Data literacy in the new EU DigComp 2.2 framework how DigComp defines competences on artificial intelligence, internet of things and data

Abstract

Purpose

The purpose of this paper is to analyse data literacy in the new Digital Competence Framework for Citizens (DigComp 2.2). Mid-2022 the Joint Research Centre of the European Commission published a new version of the DigComp (EC, 2022). This new version focusses more on the datafication of society and emerging technologies, such as artificial intelligence. This paper analyses how DigComp 2.2 defines data literacy and how the framework looks at this from a societal lens.

Design/methodology/approach

This study critically examines DigComp 2.2, using the data literacy competence model developed by the Knowledge Centre for Digital and Media Literacy Flanders-Belgium. The examples of knowledge, skills and attitudes focussing on data literacy (n = 84) are coded and mapped onto the data literacy competence model, which differentiates between using data and understanding data.

Findings

Data literacy is well-covered in the framework, but there is a stronger emphasis on understanding data rather than using data, for example, collecting data is only coded once. Thematically, DigComp 2.2 primarily focusses on security and privacy (31 codes), with less attention given to the societal impact of data, such as environmental impact or data fairness.

Originality/value

Given the datafication of society, data literacy has become increasingly important. DigComp is widely used across different disciplines and now integrates data literacy as a required competence for citizens. It is, thus, relevant to analyse its views on data literacy and emerging technologies, as it will have a strong impact on education in Europe.

Citation

Van Audenhove, L., Vermeire, L., Van den Broeck, W. and Demeulenaere, A. (2024), "Data literacy in the new EU DigComp 2.2 framework how DigComp defines competences on artificial intelligence, internet of things and data", Information and Learning Sciences, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/ILS-06-2023-0072

Introduction

The past 40 years have been characterised by strong technological innovations – some radical and some incremental – mainly based on the digitalisation of almost all sectors in society. The emergence of the internet, the integration with mobile technology, the evolution towards Web 2.0, the rise of social media and the advent of artificial intelligence or AI are all processes driven by digitalisation and innovation. Fundamentally linked to this process of digitalisation is the process of datafication.

Many authors agree that datafication is a new phase of innovation that offers huge potentials in terms of smart cities, smart environments, personalised services, user friendly recommendation systems, etc. For each process that is digitalised, new data is created both intentionally and non-intentionally (Zuboff 2015, 2019). This data can be captured, stored, analysed and used for different purposes. Advances in algorithms and artificial intelligence make that more and more meaningful insights can be extracted from (big) sets of data.

However, others point to the dangers of algorithm bias, loss of serendipity, privacy concerns, social sorting of customers and citizens, etc. At a much more fundamental level, Zuboff (2015) argues that datafication is not a technological process or an inevitable technical effect. For her, datafication plays a key role in the establishment of a new form of information capitalism that “[…] aims to predict and modify human behavior as a means to produce revenue and market control” (Zuboff, 2015, p. 75). Couldry and Mejias make a similar analysis claiming data is about “[…] the capture and control of human life itself through appropriating the data that can be extracted from it for profit” (2018, p.xi).

Together with the changes in media and technology came the call for new forms of literacy (Pangrazio and Sefton-Green, 2021). The general feeling is that especially young people need a new set of skills related to data, to be able to function in our current and future data-driven society. As Carmi et al. (Carmi et al., 2020) argue:

(This) lack of data literacy opens citizens up to risks and harms–personal, social, physical and financial–but also limits their ability to be proactive citizens in an increasingly datafied society (p. 2).

Yet, what these precise competences are or should be is currently hotly debated. This is logical as the debate on what data literacy is has not yet been stabilised.

In recent years, several new forms of literacy have been suggested in relation to the datafication of society (Van Audenhove et al., 2020). Different authors see a need for data literacy (Bhargava et al., 2015), algorithm literacy (Oldridge, 2017), coding literacy (Vee, 2017) or data infrastructure literacies (Gray et al., 2018). We prefer the term data literacy because data is the defining aspect in the process of datafication.

