risks, offering both positive and negative frames (e.g. chance of survival, chance of death), and using both the probability and descriptive terminology (Spiegelhalter, 2025). Media reports, however, seldom follow good evidentiary practices. Complementary to knowing about evidentiary practices, is learning about the rhetorical communication strategies that people can use to persuade and lead people towards particular interpretations (Souza & Araújo, 2022).

Understanding data argumentation means realizing that decision-making involves possibilities not certainties, trade-offs, and weighing potential harms and benefits which can differ across communities. Decisions are based on current knowledge, with the proviso that new information might require changes. Predictions are based on statistical models, current assumptions, and can result in changing decisions about prescribed (in)actions (e.g., Covid-19). The acceptance of uncertainty occurs elsewhere, for example, in outcomes of health screening tests, effectiveness of vaccinations, side effects of drug treatments, and modeling of cyclone pathways.

Engaging with data argumentation requires the ability to interact with diverse media and fora (e.g. blogs, vlogs, journalist items, in-person, social media posts, podcasts, AI). In these fora citizens need the skills and habits of mind to react to claims and counterclaims, and to converse with others. Engagement depends on the ability and the disposition to interact with diverse points of view, and an openness to the perspective of others. Alongside being open to others’ perspectives is recognizing how one’s own socio-political perspective (Bailey & McCulloch, 2023), values, biases, emotions, life experiences, culture, positionality in society, personal motivations, and world view lens (e.g. Eurocentric, American, Indigenous), influences one’s interpretation and critiques of data-based arguments. Considering alternative explanations and perspectives, strength of evidence, and the potential impacts and consequences of findings, claims, recommendations and decisions, are also necessary.

Media reports, however, do not provide adequate information to form a reasoned opinion, requiring the engaged citizen to seek out further information (Bailey & McCulloch, 2023; Pfannkuch et al., 2026), an important habit of mind for DataLitCit. Sourcing further information is problematic, with some academic journals not having open access, commercially paid research unavailable, and government suppression of data. Coupled with difficulties in accessing sources is ascertaining their trustworthiness. With citizens losing trust in science, institutions, social media, and others different from them (Edelman Trust Barometer, 2026) and being too overwhelmed by the media ecosystem to distinguish fact from fiction, questions arise about who and what do citizens trust and how can trust be built (Pullinger, 2020).

Aligned to distrust is dis-, mis- and mal-information, an issue that Carmi et al. (2020) said needs to be addressed when considering data literacies for citizenship. The distortion of information is determined by the intent of those creating or distributing the information. Dis-information is false and aims to create harm, mis-information is false but not distributed to cause harm, whereas mal-information is based in reality with the aim of inflicting harm. Politicians are known for cherry-picking data to support their claims. Today, however, political actors deploy information distortion tactics (accelerated by GenAI) to discredit reported facts they dislike, resulting in fact checker organizations, increasing polarization in society, and citizen misperceptions. Carmi et al. (2020) argue that to combat the distortion of information, citizens must become proactive rather than passive consumers.

Enacting data argumentation through advocacy and activism is necessary for empowering individuals and communities to understand the inter-play between data and society and its implications for social change and transformation (Castañeda et al., 2024; Bailey & McCulloch, 2023; Sander, 2024) and decision-making. Aligned with justice oriented and advocacy views of citizenship and critical literacy perspectives on criticality, activism and advocacy involve promoting collective awareness about how to use data for solving local issues such as improving the school canteen (Souza et al., 2020) and addressing inequities, injustices and environmental issues at the societal level. Proactively speaking out, influencing stakeholders, taking civil action, and becoming agents of change, however, involves considering questions such as “Am I allowed to do or say this?”, “Are the findings socially and culturally acceptable?” and “Will the desired political narrative compromise my ethical standards?” (see Büscher, 2024, p. 127). Proactive citizens can then become conflicted if the social and political environment gives rise to concerns about repercussions and fears about speaking out, resulting in self-censoring and a reluctance to speak truth to power.

In summary, data argumentation for DataLitCit involves: understanding multiple knowledge bases about evidentiary practices employed by disciplines, media and political actors; developing dispositions to engage and interact with multiple perspectives including reflecting on one’s own lens on data-based information; and countering information distortion through developing the capacity to use data argumentation for proactive participation and advocacy for social change in cognizance of the impact and consequences of findings on citizens’ lives.

