Comparing Digital,Data,and AI Literacy: A Narrative Review

Abstract

This article explores the growing need to understand what skills are required to navigate rapid technological change driven by digitalisation, datafication, and AI – both within organisations and education, and in citizens’ everyday lives. Through a narrative literature review of digital, data, and AI literacy, we analyse their defining components, synthesising these approaches into an integrated framework. Our findings position digital literacy as a foundational concept essential for understanding and engaging with both data and AI literacy. While digital literacy equips individuals with basic technological skills, data literacy and AI literacy require more specialised knowledge. Data literacy involves using data for decision-making and problem-solving, whereas AI literacy extends to understanding AI systems, their functions, and their social, ethical, and legal implications. The article identifies three interrelated dimensions across all literacies: technical, critical, and communicative-cognitive. While technical and critical dimensions are well-documented, the communicative-cognitive dimension, essential for interacting with and cognitively relating to technological resources, remains less explored. We argue that educational programs must prioritise technical, critical thinking, and communicative-cognitive skills to cultivate comprehensive digital, data, and AI literacy. Finally, we raise critical questions about how public sector officials practise these literacies and how they can be integrated into education and training across diverse organisations.

Key Points for Practitioners

• Promote inclusive literacy programmes that strengthen citizens’ digital, data, and AI skills as interconnected social practices rather than isolated technical abilities.

• Integrate critical and communicative-cognitive dimensions into training, enabling individuals to interpret algorithmic results responsibly and engage transparently with digital systems.

• Design a framework for evaluating literacy levels across different population groups to tailor educational and civic initiatives to citizens’ actual needs and capacities.

• Foster collaboration between educators, policymakers, and technology developers to ensure that literacy strategies reflect societal values and support meaningful participation in data-based governance.

Introduction

In the twenty-first century, technology's rapid developments have significantly changed how people live, work, and learn. These developments have given rise to new literacies, skills, and competencies that are essential for individuals and organisations to function effectively in increasingly digital and datafied societies. However, the question remains as to what exactly these literacies, in their various forms, represent and how they can be studied both at the school and university levels and within professions where data and AI play an increasingly significant role.

Historically, ‘literacy’ referred primarily to reading and writing (Lankshear & Knobel, 2008). However, with the rise of the knowledge-based society and the shift from physical to digital information environments, the concept has broadened considerably. This evolution has introduced digital literacy – the ability to use digital tools and navigate online information effectively (Bawden, 2008). Followed by data literacy, which involves understanding, analysing, and interpreting data (Carlson et al., 2011). And more recently AI literacy, which encompasses knowledge of how artificial intelligence functions and how it affects decision-making, ethics, and everyday life (Long & Magerko, 2020). Together, these literacies reflect the growing need to understand, evaluate, and interact with technologies in both professional and societal contexts.

These literacies have increasingly been integrated into educational curricula, reaching a broader audience (Casal-Otero et al., 2023; Rayendra et al., 2021; Schield, 2004; Sefton-Green et al., 2009). The demand for these new literacies is further fuelled by rapid technological innovations, particularly with the introduction of large language models (LLMs) and large multimodal models (LMMs). These tools are now widely employed for tasks such as idea generation, data analysis, and decision support (Potter & Palmer, 2023; Roberts et al., 2024).

Despite the growing academic attention, digital, data, and AI literacies are still often studied in isolation, leading to fragmented definitions and limited understanding of their interconnections. While it is important to acknowledge the existence of other literacies, such as media, information, ICT (information and communication technologies), and statistical literacy (Calzada Prado & Marzal, 2013; Schield, 2004; Tiernan et al., 2023), we argue that digital, data, and AI literacy should not be analysed in isolation from one another or from these adjacent domains. Together, they create a new set of interrelated skills that address current societal, organisational, and professional challenges.

Their interrelation is particularly evident, as digital, data-based, and AI-integrated tools have become deeply embedded in people's everyday lives. Individuals are now expected to possess such skills not only in many professional and personal situations but also in interactions with public institutions – for instance, when applying for social benefits, communicating with authorities via chatbots, or interpreting municipal data dashboards. A wide range of previous studies has shown that in public administration, technological change increasingly affects not only citizens’ skills but also governance, decision-making, and communication both between institutions and with citizens (Busch et al., 2025; Ferro et al., 2010; Henninger, 2016). In this way, digital, data-based, and AI-related literacies have become an essential for every citizen. At the same time, these are rapidly evolving domains, leaving educational institutions and policymakers facing crucial questions about the dynamics of these changing literacies and how to design effective educational and digital policies.

To address this gap, this article conducts a narrative literature review to explore how digital, data, and AI literacies are conceptualised and interrelated. We identify the key components of these literacies, and explore their similarities and differences. Through the narrative method, this study maps and synthesises existing dimensions, thereby supporting the development of more integrated approaches to literacies for an AI-infused age.