This discussion on new forms of literacies is not new. In the 1970s, media literacy emerged as a result of the new medium of television (Masterman, 1985). The focus of media literacy lies on understanding media and media processes (production, language, representation and audience) (Buckingham, 2003). After 20 years, from the 1990s onwards, as the result of computers, internet and digitalisation, the focus shifts to digital literacies. Early conceptualisations of digital literacy often refer to access and operational proficiency (van Dijk, 2020). More recent digital literacy competence models integrate analytical and critical skills. They list a more extensive list of specific skills and competences (Iordache et al., 2017). Competence models over time tend to become more complex, more encompassing and nuanced and more stable. The development of competence models is almost directly linked to a theoretical discussion about technological change and the social consequences of that change.

One of the more recent attempts to integrate data literacy in an overall competence framework is the new version of the EU DigComp Framework. The DigComp Framework is one of the most widely used digital competence frameworks in Europe and beyond (Barboutidis and Stiakakis, 2023; Bartolomé and Garaizar, 2022; Mattar et al., 2022). The new 2.2 version aims to “[…] engage citizens confidently and safely with digital technologies, taking account of emerging technologies, such as Artificial Intelligence (AI), the Internet of Things (IoT), (and) datafication” (Thornton, 2022). The competence framework as such has not changed. However, the framework is supported by new examples of knowledge, skills and attitudes. According to the DigComp 2.2:

These examples illustrate new focus areas with the aim to help citizens engage confidently, critically and safely with everyday digital technologies, but also with new and emerging technologies such as systems driven by artificial intelligence (AI) (Vuorikari et al., 2022, p. 2).

The questions of this article are:

Q1.

How does DigComp 2.2 define data literacy?

Q2.

What are the underlying assumptions?

Q3.

How does it define the competences needed?

These questions are important to focus on, as DigComp is widely used across different disciplines and now integrates data literacy as a required competence area for citizens. It is, thus, relevant to analyse its views on data literacy and emerging technologies, as it will have a strong impact on education in Europe. To answer the questions, we will subsequently:

• review the current theoretical discussion on data literacy;

• present the high-level data literacy competence model (DLCM) as defined by the Knowledge Centre for Digital and Media Literacy Flanders-Belgium;

• briefly introduce the EU DigComp 2.2 Framework;

• outline the methodology used to analyse DigComp 2.2;

• present our analysis of DigComp 2.2 itself in relation to the DLCM and present a thematic analysis; and

• present the conclusions of our work.

The discussion on data literacy

As indicated, the discussion on data literacy is in full flux. Different disciplines and groups of scholars hold rather different views on how data literacy should be defined and, therefore, what data literacy competences people should have. We identify three main fields of discussion: the open data field, the STEM field and the social science field.

The first field is the open data field, which focusses on providing access to as much data as possible and to make data freely available to society (Baack, 2015; Couldry and Powell, 2014). The movement started from the premise that open data can contribute to a participatory democracy in which citizens use data themselves, aimed at influencing politics and policymaking processes. This, however, assumes a high level of knowledge about data, processing, structuring, analysing and presenting data. Over the years, this community started to realise that not every citizen has the necessary knowledge to deal with data in this way (Carlson and Stowell Bracke, 2015; Frank and Walker, 2016). As Boychuk et al. (2016) note:

Data literacy is presented as imperative for understanding complex datasets, but many open government data initiatives do not address this skill, nor do they clearly lay out plans to teach this skill to the public (p. 4).

Hence, there is a growing interest within this movement for data literacy. Within this type of literature, the emphasis is mainly on being able to use data and less on understanding the role of data in society (Gray et al., 2018; Raffaghelli and Stewart, 2020; Van Audenhove et al., 2020). This approach to data literacy is often utilitarian, as most citizens do not possess the required competences to be able to properly make use of the data, because of a lack of competences focussing on understanding the data. This is something Gray et al. (2018) also state in their paper, calling for initiatives that extend beyond data science to encompass data sociology, data politics and broader public engagement with digital data infrastructures.