3.3 Models and Modeling

This section highlights our view that DataLitCit also requires that citizens can engage with and critically understand real-world messages that refer to data-based models, or products of statistical modeling. Such messages and products are prevalent across multiple key life contexts, such as media, service consumption, and the workplace, and using AI systems (Sect. 3.4). Models are usually viewed as representations aiming to capture and reflect selected key elements in a reality and the relationships between them (Doerr & English, 2003). The arena of statistical and mathematical modeling has received much scholarly and educational attention (e.g., Burrill & Biehler, 2011; Cevikbas et al., 2022; Gould, 2024). This includes discussions both of differences and connections between mathematical and statistical models (e.g. Kawakami & Saeki, 2024), and their importance for shaping or formatting people’s view of social phenomena and issues of power (Skovsmose, 1994; Barbosa, 2006; Geiger, 2024).

In media consumption contexts, messages related to models and modeling are quite prevalent. For instance, analyses by the ProCivicStat project (Gal et al., 2022) suggest that Civic Statistics reported by news channels are often based on products of official statistics producers and refer to various types of models, such as about infant mortality (health), poverty (economic equality), or climate change (environment). Gal and Geiger (2022) have examined 300 media items from four countries related to the COVID-19 pandemic, and documented many reports involving models and modeling, such as predictions, projections, and statistical scenarios they also identified demands for criticality (Sect. 3.1), e.g., regarding the quality of the evidence produced by models.

Service consumption contexts likewise involve diverse types of models and modeling, such as in recommender systems used in online purchasing, video streaming, or financial services. Recommender systems involve rather complex data-based modeling processes; they combine (personal) data retrieved from or about specific customers with data retrieved from vast numbers of other customers and mix of statistical techniques and machine learning methods (and possibly AI systems, see Sect. 3.4) to classify users and predict possible courses of action (Véras et al., 2015).

There is a need to identify fundamental knowledge elements for inclusion in a framework of DataLitCit that can be practically addressed by instructors (Fergusson & Pfannkuch, 2022) and prepare future citizens for the rich array of models and modeling across personal and civic life. Gal (2024) has argued that citizens should become familiar with at least two clusters of ideas. The first relates to the building blocks (or professional decisions) that underlie all data-based models:

  • Conceptual decisions (assumptions and rationales regarding the nature of the target phenomenon, relevant variables),

  • Data collection decisions (how to collect the data, feasible and legitimate data sources),

  • Computational decisions (how to combine or integrate the different variables in the model),

  • Reporting decisions (ways to present and explain the results so that the target audiences can derive meaning and insights from the modeling).

Gal (2024) also argues that citizens should become familiar with the core purposes or uses of statistical models, such as: aggregation (e.g. statistical indicators such as infant mortality rate, Gini coefficient, or average wait time), prediction (e.g. using regression models for predicting future sales or accidents, or classifying clients into different risk groups) and simulation, long-range forecasting and risk estimation (e.g. scenarios related to demographic projections or global warming). However, systems that rely on machine learning and AI (Sect. 3.4) involve more complex purposes and frameworks of models and data-based modeling, whether people are aware of it or not.

A core educational challenge is thus how, and how much, it is necessary to explain to the ‘average citizen’ the mathematical or statistical logic underlying messages or outputs that are based on data-based models and modeling (Gedikli et al., 2014; Höper & Schulte, 2024). Some understanding of the underlying modeling logic may arguably be needed to help comprehension, openness and trust in presented messages that involve models and modeling, whether in explicit or implicit ways (e.g. Höper & Schulte, 2026). A didactic challenge emerges because of ‘masking’, i.e., actual usage of the terms ‘model’ or ‘modeling’ may range from being explicit (e.g. “Scientists used modeling to…”), to partially obscure (“Our projections show that…”), to fully hidden when using GenAI. Such ‘masking’ affects citizens’ ability to activate criticality or engage in data argumentation.

In summary, DataLitCit should include knowledge regarding the purposes and building blocks of models and modeling, alongside a critical awareness that models are constructed representations based on selections, assumptions, and purposes rather than straightforward or value-free reproductions of reality. Knowledge and awareness demands in this regard go beyond those discussed in the literature regarding SL and DL, hence subsumed in our conceptualization of DataLitCit.