The Changing Concept of ‘Literacy’

Literacy is a subject of research across many different fields, including education, anthropology, linguistics, and social sciences. Initially, literacy referred to the ‘basic or primary levels of reading and writing’, but over time, this definition has expanded to encompass knowledge in specific areas (Graff, 2022, p. 24, note 9). Many scholars view literacy as a ‘language of possibility’, where connections are drawn between literacy, culture, education, and society, establishing literacy as a means of expanding individual rights and opportunities (Freire & Macedo, 2005, p. 36). Therefore, beyond its classical meaning, literacy evolved in the twentieth century into a more complex construct, situated at the intersection of various disciplines and theoretical approaches. Later, the concept of literacy has become more flexible: Barton and Hamilton (2012) viewed it as a social practice shaped by local contexts, while May and Street (2017) emphasised it as a network of social interactions rather than an individual skill. These intellectual developments gave rise to new interdisciplinary frameworks for studying literacy, such as the ‘New Literacy Studies’ or ‘Literacy as Social Practice’ (May & Street, 2017, p. 4). Therefore, the understanding of literacy shifted from its classical definition.

Diverse concepts are employed to explain literacies in the context of digital transformations, encompassing both practical (‘skills’, ‘capabilities’) and theoretical (‘knowledge’, ‘awareness’) notions, each offering distinct perspectives. For instance, ‘internet skills’ are part of the broader ‘digital skills’ framework (van Deursen & van Dijk, 2014), focusing on specific abilities to navigate the online world. In contrast, ‘digital capabilities’ are often discussed in the context of optimising and developing business processes and are synonymous with ‘competencies’ that enable employees to fully leverage technology (Arkhipova & Bozzoli, 2017). Similarly, ‘digital competencies’ are understood as the ability to use information technologies appropriately (Levano-Francia et al., 2019; Pettersson, 2017; Rizza, 2023). Meanwhile, ‘digital knowledge’ encompasses both knowledge acquired through digital means and that processed with digital tools, closely related to ‘awareness’ (Belisle, 2006).

Choosing a precise term for the general – whether full or partial – capacity to engage with a technology matters. Graff (2022) critiques the proliferation of ‘multiple literacies’ as boundary-diluting and contends that digital technologies do not alter the core of literacy, so continual redefinition is unwarranted. By contrast, ‘knowledge’ denotes awareness or familiarity and ‘skills’ denote concrete techniques; in competence frameworks, they are often paired (e.g., ‘digital knowledge’ and ‘digital skills’) to capture theory and practice, yet they remain relatively static and task-specific. ‘Literacy’ is the broader construct: it integrates understanding with context-sensitive application, enabling both active use and informed interpretation. In this article, we therefore use ‘digital literacy’ to denote an integrated, adaptable capacity to engage with digital technologies while recognising the ongoing debate flagged by Graff. The concept is well established in the literature and has been used for decades (Bawden, 2008; Eshet-Alkalai, 2004; Gilster, 1997; Naamati-Schneider & Alt, 2024).

The modern discourse increasingly highlights the rapidly expanding world of data and the growing need to understand and work with it effectively. This dimension represents an essential component of digital literacy while also requiring a distinct educational focus. The term ‘data literacy’ has already been in active academic use for more than a decade (D’Ignazio, 2017; Mandinach & Gummer, 2013; Schield, 2004), making it a well-established and relevant concept for this context. As the data landscape evolves and most information becomes digitised, individuals require not only practical skills but also a theoretical understanding of data and its broader societal implications. To bridge the gap between technically trained specialists and other groups of citizens, it is therefore crucial to raise awareness and promote educational initiatives in the field of data literacy (D’Ignazio, 2017). So, ‘data literacy’ is also an appropriate term for the aim of our research.

With the increasing integration of AI into daily life, research has begun examining AI's impact on modern literacy concepts, often treating it as a subset of digital literacy. For instance, Tiernan explores AI's influence on digital, information, and media literacies (Tiernan et al., 2023). Other studies emphasise the need to teach AI literacy in both schools (Casal-Otero et al., 2023) and higher education (Naamati-Schneider & Alt, 2024), preparing students for a society where AI plays a significant role in their daily lives and interactions with state institutions. At the same time, the growing wave of AI use should be more fully reflected in public institutions, and recent studies point to the need for a structured framework that defines AI-related competencies for school learners and informs the development of curricula and didactic initiatives (Casal-Otero et al., 2023). Such frameworks can also serve as a reference for education authorities and other public actors responsible for planning and supporting these initiatives. Digital and AI literacies are, therefore, more crucial than ever (Hong & Kim, 2024), potentially reshaping literacy concepts in the future. Consequently, ‘literacy’ is a more fitting term for both theoretical understanding and practical application.