The second field is the field of STEM education in its broadest form, both at the level of primary, secondary and tertiary education. The emphasis here is on educational innovation through the integration of (big) data, algorithms and AI, into existing curricula and subjects that are aligned with these topics (Bhargava et al., 2015; Csernoch and Biró, 2015; MacMillan, 2015). The overall objective is, on the one hand, to teach students how to use data sets, algorithms and AI in a playful way (secondary and higher education). On the other hand, students are taught to critically reflect on data and the role of data in society. In general, however, much of this literature focusses heavily on being able to use data and less on understanding data. For tertiary education, for example, Carlson and Stowell Bracke (2015) observe that:

With researchers facing new requirements and expectations for managing, sharing and curating their data, it is critical that they have the knowledge and skills needed to respond effectively. However, competencies in working with data are often not included as a part of a student’s formal education. Students that do acquire proficiencies with data generally gain their skills in an ad hoc manner on the job and at the point of need (our emphasis) (p. 96).

Related to this field is also the discussion on data literate educators and schools that focusses on using data more generally throughout the curriculum (Mandinach and Jimerson, 2016; Ndukwe and Daniel, 2020; Reeves and Honig, 2015).

The third field is the field of social sciences reflection on data. This field is consistent with the reflection on data, people and society that takes place in the social sciences. Some of this literature looks at personal data literacy (Pangrazio and Selwyn, 2019) and the role of personal data and its consequences for the individual user. Other authors discuss the role of data in broader social processes (Zuboff 2015, 2019; Gray et al., 2018). This field is very much in line with the discussion on media literacy and the role of media in society (Van Audenhove et al., 2020). To come to grips with a broader understanding of data literacy, Carmi et al. (2020) introduce the concept of data citizenship as “[…] a framework that outlines the importance of citizens having a critical and active agency, at a time when society’s datafication and algorithmically-driven decision-making has become normalised” (p. 10). The framework consists of three areas:

1. data thinking – citizens’ critical understanding of data (e.g. understanding data collection and data economy);

2. data doing – citizens’ everyday engagements with data (e.g. deleting data and using data in an ethical way); and

3. data participation – citizens’ proactive engagement with data and their networks of literacy (e.g. taking proactive steps to protect individual and collective privacy and well-being in the data society as well as helping others with their data literacy) (Carmi et al., 2020).

We observe that within social and educational sciences, the prevailing focus in the literature is on the proficiency in using data, especially in relation to statistics, mathematics, data science and big or open data. We see a similar conclusion in the systematic review of Godaert et al. (2022), who note that digital competence assessments of children focus much less on the DigComp 2.1 areas of “problem solving” and “safety”. They also see emphasis on skills and use, less on knowledge and attitudes. Gebre (2022), who reviewed the literature on K-12 data literacy education and how the included articles define data literacy, finds that data literacy education research focusses on data skill development in relation to data use and looks less at data literacy from a perspective of “human agency” and how to live in a datafied society.

Data is always strongly influenced by context, and it is collected, handled and interpreted by people with their own values, norms and prejudices, which can perpetuate inequality and exclusion. So even in “using” data, citizens need to be able to critically “understand” and look at the broader narrative of data in society and the contexts to which the data belongs, from collecting to interpreting to presenting the data. “Although reasoning with data is rooted in epistemic processes, the stances people take are—and should be—influenced by values [and] concerns” (Polman et al., 2022, p. 1). As Lee et al. (2022) state, data literacy or data science education, need to put more emphasis on data justice, inclusion and plurality. Data literacy is moving toward this more critical focus, however, as was found in other articles, educational activities primarily focus on data use without necessarily critically reflecting on societal repercussions, bias in data, data fairness […] (Godaert et al., 2022; Gebre, 2022). The DLCM, which will be discussed below, seeks to tackle these issues by offering a comprehensive outline of essential competences required for a well-rounded data literate citizen and enabling educators to map their initiatives on a user-friendly model.