3.4 A Bridge to AI Literacy

The emergence of AI, particularly machine learning, has changed the character of models and modeling. Machine-learning models are data-driven, typically contain many parameters determined from data, and are often less transparent than traditional statistical models (Biehler et al., submitted). These developments require DataLitCit to encompass citizens’ engagement with the data-related foundations, outputs, and societal uses of AI. At the same time, AI literacy draws on several literacies beyond DL, including mathematical, ethical, media, computational, linguistic, visual, scientific, design, and domain-specific literacy (Chiu, 2025, p. 3226). The goal here is not to absorb AI literacy entirely into DataLitCit, but to highlight key overlaps where critically engaging with AI depends on data-related knowledge and practices.

AI and algorithmic decision systems have become relevant to citizens’ private, public, and professional lives. Yet conceptions of AI literacy often lack an explicit and elaborated conception of citizenship. The OECD and European Union (2026, p. 53) framework refers to citizenship only as digital citizenship: “The capacity to participate actively, continuously and responsibly in digital environments (local, national, global, online) at all levels (political, economic, social, cultural and intercultural.” Its detailed competencies implicitly address different forms of citizenship in the sense of Westheimer and Kahne (2004). Many other AI literacy frameworks, however, specify a common set of competencies for different areas of life without clearly distinguishing democratic citizenship from preparing for the workforce. For example, Long and Magerko (2020), Chiu (2025), and Almatrafi et al. (2024), place more emphasis on workforce preparation than on citizenship. DataLitCit can complement these frameworks by foregrounding the civic purposes, societal consequences, and power relations associated with data-driven AI.

Researchers in AI literacy have identified the DL knowledge and competencies required for an adequate understanding of AI (Chiu, 2025). Conversely, researchers in SL and DL have expanded their conceptions to address the changing roles of data and AI in society and individuals’ everyday lives (Biehler, 2026, accepted; Gould, 2017). AI literacy also intersects with personal data literacy, because some AI systems rely on the collection and processing of personal data (Pangrazio & Selwyn, 2019), and with critical datafication literacy, because such systems transform social practices and institutions into data and make them subject to computational analysis and intervention (Sander, 2024). To clarify how these developments elaborate DataLitCit, it is useful to distinguish between pre-GenAI and post-GenAI conceptions of AI literacy.

Pre-GenAI conceptions developed alongside automated decision systems based on machine learning. Their prototypical situation is an algorithmic system whose construction, application, and evaluation depend on data (Long & Magerko, 2020). In Touretzky et al.’s (2019) five big ideas for AI in education, idea three “computers can learn from data” is most relevant to SL and DL. Nevertheless, machine learning introduces forms of inference and predictive modeling that differ from traditional statistical modeling and inference (Breiman, 2001; Biehler, 2022). Although DL is indispensable for understanding these processes, holistic approaches connecting AI and DL remain underdeveloped (Olari & Romeike, 2024).

For DataLitCit, this means that citizens should understand how algorithmic systems use data to classify, predict, and recommend. Relevant knowledge includes the role of training data in constructing a model, the use of separate test data in evaluating it, and the dependence of outputs on the quality, composition, and provenance of data. Citizens should also recognize that predictive accuracy and error differ from conventional statistical considerations of model fit. Bias, fairness, and transparency are central to evaluating algorithmic decisions and their consequences. These competencies enable citizens to interpret, question, and contest AI-supported decisions in areas such as credit scoring, recommender systems, employment, and risk prediction (Casal-Otero et al., 2023; Long & Magerko, 2020). Experience with constructing simple machine-learning models may support this understanding, but the primary civic objective is critical engagement with their uses and effects.

DataLitCit contributes several foundations to AI literacy: understanding data types and sources, collection and cleaning, manipulation and visualization, basic analysis, and data quality. These competencies support the identification of biases in datasets, interpretation of AI outputs, and recognition of the limitations of data-driven conclusions (Chiu, 2025). From a citizenship perspective, they must be connected to questions about who collects and controls data, whose interests are represented, which groups bear errors and harms, and how automated decisions can be challenged.

GenAI changes this relationship because it not only supports decisions but also participates in producing texts, images, code, explanations, and other forms of knowledge. The prototypical situation is active interaction with a generative system in producing or revising digital artifacts. Through prompting, iterative refinement, and evaluation, users become co-producers. This requires awareness of how such systems function, their limitations, and their epistemic risks, even though their training data and internal processes remain largely in the background. These competencies apply when GenAI supports learning across subject areas, including mathematics (Pepin et al., 2025).