We therefore adopt the terms data literacy and AI literacy in this article. As the data landscape evolves and most information is digitised, people need not only practical skills but also a theoretical understanding of data and its social implications. AI is advancing just as quickly – from everyday interactions (e.g., voice assistants) to tasks such as interpreting and critically evaluating AI-generated content. The term literacy is apt: it signals an integrated, adaptable competence – understanding, application, and reflection. This aligns with the digital-literacy literature; extending that framework to data and AI follows Graff's (2022) view of expanding established concepts rather than coining new ones.

Narrative Review Method

In this article, we employed a narrative review method (Sukhera, 2022) to review and compare concepts related to digital, data, and AI literacy. The narrative approach is particularly suitable for exploring conceptual overlaps and theoretical complexity in relatively unexplored fields, allowing for more synthesis and contextualisation (Baumeister & Leary, 1997; Ferrari, 2015; Sukhera, 2022).

This analysis is based on a total sample of 25 articles. Article selection followed a multi-step textual analysis integrating sampling principles of narrative review (Ferrari, 2015) and purposive sampling (Suri, 2011). Instead of a meta-analytic or other quantitative review approach (Bourhis, 2017), articles were selected through a stepwise process grounded in substantive and conceptual criteria. We began by defining the search terms for article selection. Following an initial mapping of terms from the literature and exploratory searches, we adopted three core term sets: ‘digital literacy’, ‘data literacy’ and ‘AI literacy’. Our initial search was conducted in the Scopus database, as it offers a wide spectrum of peer-reviewed research articles in the social sciences, and it provides access to both theoretical and empirical studies.

The screening process started in 2024 and included several stages to identify the most relevant articles. For the literature review, we determined the publication data filter and we opted for articles which were: 1) published between 2010 and 2024; 2) published only in the English language; 3) both peer-reviewed journal articles and conference papers as the main sources; 4) published only in the social sciences subject area. This field was prioritised in our research and represents the largest body of publications relevant to literacy, thereby providing the most substantial sample for this topic.

Following a narrative review approach (Baumeister & Leary, 1997; Green et al., 2006), we did not maintain PRISMA-style flow counts at each stage. At the final screening stage, we examined titles and keywords for the three target literacies. For AI literacy, Scopus returned approximately 1,000 records; about 300 were title/keyword-screened, and 60 with ‘AI literacy’ as a distinct keyword were examined in depth. For data literacy, the initial 36,000–37,000 records were reduced to about 250 using a restrictive keyword filter – necessary given the volume; we did not apply this filter to AI at the outset, as doing so would have yielded only 65–70 items and undermined cross-topic comparability. For digital literacy, the initial 18,000–19,000 records were narrowed to about 1,500 under the initial criteria; further refinement using Scopus's AI-assisted search (e.g., ‘digital literacy AND education’, ‘digital literacy AND literature review’) yielded a manageable, methodologically comparable corpus of approximately 310 items for analysis.

Then, we screened each article's abstract for relevance to the literacy concepts. Finally, we assessed each article's full text to determine whether it addressed the theoretical and/or empirical basis of explaining, describing, or analysing a specific literacy type. For each literacy, we aimed to include at least five conceptually rich or seminal articles.

Following the process, 13 articles were retained from the Scopus database. To complement the sample and ensure inclusion of the key words not always retrievable via keyword searches, we employed backwards snowball sampling (Green et al., 2006; Wohlin et al., 2022). Reference lists of the included articles were examined for additional studies in which the main search term appeared in the title, abstract, or keywords. This method yielded 12 additional articles, producing a final dataset of 25 studies (13 being from Scopus and 12 from snowballing, see Table 1), representing a total of about 960 pages analysed.

Table 1.

Articles Used in the Analysis by Database and Literacy Type.

	Articles Published in the Scopus Database	Articles Found using the Snowball Method
Digital literacy	3	6
Data literacy	4	3
AI literacy	6	3
Total number of used articles:	13	12

The analysis proceeded in three main steps. First, we introduced and described theoretical perspectives and definitions for each type of literacy, summarising key findings and identifying essential practical and theoretical skills. Second, we identified commonalities and differences across the three literacies. Using the final sample of 25 articles, we compiled keywords representing core theoretical and empirical components for each literacy. These keywords were constructed through thematic synthesis, meaning they were developed by identifying and combining common themes from the literature review, rather than copied word-for-word from the texts (Braun & Clarke, 2006). These were organised into three categories, with 18 keywords per literacy (54 in total), reflecting field-specific skills. Finally, we compared these keywords to identify underlying latent dimensions, which were grouped into sub-categories. We analysed and compared the concepts in their usage contexts, distinguishing, for instance, between individual and institutional settings in which these literacies are applied. Comparing similarities and differences among these dimensions helped clarify how the literacies interrelate and evolve, contributing to prior research by synthesising knowledge on digital, data and AI literacy in the context of technological transformation.

Literature Review on Digital, Data and AI Literacies

Digital Literacy

Digital literacy, a term first introduced in 1997 by Gilster, is closely linked to twenty-first-century skills essential in our rapidly digitalising world. Gilster defined it as ‘the ability to understand, evaluate, and integrate information in multiple formats delivered by computers’ (Gilster, 1997, p. 1). Beyond evaluating and comparing information found on the internet, digital literacy also involves filtering and categorising information (Kreinsen & Schulz, 2023; Rayendra et al., 2021).