The data literacy competence model

In this article, we start from the DLCM, developed by the Knowledge Centre for Digital and Media Literacy Flanders-Belgium (Figure 1) (Seymoens et al., 2020). This model is largely based on the model of media literacy by the same centre. Based on existing literature, a model was presented and validated by a group of social science scholars with a background in media and data literacy. The aim of the model is to provide a high-level common understanding of data literacy in Flanders-Belgium. The model can be applied to evaluate specific data literacy initiatives in terms of competences they address (Seymoens et al., 2020). It can also be used by educators to develop and map specific data literacy initiatives that focus on the different competences. In Flanders-Belgium, the model is used to map and develop data literacy initiatives developed by the Knowledge Centre. It is also used to introduce and explain the concept of data literacy to policymakers.

The model comprises two major competence clusters using data and understanding data (Mediawijs, 2020b). Competences refer to the knowledge, skills and attitudes that allow individuals to act adequately in a given situation (Mediawijs, 2020a). For each of those two clusters, four broad competences are identified for data literacy. The competence clusters are defined in more detail as:

Using data or the knowledge, skills and attitudes to use data actively and creatively:

1. interpreting: being able to read a graph, table or list of data and understand what they mean;

2. navigating: finding your way through a collection of different types of data and ways they were processed and being able to extract the message or what you need;

3. collecting: being able to set up a process to collect raw data and organise an analysis; and

4. presenting: being able to present and visualise the results of a data analysis in a targeted manner, tailored to an audience.

Understanding data or the knowledge, skills and attitudes to critically and consciously assess the role of data:

1. observing: being able to observe how data is communicated and used;

2. analysing: being able to analyse the individual and social consequences of the way how data is communicated and used;

3. evaluating: being able to evaluate whether those consequences are harmful or constructive; and

4. reflecting: being able to reflect on how you and others communicate and use data, being able to adjust to minimise harmful consequences.

The cluster of competences for using data is more practice-oriented. This is in line with the focus of the open data movement field and partly also the education field on being able to actually use data. These fields see data literacy often as functional or goal-oriented. Starting with a problem that can be or needs to be solved by using data. In its definition of data literacy, the Education Development Centre lists data literacy competences as: “(He/she/they) can identify, collect, evaluate, analyse, interpret, present and protect data” (EDC, 2016). The order in which the competences are listed reveals the goal-oriented way of thinking. Collecting comes before interpreting. Yet, being able to collect data requires a lot more knowledge and experience than just interpreting data.

The model of the Knowledge Centre (2020), therefore, follows the levels of literacy in relation to data. It starts with the questions: Can I read data? Can I navigate different types of data? Can I organise data to analytically understand them? Can I collect existing and new data? Can I present and communicate those data? The cluster of competences for understanding data is more oriented towards critically and consciously understanding the role of data in society, personal life, etc. This focus is often neglected or underdeveloped, especially by those authors who have a background in statistics, data analytics or computer sciences. These competences are closely linked to reflections on the role of data in social science. The DLCM (2020) highlights the importance of bringing both competence clusters into one overarching model, as citizens need to possess competences in both using and understanding data to effectively participate and navigate our data-driven society. However, we see a much broader focus within society on being able to use data and its relation to numeracy and statistics, than on understanding data and critically looking at these data (Gebre, 2022; Godaert et al., 2022).

The EU DigComp 2.2 framework

On March 22, 2022, the Joint Research Centre of the European Commission published a new version of the Digital Competence Framework (EC, 2022) (Figure 2). The DigComp Framework is probably one of the most widely used digital competence frameworks. The tool can be used by member states and organisations within member states to develop digital competency policies and/or to assess the digital competences of its citizens, as mentioned in the document. It strives to create a common baseline for digital competences. The current model is based on 21 digital competences, which are divided in separate categories, or components. Within this framework, the model identifies five main components, namely, information and data literacy, communication and collaboration, digital content creation, safety and problem-solving (Figure 2).