Of particular relevance to DataLitCit is the use of GenAI to locate datasets, produce visualizations, conduct analyses, write programs, and generate data-based reports (Ellis & Slade, 2023). These applications shift parts of the analytical process from humans to AI systems. Citizens must therefore supervise AI-mediated data practices, validate their products, and determine whether they can responsibly inform reasoning and decisions. This entails checking sources, methods, calculations, representations, and interpretations rather than treating fluent outputs as trustworthy evidence.

Post-GenAI DataLitCit consequently extends beyond conventional statistical inference. It includes managing epistemic uncertainty in AI-generated representations, distinguishing empirical data and analyses from synthetic products, and evaluating AI-mediated knowledge production. Analyses may be fabricated or based on nonexistent data, while opaque procedures may violate standards of transparency and reproducibility. Competencies in SL, evidentiary reasoning, and data argumentation therefore acquire renewed importance in evaluating AI-generated reports and responding to low-quality or manipulative AI-generated content in public discourse.

Post-GenAI DataLitCit also connects personally responsible and justice-oriented citizenship (Westheimer & Kahne, 2004). Responsible practices include reflective prompting, verification, disclosure of AI use, and transparent attribution. A justice-oriented perspective examines the conditions under which GenAI operates, including concentrated power among foundation-model providers, the extraction of personal and public data, environmental costs, unequal distributions of benefits and harms, and the normalization of AI-generated knowledge in public discourse. DataLitCit thus connects individual responsibility with critical examination of the structures governing AI-mediated data practices (Weidinger et al., 2021).

In summary, AI literacy elaborates DataLitCit by extending it from conventional data and statistical messages to the data foundations, outputs, and practices of machine-learning and GenAI systems. DataLitCit encompasses understanding how data shape algorithmic models, evaluating and contesting automated decisions, supervising AI-mediated analyses, and assessing their epistemic, ethical, and societal consequences. It frames these competencies explicitly as capacities for responsible and justice-oriented citizenship.

4 Special Issue Contributions to DataLitCit

A key contribution of this paper is introducing the construct of DataLitCit as a target for research and goal for educational efforts for all citizens. This section provides further perspectives regarding the key areas encompassed in DataLitCit by illustrating how the 17 other papers in this Special Issue of ZDM relate to and support the key areas introduced in Fig. 1.

4.1 Integrating Conceptions of Citizenship and Data Literacy

An important implication of DataLitCit is that scholars in SL and DL must explicitly identify the conception of citizenship framing their work, moving beyond superficial rationales. While existing typologies help link citizenship to DL, substantial room remains to explore deeper connections. Work in this issue illustrates this potential:

  • Critical Data Education: Weiland et al. (2026) trace CDE’s roots to Freire’s (1970) critical pedagogy and Skovsmose’s (1994) critical mathematics education, detailing how distinct socio-political origins shape educational frameworks.

  • Relational Citizenship: Gargroetzi et al. (2026) center civic activity around relationality and collective answerability (“What should we do?”) rather than specific political positions.

  • Applied Learning: R. Ridgway and Reading (2026) show that integrating evaluative, social, and ethical considerations into technical coursework heightens students’ awareness of team dynamics and multifaceted AI impacts.

Each of these papers presents different perspectives on how citizenship can be considered explicitly in DL work, which is important for adding coherence to the field and can serve as models in future research.

4.2 Perspectives on Criticality

Fostering criticality requires designing and implementing effective learning experiences across formal and adult education. Although critical thinking is a common mathematics education objective, it does not always occur in practice, and critical literacy is seldom taught explicitly in mathematics courses. As the role of data in society evolves and widespread surveillance grows, curricula must integrate emerging frameworks and educators must engage students in criticality holistically (Sect. 3.1). Work in this issue illustrates this potential:

  • Teacher Preparation: Bailey and McCulloch (2026) show that preservice math teachers rarely exhibit advanced criticality or Critical Statistical Literacy Habits of Mind (CSLHM), highlighting the urgent need to intentionally cultivate these skills in future educators.

  • Bridging Literacies: Höper and Schulte (2026) offer a conceptual framework linking computer science data awareness with critical statistical literacies, pointing toward practical classroom integration.