However, the scope of digital literacy extends beyond these informational skills. Martin (2006) refers to digital literacy, describing it as people's awareness, mindset and ability to use digital tools and technology appropriately to identify, access, evaluate and analyse digital resources, as well as create new knowledge and interact with other individuals (see Table 2). Hague and Payton (2011) further define digital literacy as the ability to use a wide range of practices and cultural resources through digital realms. This includes collaborating, communicating effectively, and understanding when and how digital tools and technologies can support these activities. In addition to electronic technologies, digital literacy also includes different online, informational and multimedia resources that are being used for communication and collaboration purposes.

Table 2.

The Main Keywords Used for Characterising Digital, Data and AI Literacy.^a

^a

Colours in the table are used to group keywords: blue for technical, red for critical, and yellow for communicative-cognitive.

In addition to being more than just technical skills and their conscious use, digital literacy is also seen as an evolutionary extension of other literacies. For example, Ala-Mutka (2011) defined digital literacy as an evolution from other literacies, such as media literacy, internet literacy and ICT literacy. Similarly, Ng (2012) refers to digital literacy as a variety of literacies related to the use of digital technologies. These technologies include both hardware and software that people use for educational or entertainment purposes. Digital literacy further includes cognitive, sociological, emotional and critical thinking skills that enable people to use technology effectively, understand digital content and make informed decisions in the digitalised world (Eshet-Alkalai, 2004; Tinmaz et al., 2022). Later, Law et al. (2018) similarly define digital literacy as a set of skills that includes computer literacy, ICT literacy, information literacy and media literacy. In addition, they argue that digital literacy gives individuals the ability to safely access, understand, integrate, and communicate information through digital technologies, thereby opening work opportunities.

In addition to treating digital literacies as a general phenomenon or a personal skill, other articles emphasise the institutional level – the emerging capabilities organisations need to navigate digital transformations. For example, in the context of public administration, digital literacy is essential for both individual officials and public institutions. As the majority of public services and communication with citizens increasingly take place online, as the analysed articles indicate, officials require skills not only to use and integrate digital platforms but also to analyse information effectively and transparently, communicate clearly, and make informed decisions based on digital inputs (Floridi & Cowls, 2019; OECD, 2021). Since not all citizens are familiar with or educated to use these digital services, it also becomes part of officials’ responsibilities to guide and educate citizens on how to access and navigate online platforms (Kaun & Männiste, 2025). From the perspective of digital governance, where traditional forms of public administration and governmental services are embedded with technological tools, digital literacy, together with data and AI literacy, serves as a key enabler of digital transformation (Hanisch et al., 2023).

Different countries adopt digital literacy development programs for various reasons – from reducing the digital divide to increasing the efficiency of public officials (Law et al., 2018). Nevertheless, the existence of such programs and their active support at the state level are crucial, as only this can produce a genuine rise in digital literacy within each society. Digital competence also benefits society more broadly: through internet-based communities of practice, workers and professionals now have an effective means of collaborating on tasks and expanding their knowledge with peers worldwide (Ala-Mutka, 2011).

Building on this perspective, an analysis of the literature shows that while the term digital literacy initially seems broad, it has acquired increasingly clear characteristics with the rapid development of technology. The term originated at a time when the potential for computers, including the creation of smartphones and social networks, was not yet evident. The evolution of this concept, from a simple understanding of information and interaction with it to the need for digital skills in communication and work, began around the era of social media, such as Facebook's creation and widespread use. Based on this, the concept of digital literacy will likely continue to evolve as the internet and technological developments foster a more digitally open society, where information and data are readily accessible. This shift will give rise to the development of what has been defined in the literature as ‘data literacy’.

Data Literacy

Society's growing reliance on data is evident – every service and decision that is being shaped and informed is data-based, showing the shift towards a datafied society (see, e.g., Masso et al., 2020b; Schäfer & van Es, 2017). Decisions based on data and evidence tend to be more successful and trustworthy, as they help to address problems and resolve them effectively (Schield, 2004; Schüller, 2022). The transformation to the digital realm has also made people more conscious of the data they consciously or unconsciously create, share and use, prompting the need to be better able to orient the datafied world (Deahl, 2014).

Whereas the definition and process of social datafication have gained a lot of scholarly attention and there is at least some agreement on the definition, the concept of data literacy is still unclear. At its core, the definition revolves around the ability to understand and/or use data effectively in everyday life (Carlson et al., 2011; Mandinach & Gummer, 2013; Masso et al., 2020a; Vahey et al., 2006). However, each author complements and enriches this understanding of data literacy with further practical insights. Vahey et al. (2006) and Carlson et al. (2011) emphasise that data literacy provides individuals with important technical skills and knowledge needed to shape positions, understand graphs and charts, draw conclusions from data, and identify instances of data misuse or manipulation. Mandinach and Gummer (2013, p. 30) define the practical parts of data literacy from the research standpoint of ‘how to develop hypotheses, identify problems, interpret the data, and determine, plan, implement, and monitor courses of action’. Schüller (2022) later complements this definition by emphasising the critical ability to solve problems and communicate their solutions (see Table 2).