The first European digital competence framework was published in 2013 by the Joint Research Centre in collaboration with relevant stakeholders, such as policymakers, educators and researchers. The framework has been updated and extended several times to come to grips with the ever-changing digital society (Carretero Gomez et al., 2017; Vuorikari et al., 2016). In 2021, the European Commission established several working groups to update the framework yet again. The new version 2.2 aims to “[…] engage citizens confidently and safely with digital technologies, taking account of emerging technologies, such as Artificial Intelligence (AI), the Internet of Things (IoT), (and) datafication” (EC, 2022). The competence framework as such has not changed. However, the framework is supported by new examples of knowledge, skills and attitudes in the three identified domains. According to the DigComp 2.2:

These examples illustrate new focus areas with the aim to help citizens engage confidently, critically and safely with everyday digital technologies, but also with new and emerging technologies such as systems driven by artificial intelligence (AI) (Vuorikari et al., 2022, p. 2).

Europe has widely recognised and embraced DigComp as the main model for digital competences. In Flanders, for example, Digisprong, the government's knowledge centre on digitalisation in education, used Digcomp 2.1. to develop their teacher guides, an opensource learning platform explaining teachers how to self-assess their digital competences. For students, the framework was used to help ground the educational target goals around “digital competences”. DigComp is also used within the educational context of other countries. In The Netherlands, it serves as the basis of a competence measurement tool for educational innovation, mapping how digital literacy is covered in higher education curricula (Versnellingsplan Onderwijs met ICT, 2022), and in France, the ministry of education developed a reference framework for digital skills from primary school to university, inspired by DigComp (éduscol, 2023). DigComp is also used by researchers to support and develop their digital competence assessments of citizens or to analyse educational practices. Because DigComp is widely used as an assessment tool in research and in education throughout Europe, it is important to analyse the framework to understand its potential and limitations.

Research set up, methodology and coding

What is new in the DigComp 2.2 Framework are the examples that reflect new emerging technologies and datafication in the domain of AI and IoT. In our contribution, we aim to critically analyse the DigComp 2.2 Framework. We do this by focusing on an analysis of the examples, as they are meant to: “[…] motivate education and training providers to update their curriculum and course material to face today’s challenges” (Vuorikari et al., 2022, p. 2).

The DigComp 2.2 Framework lists 259 concrete examples. As a first step of our approach, we carefully read all examples and selected all those that have a bearing on data, algorithms, IoT and AI. As mentioned, AI and IoT are the two new domains the DigComp framework explicitly mentions, so we included these together with data and algorithms, as algorithms are linked to decision-making (which is one of the thematic codes). Algorithms can also include AI, making it relevant to focus on. In principle, if the word data appears in an example, even as an explanation or sub-example between brackets, then we have included the example. We have left out examples that refer to digital skills and technology more generally. This led to a sample of 84 examples for our further analysis.

Next, the 84 examples on data were coded according to the following process:

• The first author of this article coded the 84 examples according to the eight competences identified in the DLCM: interpreting, navigating, collecting, presenting and observing, analysing, evaluating and reflecting. Each example was given only one code for one of the eight competences. If an example referred to two possible competences, then a decision was taken to code the most prominent one. The first author introduced a separate competence of processing between collecting and presenting, as this seemed to be missing (see later).

• The author also coded the 84 examples thematically based on the content of the example. For the latter, a coding scheme was progressively developed while coding. In line with the logic of DigComp 2.2, a broad distinction was introduced between data and artificial intelligence with more specific sub-themes identified and added while coding.

• Two of the co-authors independently coded all 84 examples according to the DLCM (with the extra competence of processing) and using the thematic coding scheme developed by the first author. A field for comments was added to note any potential questions or issues during the coding. This led to three separate fully coded files.