  • Ethical Frameworks: Register et al. (2026) address the moral dilemmas of modern data by outlining a framework grounded in social norms and moral theories to characterize student ethical reasoning.

These studies point to ways to empower students to proactively reflect on and critique the ethical and societal issues underpinning data-based information.

4.3 Perspectives on Data Argumentation and Evidentiary Practices

Data argumentation and evidentiary practices—extensively studied in science education and media literacy—require further translation into mathematics and statistics curricula. Learners need robust statistical thinking to construct quality arguments, evaluate mis- or disinformation, and adapt communications for diverse audiences. Studies in this issue highlight key approaches and challenges in developing these skills:

  • Checking Media Claims: Digan et al. (2026) adapt science evidentiary practices to show how students can evaluate controversial media claims about poverty using big data sets.

  • Civic Data Visualization: Gómez-Blancarte (2026) demonstrates a web-based activity structure that helps teachers prompt beginner students to visualize, discuss, and reason about civic data.

  • Structuring Arguments: Louie et al. (2026) emphasize that teaching statistical argumentation requires explicit focus on argument structure, numerical evidence, and clear warrants when confronting social claims.

  • Communicating Evidence: Kazak et al. (2026) analyze preservice teachers’ data communications to reveal key gaps in argumentation knowledge, while Podworny (2026) explores using social media posts to encourage students to interact with alternative perspectives.

  • Dispositional Factors: Engel et al. (2026) show that merging statistical and socio-political learning boosts statistical self-efficacy, whereas St-Denis et al. (2026) find a negative correlation between DL and data anxiety—underscoring the need to mitigate anxiety to improve argument evaluation.

These papers stress the need for empirical classroom research to establish effective methods for enhancing students’ evidentiary practices, dispositions, and debate skills.

4.4 Perspectives on Models and Modeling

Although modeling is a core topic in mathematics curricula, it is treated inconsistently across disciplines. Mathematics education often treats modeling as a generic practice without differentiating types, and data-based modeling remains the exception rather than the rule despite decades of supportive research. Recent studies in this issue address these gaps and highlight key directions for practice:

  • Computational Notebooks to Support Data Modeling: Wilkerson et al. (2026) show that using Jupyter notebooks enables pre- and in-service teachers to leverage computational actions to make sense of complex scientific phenomena with minimal programming experience.

  • Simulation-Based Inference: Fernandes et al. (2026) demonstrate that simulation-based inference supports student thinking and argumentation when investigating sociopolitical contexts.

These papers suggest ways to improve the training of teachers in developing their data modeling skills. This is an important area given the increased emphasis on data modeling in secondary mathematics curriculum and the increased complexity of models that impact citizen’s daily lives.

4.5 Perspectives on bridges to AI Literacy

The emerging intersection of AI literacy and DataLitCit requires significant research, balancing technical understanding with practical application. AI literacy spans two main areas: understanding foundational technical mechanics (such as how GenAI is trained and its limitations) to support critical evaluation and leveraging AI tools to scaffold complex data exploration and modeling for everyday citizens. Key research and challenges in this domain in papers in this issue include:

  • Demystifying AI Models: Fleischer and Biehler (2026) demonstrate methods for making machine learning models transparent rather than “black boxes,” showing how model training can be taught effectively to school-aged students.

  • Student Discourses on AI: Investigating how students view predictive models and their societal impacts, Andersson et al. (2026) identified four distinct student discourses—ranging from using mathematics for societal critique to viewing math and ethics as completely incompatible—highlighting how existing beliefs shape AI-informed citizenship education.

These papers present exciting possibilities and disturbing challenges in this area that need further consideration. For example, evolving AI tools could reduce technical barriers in statistics and data science, allowing citizens to engage in more sophisticated data analysis and argumentation than was previously feasible. However, the prevalence of AI-generated content (“AI slop”) used for political misinformation creates an urgent need for strategies to help citizens detect and critique subtle, AI-driven manipulations.

5 Discussion and Contributions

This survey paper, and the Special Issue that it introduces, were motivated by an identified gap in conceptualization and empirical research regarding SL and DL in relation to conceptions of citizenship. This is a multi-disciplinary arena that many scholars and education practitioners interested in statistics and data science education, mathematics education, adult education, and related disciplines, have contributed to over several decades, but not in coordinated ways. Furthermore, the role of data in society is changing rapidly and requires rethinking SL and DL. This survey paper makes two contributions towards addressing the gap identified: integrating conceptions of citizenship and statistical and data literacies (Sect. 2), and conceptualizing DataLitCit (Sects. 3 & 4). In this section we holistically discuss these contributions.