Therefore, it becomes evident that data literacy encompasses more than just technical skills such as data interpretation, graphs and charts reading, but also critical skills like understanding how data is collected, how to verify its veracity, and how to learn to use it in general. All of this, as D’Ignazio (2017) points out, is an important part of modern data literacy, and people who are working with data (e.g., public sector officials, scientists, journalists) should be trained to interact with it correctly. But it is essential to select data that are relevant to the group being trained: for instance, municipal government officials require data drawn from citizen surveys (D’Ignazio, 2017). This also raises issues of inequality, as data is largely controlled by a small minority, leaving most citizens unable to collect, analyse, or use it effectively. It is equally important not only to train civil servants and address the issue institutionally but also to ensure that digital literacy is cultivated at the personal level by establishing organised systems of digital literacy education for all citizens (Ala-Mutka, 2011). This is especially relevant as approaches to data literacy are evolving: it is now required at the personal as well as the institutional and state levels, since citizens are no longer merely data objects but increasingly act as data subjects, learning to use data independently (Schüller, 2022) and needing support in this process. For example, institutions already encounter students with widely varying levels of data management skills, which underscores the need to expand training in this area (Carlson et al., 2011).

Therefore, previous studies note that government officials could play a far more active role in promoting and implementing data literacy, particularly among educators (Mandinach & Gummer, 2013). Achieving progress, however, requires a coordinated rather than piecemeal approach – one that brings together politicians and officials from federal and state education agencies with professional education bodies and practitioners (e.g., university deans). Without such cooperation, a significant improvement in teachers’ data literacy is unlikely.

In sum, there are markedly different approaches to defining what is included in data literacy. While some researchers focus on specific practical and technical interactions with data, including its storage and processing, others take a broader view, examining the role of data in social interaction and societal interpretation. However, these aspects are tightly interrelated: working with data builds personal understanding and societal knowledge, making data competencies a social process in which collective development depends on individual skill.

AI Literacy

Digital data, created through human-generated data traces, has provided growing opportunities for the development of data models, such as machine learning models, often referred to as AI (McCarthy, 2007). Although the concept of AI has been in use for decades, the concept of AI literacy is relatively new.

AI has started to evolve significantly, especially with the recent rapid technological development, such as machine learning, LLM and LMM, which have also influenced the increasing discussions about AI literacy (Sidra & Mason, 2024). Since the first use of the concept ‘AI literacy’, definitions have varied significantly, based on the articles analysed in this study. For example, Ali et al. (2019) and Steinbauer et al. (2021) refer to AI literacy as a skill that enables people to understand and identify the AI solutions, tools, methods, and algorithms in the diversity of AI applications. This may include the ability to understand how AI algorithms function, and how to use these algorithms in order to analyse and identify benefits, solve problems and evaluate the social and ethical impact and aspects of AI solutions on society. In addition to the engineers’ and developers’ perspectives, Long and Magerko (2020) define AI literacy as a set of skills that enable citizens to effectively interact with AI tools and technologies, such as voice assistants, smart devices, or robot assistants. A more recent definition highlights understanding the basics of AI solutions and concepts across various services (Burgsteiner et al., 2016; Casal-Otero et al., 2023; Kandlhofer et al., 2016). For example, Casal-Otero et al. (2023) propose integrating AI literacy into core subjects as a cross-curricular topic, shifting the focus to the practical, everyday use of AI across a wide range of applications.

While some researchers focus on methods to differentiate AI-created content from human-generated content (Kamali et al., 2024; Yi, 2021; Zhang et al., 2022), others concentrate solely on defining and conceptualising AI literacy (Casal-Otero et al., 2023; Long & Magerko, 2020, etc). It becomes increasingly difficult to distinguish between AI-generated and human-created content (Casal & Kessler, 2023). Consequently, teachers may struggle to assess students’ work and identify when AI has been used. It is crucial to differentiate between unauthorised AI usage, insufficient engagement (e.g., verbatim copying), and intentional plagiarism. Although AI-generated content and plagiarism share similarities, they are distinct issues requiring tailored assessment and responses. Clear guidelines and evaluation frameworks are essential to effectively address these challenges in educational settings.