• In a joint session, the three coding authors then compared their coding, discussed the choices made, discussed divergences in the coding, agreed on interpretations of coding, clustered the thematic coding into a more comprehensive system and finalised the coding into a joint table for analysis (Appendix). We chose this approach to allow for an active discussion and final agreement on the coding between the three coders.

In what follows we describe the discussion by the coding authors, the reflections made and the decisions taken in relation to coding. For most entries (around 80% of the cases), there was convergence on coding both in terms of the DLCM and thematic coding. Entries starting with words like Aware or Knows clearly refer to competences in the cluster understanding data. Other entries starting with words such as Knows how, Can, Avoids and Are able to, most often refer to using data. For some entries, however, there was divergence on the interpretation between using data or understanding data and its sub-competences, leading to different codes by the three coders. The authors then decided to categorise entries that presume an active behaviour by the user under the cluster using data, even when elements of understanding data were also present. Again, only one code was attributed per example.

In the cluster using data, a discussion unfolded around the competence of collecting data. In the DLCM, collecting data is described as: being able to set up a process to collect raw data and organise an analysis. Yet, in DigComp 2.2, only one example explicitly refers to collecting data in the strict sense of the word. Example 36 states: Knows how to collect digital data using basic tools such as online forms and present them in an accessible way (e.g. using headers in tables). In our view, nine examples in the cluster collecting data would better fit under a new heading of processing data. Different examples in the DigComp 2.2 Framework, such as Competence 4.2: Protecting Personal Data and Privacy would fit in such a new category. Example 175 states: Able to encrypt sensitive data stored on a personal device or in a cloud storage service. This is clearly a competence that requires the user to act but has less to do with collecting data as such. The DigComp 2.2 Competence 2.6: Managing Digital Identity raises the same issue. Example 111 states: Able to verify and modify what type of metadata (e.g. location, time) is included in pictures being shared to protect privacy. In general, the competences related to identity and privacy are not easy to code in the competences of the DLCM. We, therefore, suggest introducing a new category processing data which could encompass not only processing of raw data but also protecting personal data.

In the cluster understanding data, coding seemed to be more straightforward. Discussion sometimes arose between evaluating and reflecting. Evaluation in the DLCM refers to being able to evaluate whether consequences are harmful or constructive. Example 42 related to managing data states: Watchful of accuracy when evaluating sophisticated representations of data (e.g. tables or visualisations as they could be used to mislead one’s judgement by trying to give a false sense of objectivity). Reflecting in the DLCM refers to the capacity of the individual to ethically assess the social implications of data. Example 165 related to programming is a good illustration: Considers ethics (including but not limited to human agency and oversight, transparency, non-discrimination, accessibility and biases and fairness) as one of the core pillars when developing or deploying AI systems (AI). But the line between the two is not always that clear and often needs interpretation. Example 54 in relation to the interaction through technology states: Open to AI systems supporting humans to make informed decisions in accordance with their goals (e.g. users actively deciding whether to act upon a recommendation or not) (AI). We considered that this goes beyond just evaluating the consequences of using technology and coded this under reflecting.

For the thematic coding, we first introduced the distinction between data and AI. This is a distinction made within the DigComp 2.2 Framework itself. All examples related to artificial intelligence are followed by a symbol (AI) in the document. At first, we coded examples with very specific thematic codes. However, often this resulted in one or two codes per theme. We, therefore, decided to cluster examples in larger themes. Identity, Privacy and Ethics for instance is a larger theme that well describes the importance DigComp 2.2 allocates to this theme in the framework. We kept some of the smaller themes such as Citizenship and Environmental Impact because it is important to indicate that DigComp 2.2 does reflect on data and AI in a broader societal context. As already mentioned, only example 36 refers to actually collecting data which would get much more attention in a STEM, educational or scientific approach to data literacy. Example 36 states: Knows how to collect digital data using basic tools such as online forms and present them in an accessible way (e.g. using headers in tables).