5.1 Integrating Conceptions of Citizenship and Data Literacies

We view DataLitCit as a public good, serving different forms of democratic citizenship. We rely on a typology of conceptions of citizenship (Fig. 2), which can support situating SL and DL scholarship within educational practices. We do not advocate for a specific type of citizenship. Instead we envision there could be different variants of DataLitCit to support different conceptions of citizenship. It is the explicit consideration and integration of SL, DL and conceptions of citizenship coordinating the key areas (Fig. 1) that define the structure of DataLitCit.

While DataLitCit overlaps with established frameworks of SL and DL (e.g., Castañeda et al., 2024; Gal, 2002; Gould, 2017; J. Ridgway, 2022; Weiland, 2017), it differs from them by explicitly considering their necessary integration with and differing alignment to conceptions of citizenship outlined in Sect. 2, which remains a rarity in DL literature. We view this oversight as problematic since differing types of citizenship may require different combinations of specific knowledge, skills, values, and dispositions to fulfill them. For instance, a personally responsible citizen primarily requires the competencies to understand basic data messages relevant to their immediate context and maintain awareness of personal data privacy threats. Conversely, they have little need for the structural focused capabilities required of a justice-oriented citizen—such as understanding for the community relevant data-based reports, gathering and analyzing community-level data, constructing data-driven policy arguments, or sacrificing personal advantage to advocate for equitable community outcomes.

In summary, ignoring the underlying assumptions of different conceptions of citizenship and their demands on DL ultimately impedes DL initiatives from building a coherent knowledge base for impacting meaning change in educational contexts. DataLitCit helps to address this issue by providing a typology of conceptions of citizenship to support researchers in coherently aligning the knowledge, skills, values, and dispositions they are studying with the type(s) of citizenship they state are their goal so there is integration and coherence between them.

5.2 Conceptualizing DataLitCit

The necessity for a distinct conceptualization of DataLitCit arises from a persistent gap in the SL and DL literature, which rarely addresses the nuances of integrating statistical and data practices and education with the demands of democratic citizenship. While foundational initiatives like Civic Statistics (J. Ridgway, 2022) established early connections, the rapid expansion of data science as a standalone field, and emergence of many technological innovations (e.g., cloud computing, GenAI) has fundamentally reconfigured the relationship between data, society, and civic life. This evolving space requires a new conceptualization that provides some structural coherence but is also flexible to evolve with the rapidly changing landscape of data and society.

We proposed a working definition for DataLitCit (Sect. 1), which builds on prior conceptualizations of SL and DL, but goes beyond them based on our integration of conceptions of citizenship and statistical and data literacies (Sect. 2). We also identified four additional key areas of DataLitCit—criticality; data argumentation and evidentiary practices; models and modeling; and creating a bridge to AI literacy (Sect. 3). In Sect. 4, we show that all 17 papers in this ZDM Special Issue highlight how one or more of the key areas in the definition of DataLitCit are important and illustrate how they can be addressed in practice at various educational levels. Most of these areas have been discussed in SL or DL scholarship before in some form, but they are not connected and integrated like in DataLitCit.

We also point to the necessity to expand conceptual frameworks beyond prior SL and DL scholarship, in four areas (see Fig. 1). For example, DataLitCit extends to deal with data-based models and modeling because many public messages, services, and decisions rely on models that aggregate, predict, classify, simulate, recommend, or forecast. Citizens should understand that models are selective and simplified representations of reality whose outputs depend on decisions about how phenomena are conceptualized, which data are collected, how variables are combined, and how models impact people’s lives and society at large. Citizens do not necessarily require advanced technical expertise regarding models and modeling, but they should be able to identify assumptions, uncertainty, limitations, and consequences and recognize how models shape knowledge, decisions, opportunities, behavior, privacy, and power—even when the models operate invisibly. This also points to the focus on literacy for citizenship not for the workforce.