Guides are already being developed to help users and researchers distinguish between ‘artificial’ and ‘real’ artefacts, such as images (Kamali et al., 2024) or texts (Casal & Kessler, 2023). As more concrete criteria emerge, this ability to differentiate between human- and AI-created content could become a key aspect of AI literacy (Yi, 2021; Zhang et al., 2022). However, this area remains in its early stages and lacks clear, universally accepted guidelines. Additionally, as AI becomes more integrated into daily life, it is increasingly difficult to define where the line is between products generated entirely by AI, created solely by humans, or developed with AI assistance (Elkhatat et al., 2023). If AI becomes widely used for tasks like writing, it may normalise reliance on AI, blurring the distinction between human-written and AI-assisted texts, and potentially reducing independent writing altogether.

These challenges are not limited to individual users; they also reshape organisational practice across sectors (e.g., public administration, healthcare, education). Public administration offers one illustrative case. AI-supported tools – chatbots, dashboards, decision-support systems – are already in use, requiring officials to develop a mix of technical, critical, and ethical competencies (Kaun & Männiste, 2025; Masso et al., 2022; Masso & Kasapoglu, 2020c). Officials need to understand how these systems operate, critically evaluate outputs and inputs from both staff and citizens, and remain alert to potential biases or errors. They may also act in multiple roles: as developers ensuring tools function, implementers integrating AI into workflows, and educators guiding citizens in using AI-enabled services.

Therefore, although private educational courses tend to respond more rapidly to technological change, it remains essential to embed AI literacy at the institutional level from the earliest school grades (Steinbauer et al., 2021). Like data literacy, AI literacy should be integrated across all areas of education and made an inherent component of both school and university curricula, with dedicated programs also developed for vocational and extracurricular training (Casal-Otero et al., 2023; Schüller, 2022).

Underlying Literacy Dimensions in the Digital, Data and AI era

Based on the analysis conducted above, we can propose three main overlapping and tightly intertwined dimensions of digital, data and AI literacies: technical dimension, critical thinking dimension, and communicative-cognitive dimension (see Table 3). We have named them based on the key terms and characteristics that emerged across different types of literacy in our narrative analysis, as well as additional articles used to provide a contextual interpretation of conceptual evolution. These names reflect the tightly interconnected and overlapping nature of the dimensions, which, while essential to digital, data, and AI literacies, can be understood and practised in different ways.

Table 3.

Comparison of Keyword Groups Across Literacy Types.^a

The first dimension emerged in our analysis, technical, characterises the practical, hands-on abilities needed to interact with technology, manage data, and use AI services or solutions. This dimension encompasses several commonalities with general digital skills, such as understanding and using various technological tools (e.g., smartphones, PCs, data analysis or programming tools), managing different types of content, information, or data, and grasping the basics of everyday technology, including both software and hardware. It is primarily concerned with instrumental skills, such as how to operate, set up, and troubleshoot technological tools and systems. However, there are significant differences across these three literacies regarding the technical dimension. In data literacy, this entails developing skills for working with information, such as managing, collecting, visualising, and making decisions based on it. Similarly to data literacy, AI literacy expands to understanding technical and functional concepts of AI solutions and how they work, for example, algorithms and statistics. In an institutional context, public administration officials need hands-on capabilities to operate and maintain AI-enabled service channels (e.g., chatbots and dashboards), integrate and visualise administrative data, and troubleshoot data flows between registries.

The second dimension identified in the analysis is critical thinking. Overall, this dimension emphasises the importance of critically evaluating content across online and digital sources, being aware of how information and data are created, shared, and manipulated, and reflecting on their validity. In digital literacy, this dimension includes evaluating, understanding and comparing information, or in other words, being informed of different biases in the online information, for example, being able to distinguish the correct information from the fake news. In data literacy, critical thinking involves solving problems with data, recognising potential bias, misuse, and manipulation, and assessing the consequences of unethical practices – for example, when public institutions deploy data-based systems to administer child-benefit allowances. It also includes assessing online information and questioning data sources, particularly in the context of decision-making, sales planning, and advertisements. For example, this could mean understanding how internet platforms use shared personal data to display targeted ads. And lastly, in AI literacy, the critical dimension considers estimating potential ethical, legal and societal impacts related to AI solutions, content, technologies and services. This may include issues with copyright while using AI for producing visual, written or audio content.

The third dimension is communicative-cognitive. This dimension involves the ability to communicate with different technological resources (e.g., devices, internet channels and social media groups) and to interpret and contextualise information, as a common characteristic in all analysed literacy types. In digital literacy, communication focuses on individuals using technologies like smartphones or PCs to engage actively in online environments. This requires understanding how to communicate effectively based on the platform and the user's specific goals, as well as cognitively positioning oneself to these technologies. For example, it involves adapting communication styles or language for different contexts, such as consuming social media content, building relationships, or participating as an active citizen in online community discussions and local internet channels. In data literacy, the communicative-cognitive dimension involves actively engaging with data to support arguments, making informed decisions using openly accessible data, and interpreting the content and context behind the numbers shared online. In AI literacy, the communicative-cognitive dimension emphasises collaboration with AI systems, such as those using LLMs (e.g., ChatGPT). This involves not only understanding how to use these platforms effectively but also interpreting and integrating AI-generated content into everyday personal or professional processes. Or in public institutions practice, this means translating model outputs into plain language for citizens, crafting safe, effective prompts for LLM-assisted drafting without inserting sensitive data, and documenting interactions to support transparency and accountability.