DataLitCit also includes a critical and reflective stance toward data, data-based messages, and data products in relation to society and people’s everyday lives. It therefore extends beyond assessing technical competencies to questioning power relations, established norms, and the ways data may reproduce or challenge injustice. DataLitCit also extends the interpretation or communication typically discussed in SL and DL to examine how evidence is described and used to support claims, recognize rhetorical framing and information manipulation, assess the trustworthiness of sources, and distinguish mis-, dis, and mal-information.

A key area of DataLitCit that is not seen in previous conceptions of DL and SL is the connection to and integration of aspects of AI literacy. We do not see DataLitCit as encompassing all of AI literacy which is an enormous and rapidly changing field in its own right. However, there are crucial connections through machine learning and GenAI’s use of data that need to be considered and incorporated into DL. Citizens should understand how training and evaluation data influence classifications, predictions, recommendations, and automated decisions and should be able to assess accuracy, error, opacity, bias, fairness, and transparency. In relation to GenAI, they should be able to prompt and make sense of AI-generated data analyses, visualizations, and reports; recognize hallucinations; distinguish empirical from synthetic products; and consider uncertainty, reproducibility, attribution, and privacy. This also entails examining broader societal consequences, including unequal distributions of benefits and harms, corporate concentration of power, environmental costs, the growing influence of AI-generated knowledge on public discourse, and its strong formatting power on society today.

5.3 Towards an Elaborated Definition of DataLitCit

To solidify the contribution of DataLitCit, discussed in detail in Sect. 3, we expand upon our working definition to present a more elaborate and complex definition that captures the broader issues presented in the literature we discussed.

DataLitCit encompasses citizens’ knowledge, skills, values, and dispositions that enable them to

  • locate data or data-based information relevant to their lives or the world they live in;

  • make sense of data-based messages/products encountered in daily life to gain insights about the world and make informed decisions;

  • evaluate data and data-based messages for their quality and trustworthiness and to identify mis-, dis-, mal-information;

  • critically interrogate data and data-based messages using multiple social, historical, cultural, political, and epistemological lenses to better understand the world, make informed and ethical decisions, and/or take transformative action;

  • be aware of how data are collected on people, how third parties commonly use data for commercial or governing purposes, and how to protect one’s data privacy or advocate for structural changes to protect personal data;

  • ask and answer questions about relevant real world phenomena using data including asking critical questions when querying all phases involved in producing data-based information;

  • understand how data investigative, modeling, algorithmic, and AI processes treat data in different ways and how they work from a system perspective to impact people’s day-to-day reality and society and how those impacts vary for different people;

  • understand the limits of data-based evidence and products – based on current knowledge, uncertainty, assumptions, caveats, training data, etc.—and be open to changing decisions/actions based on new information and insights;

  • be aware of how one’s values, life experiences, emotions, culture etc. influence data creation and interpretation and of one’s biases and be willing to challenge one’s preexisting beliefs based on data-based evidence;

  • communicate data-based insights ethically, transparently, and persuasively to improve one’s life, community, society, or world.

This definition is expanded, but not exhaustive, and it is flexible to fit various conceptions of citizenship. For example, what ethics a person uses to make decisions based on data may change based on the ethics associated with their conception of citizenship. Someone from an environmental global citizenship perspective may push back against using AI because of its environmental impacts on water, electricity, and land usage, while someone from an economic perspective may embrace using AI for its perceived increase in economic production and profit generation. We acknowledge that our conceptualization of DataLitCit is ambitious, and in need of further unpacking, which points to limitations of its current conceptualization and directions for future work that we discuss in the next section.

6 Implications and Future Directions

We end our paper by discussing limitations and implications for future theoretical and empirical research and educational practice.

6.1 Limitations

Our description of DataLitCit and its areas should be viewed as a tentative beginning that needs to evolve over time and through dialogue across disciplines, especially because of the dynamic nature and changing demands across all of the areas listed earlier. For example, the notion of data is rapidly expanding to include new sources of data (see Fielding et al., 2025) such as administrative data or data emerging from sensors and other sources. GenAI systems and large language models are evolving in leaps and bounds, and the notion of what a ‘model’ may mean is expanding yet also becomes fuzzier in such contexts. That means that while our general characterization of DataLitCit may prove stable, underlying details “under the hood” such as those in Sect. 3, need to be revisited and updated frequently. It also means that forms of criticality or data argumentation for citizens will also need to change over time, especially as models and GenAI continue to evolve rapidly and increasingly shape society (O’Neil, 2016; Skovsmose, 1994; Zuboff, 2019).

Another limitation is that we present an idealistic conceptualization of DataLitCit, which would be challenging to impossible to achieve in most current educational contexts outside of a dedicated postsecondary degree. Much work is still needed to ground this conception and unpack its complexities. We mapped all the papers in this Special Issue to the key areas, but as other scholars engage with our conceptualization of DataLitCit they may identify further areas of importance. A related limitation due to time and space is that we focus on top level conceptualization of DataLitCit and its key areas without diving deep into unpacking all the nuances of integrating conceptions of citizenship and DL and the specific knowledge, skills, values, and dispositions related to each. We also do not parse conceptual trajectories based on age, educational-level, ability, etc. For example, it is necessary (see Gal, 2025) to develop relevant learning goals and teaching trajectories that would serve learners (including adult learners) who have low or relatively low basic competencies in literacy and numeracy. These limitations all point to areas of needed future research.

6.2 Implications and Future Work

As mentioned in previous sections, there is the need to consider numerous implications of DataLitCit for future theorizing, research, and instruction. First, the prior discussions point to the need for additional theoretical work regarding DataLitCit. Expanding the conceptualization of DataLitCit to address limitations mentioned in the previous section is important to guide further research and instructional efforts, yet it is a challenging undertaking. For example, the rapidly increasing technical complexity and use of data by systems that combine data science, machine learning and AI make understanding them increasingly challenging and possibly out of reach for many citizens. This poses both challenges and limitations on how DataLitCit can be conceptualized at present and in the future, and for educational design.

Research is also needed to support fine-tuning and deepening our understanding of education (teaching and learning) for DataLitCit. This includes basic research on questions such as:

  • How do people learn concepts such as bias in model training and evidentiary practices for citizenship?

  • What are developmental trajectories for knowledge, skills, values, and dispositions of DataLitCit?

  • How much do people need to learn about engaging in data enquiry to be able to critically engage with and interrogate data-based messages?

There is also a need for applied research to test basic research in pragmatic ways in actual learning settings and work done at scale. This includes research questions such as:

  • How can learning environments be designed to support K-12 students or adults in learning key knowledge, skills, values, and dispositions for DataLitCit?

  • How can DataLitCit be incorporated into school or general undergraduate curriculum?

  • How can instructors be supported to teach DataLitCit?

In general, there is also a need for developing measures for aspects of DataLitCit to help scale research and measure the impact of interventions. Additionally, there is a need for research on educational policy to support DataLitCit. Research on DataLitCit requires teams of researchers with diverse expertise across disciplines to work together to study this complex interdisciplinary space.

Finally, we consider implications for instruction and instructional systems. Much of what we subsume in DataLitCit is not commonly in the school curriculum or general undergraduate education and differs to some extent from the visions in the growing calls and views about what data education for the general public should entail, from professional organizations, public agencies, and think tanks (e.g. Data Science Learning Progressions, 2025; IDSSP, 2019; NASEM, 2023; NCTM et al., 2024; NRC, 2026). Considering the audience of ZDM, we note that mathematics and statistics curricula have not kept pace with emerging trends (Burrill & Pfannkuch, 2024), such as how (future) citizens can be prepared to engage with increasingly pervasive data, navigating social media feeds and understanding their underlying algorithmic drivers, or using AI to improve real-world tasks and decisions (Biehler & Wilkerson, 2025). Such complex developments require educational responses or reforms (Gould, 2017; J. Ridgway 2016, 2022).

The conception of DataLitCit we present is idealistic and in need of pragmatic grounding as it faces various realities and pressures in educational systems, given fully packed curricula in all subjects/majors, restricted access to technology, and gaps in teacher preparation (Schreiter et al., 2024). For example, tools to help in planning curriculum like those seen in the radar plots from Civic Statistics (J. Ridgway, 2022) or tools for planning AI literacy from Long and Magerko (2020) are needed. Moreover, teaching DataLitCit in an interdisciplinary way requires instructors with an open mind and relevant backgrounds, which is challenging even when instructors of different disciplines collaborate. The papers in this Special Issue present some first steps toward DataLitCit by presenting empirical examples of different aspects of it (Sect. 4), but much work is still needed. We hope the contributions of this survey paper and associated Special Issue move conversations towards viewing Data Literacy for Citizenship as a separate and valuable main goal, given the criticality of DataLitCit for individuals and societies alike.