An important aspect of the communicative-cognitive dimension is awareness of the social and psychological dynamics of human-AI interaction, which helps individuals maintain healthy boundaries when using generative AI tools (Wang et al., 2024). Individuals with high communicative-cognitive abilities are less likely to share personal information or be influenced by the emotional appeal of short-form video platforms and the need for constant validation. This dimension supports users in critically positioning themselves and resisting the anthropomorphisation of AI. Thus, communicative-cognitive AI literacy involves not only functional skills but also digital self-awareness and resilience.

Based on the analysis above, we can conclude that all three dimensions identified are essential for achieving digital, data, and AI literacy. While each dimension may be applied and practised differently across these literacies, they are interconnected and should not be viewed in isolation. While the communicative-cognitive dimension in data literacy requires communication in an online environment, in AI literacy, this dimension requires communication with the technology, being in an online or offline environment. Technical dimension, on the other hand, can be broken down into varying degrees of proficiency, ranging from basic to advanced. In data and AI literacy, it demands somewhat more field-related and specific expertise, while in digital literacy, it requires rather basic or even entry-level skills.

Discussion

The analysed literature suggests that the term ‘AI literacy’ will soon become more defined, as its scope narrows within the broader category of ‘digital’. Rooted in general digital and data literacy, this emerging concept is closely connected not only to these three forms of literacy but also to the three dimensions identified in the analysis.

Based on the conducted research, we suggest that the foundational role of digital literacy is a prerequisite for evolving AI literacy. This digital literacy consists of basic skills and knowledge required for everyday functioning within the technological advancements of the twenty-first century. It includes technical abilities like navigating online resources, basic critical skills like the ability to evaluate information, social and emotional skills, such as online communication and collaboration or digital awareness. For instance, Gilster (1997), Eshet-Alkalai (2004), Hague and Payton (2011), and Tinmaz et al. (2022) argue that digital literacy is the foundational skill for understanding technology, digital content, and digital tools, serving as a prerequisite for all other forms of literacy in the digital era.

However, AI literacy has introduced many essential additional competencies needed, besides general digital literacies, for engaging with, understanding, and critically assessing AI tools and technologies. This includes not only technical aspects but also ethical, cultural, and social dimensions that are unique to AI solutions (see Table 3). Unlike the broader concept of digital literacy, AI literacy, much like data literacy, on which it tightly relies, involves somewhat more field-specific and technical skills. For example, data literacy is a prerequisite for developing AI literacies as the development of machine learning models inherently relies on the data. Therefore, AI literacy also increasingly means turning attention to technical skills related to collecting, analysing, and interpreting data.

As our analysis indicates (Hong & Kim, 2024; Kreinsen & Schulz, 2023; Schüller, 2022), the evolution of literacies from digital to data and AI occurs in multiple forms and across levels. Beyond individual learning pathways, through formal training and self-directed study, public awareness-raising and organisational capability building are critical (Floridi & Cowls, 2019; OECD, 2021). Public communication about risks associated with data and AI is often reactive rather than proactive, meaning that authorities tend to respond after incidents occur rather than anticipating risks and informing citizens beforehand, as illustrated by several data scandals in Estonia (Delfi, 2023; Kiisler, 2025; Republic of Estonia Ministry of Justice and Digital Affairs, 2025).

Our analysis also shows that digital and AI literacies are often treated as abstract umbrella terms, with little effort to situate them in specific contexts – this contrasts with prior work on data, digital and AI transformations, which foregrounds context (Kitchin & Dodge, 2014; Loukissas, 2019; Masso et al., 2022; Engvall and Flak, 2022). Only a few studies (Law et al., 2018; Wirtz and Müller, 2018) have emphasised the importance of institutional and organisational contexts, for example, public administration, rather than referring generically to formal or informal education systems as responsible for developing digital, data, and AI literacies. However, the literature seldom addresses either the changing nature of these literacies or the need to integrate technical, critical and communicative–cognitive dimensions – instead, digital, data and AI literacies are typically considered in isolation rather than as interdependent (Busch et al., 2025; Ferro et al., 2010; Henninger, 2016).

Therefore, a key question is how the three dimensions identified in our narrative review are embedded in, and can be integrated across, everyday situations and institutional contexts. As digital, data, and AI literacies have become essential for citizens to navigate daily life and interact with public institutions, public administration increasingly operates as a mediator that shapes how these literacies are developed and applied in society. Three interrelated roles can be identified in this process. First, civil servants themselves must be trained in these literacies to ensure that they can design, implement, and communicate data-based policies in transparent and inclusive ways. Second, public institutions function as facilitators that promote these literacies among citizens – for instance, by supporting digital training programmes, and the creation of accessible, user-oriented interfaces that enable citizens to engage confidently with data-based services. Third, public officials act as regulators and advocates for the ethical and responsible use of technology, helping to build public confidence in both digital systems and the institutions governing them. Existing studies have examined how AI tools can improve the performance of civil servants (Casal-Otero et al., 2023; Kaun & Männiste, 2025) and how technological adoption influences citizens’ trust in public institutions (Kaun & Masso, 2025; Kaun et al., 2023; Masso & Kasapoglu, 2020c). Future research could further explore how these literacies can be institutionalised as part of a broader model of datafied citizenship – through continuous training for public officials, integration into university curricula, and inclusive civic initiatives that strengthen citizens’ participation, and digital inclusion.

Public institutions, such as national statistical offices, also have a central role in proactively developing citizens’ skills, particularly technical literacies, for example, guiding the use of smart-city dashboards to find relevant information and supporting interactions with government services via chatbots (Kaun & Männiste, 2025). Kaun and Männiste (2025) also note that officials must navigate ‘AI frictions’ between high expectations and limited functionality, which necessitates a reflective stance toward the tool's actual capabilities and societal implications. For instance, when teaching how to use tools like ChatGPT, it is not sufficient to focus solely on how to formulate requests correctly. This should include: 1) a technical dimension – knowing how generative AI systems work, and how to create and customise conversations; 2) a critical thinking dimension – learning to approach the tool with both trust and scepticism and taking responsibility for verifying its outputs; 3) a communicative-cognitive dimension – being able to interpret and communicate the outputs correctly, recognising that the system might make mistakes or reflect biases.

Furthermore, it entails recognising that the use of AI is bound up with how we understand ourselves: prior skills, perceptions and institutional memory (such as organisational culture) shape how we engage with and make sense of technology. The relationship is bidirectional: deploying AI can, in turn, reshape how we define ourselves within the organisation – the skills we foreground, our everyday practices and tasks, and even the nature of work.

Therefore, future research could further explore how organisations such as public administration can institutionalise the development of digital, data, and AI literacies both through continuous professional training for current public officials and through integrating these competencies into university programs that prepare future public servants. In addition, it would be valuable to examine how such initiatives influence citizens’ trust, participation, and digital inclusion.

Conclusion

This article aimed to contribute to ongoing discussions on the essential literacies required to navigate rapid technological transformations, particularly those driven by digital technologies, increasing datafication, and the emergence and integration of AI tools into everyday life. Based on a narrative literature review of digital, data, and AI literacy, we examined their distinct components, synthesised these perspectives, and proposed an integrated framework to understand their interrelationships.

Through the analysis, we identified digital literacy as a central concept or umbrella term encompassing both data and AI literacy. While digital literacy equips individuals with foundational skills and knowledge needed in the twenty-first century, such as understanding and comparing digital content and using basic digital tools, data literacy and AI literacy require more field-specific or specialised knowledge. For instance, data literacy provides individuals with the skills to work with and understand data, such as using it for decision-making and problem-solving, visualising data, and effectively communicating the insights it conveys. AI literacy goes further, enabling interaction with AI tools and services, understanding their functionality, and assessing their potential social, ethical, or legal implications.

Between 2010 and 2024, these literacies evolved from being defined by mostly technical to more complex, multidimensional literacies. Early work emphasised technical proficiency, whereas more recent papers focused on the importance of critical awareness of biases and consequences, as well as the ability to interpret and communicate technological outcomes. This shift emphasises the importance of integrating three interrelated dimensions across all literacies: the technical (practical skills for using digital tools), the critical thinking (ability to assess implications and risks), and the communicative-cognitive (capacity to interpret and contextualise insights). While the first two dimensions are frequently discussed in the literature (Carlson et al., 2011; Kreinsen & Schulz, 2023; Ng, 2012; Rayendra et al., 2021), the communicative-cognitive dimension has received less scholarly attention (Casal-Otero et al., 2023; Long & Magerko, 2020). It should also be noted that for all literacy types, the number of publications exhibited a marked increase beginning in 2024. However, as our analysis was conducted at the end of the first quarter of 2024, publications appearing later in the year were not included in this study.

Our analysis also revealed that digital, data, and AI literacies need to be studied more concretely in the institutional context. For example, public institutions’ officials make decisions that directly affect decision-making, services, and citizen trust, and thus must be able to use emerging technologies responsibly. This highlights the need for a deeper understanding of how systematic professional training programmes for current public officials should be designed to enable them to effectively navigate and manage emerging technologies in their daily work and policy decisions. To ensure these programs address real needs, a diagnostic tool should be developed to assess officials’ levels of digital, data, and AI literacy, with training subsequently tailored to the identified gaps. At the same time, integrating these three types of literacy into university curricula for future public servants is essential to build long-term institutional capacity. By strengthening the technological competencies of both current and future officials, governments can facilitate citizens’ adaptation to new technologies and foster stronger tconnections between society and public institutions.