Critical analysis of the DigComp 2.2 framework

From the 84 examples referring to data literacy in the DigComp 2.2. Framework, 31 can be coded under using data and 53 under understanding data. A first observation to make is that DigComp 2.2 clearly emphasises understanding data more than being able to use or handle data. In this respect, DigComp 2.2 approaches data literacy from the more sociological field and less from an open data field or STEM field, as will be confirmed by our further analysis. In the first subsection, we will analyse DigComp 2.2 against the DLCM. In a second section, we will provide a thematic analysis of the examples related to data and AI. From the 84 examples, 49 refer to data and 35 specifically to AI. The focus on AI does not come as a surprise as DigComp 2.2 explicitly states that version 2.2 would include a focus on AI.

Analysing DigComp 2.2 against the data literacy competence model

In the coding of the cluster using data (Table 1), coding is uneven over the different categories interpreting, navigating, collecting, processing and presenting. Interpreting (3), navigating (7) and collecting data (1), which refer to an active and goal-oriented use of data, only encompasses a limited number of examples. Within navigating data, examples 35 and 75 explicitly refer to an open data interpretation of data literacy. Example 35 states: Knows that open data repositories exist where anyone can get data to support some problem-solving activities (e.g. citizens can use open data to generate thematic maps or other digital content). Example 75 notes: Knows how to monitor public spending of local and national government (e.g. through open data on the government’s website and open data portals). The rest of the examples are more hands-on navigating or searching information and data.

Processing and presenting data are mentioned 9 and 11 times, respectively. At first sight, one would presume that DigComp 2.2, thus, puts a lot more emphasis on using existing data and data sets than on creating or collecting own data. However, even within these categories, not that much emphasis goes to actively handling data. Example 40 is such an example in processing data: Can use data tools (e.g. databases, data mining and analysis software) designed to manage and organise complex information, to support decision-making and solving problems. However, within the category of processing data, seven examples have a privacy and security focus. These include the handling of sensitive data on personal devices (example 175) or verification of metadata (example 111).

In the category of presenting data, several examples focus on processing data to present data to an audience. This ranges from applying basic statistics to visualise data (example 37), to interacting with and manipulating data visualisations. Example 38 says: Knows how to interact with dynamic data visualisation and can manipulate dynamic graphs of interest (e.g. as provided by Eurostat and government websites). In this category, multiple examples refer to the more creative and goal-oriented handling of data, such as supporting your own ideas and opinions by creating digital content (example 124) or combining various types of digital data (example 127).

In relation to creative use, example 131 focuses on the creation of infographics and posters with software and example 133 focuses on the integration of digital technologies, hardware and sensor data to create new artefacts. Some of the examples used require highly complex skills and almost move into professional use of data. Example 134 focusses on: Knows how to incorporate AI edited/manipulated digital content in one’s own work (e.g. incorporate AI generated melodies in one’s own musical composition). This use of AI can be controversial as it raises questions about the role of AI in artworks and, for example, who should be credited. Or 246: Open to engage in collaborative processes to co-design and co-create new products and services based on AI systems to support and enhance citizens’ participation in society (AI). It becomes clear from this analysis and the concrete examples that a lot of emphasis is placed on presenting data in a creative and clear way. The processing of data is focused not only on the analysis of the actual data, but also on processes related to dealing with privacy sensitive information and being able to check metadata for example. Only one example is focusing on data collection.

As indicated, 53 examples are coded in the cluster of understanding data (Table 2). Coding is more equally distributed over the categories observing (18), analysing (9), evaluating (16) and reflecting (10). The category observing predominantly comprises rather simple things users need to know about data and how data is created in a digital society, such as knowing that communication services are free because they are partly paid for by user data (example 43) or that data needs to first be digitally encoded before it can be processed and analysed (example 32). Example 43 states: Knows that many communication services (e.g. instant messaging) and social media are free of charge because they are partly paid for by advertising and monetising user data. Observing also explicitly refers to knowledge about EU-policy in relation to data, especially with a focus on privacy and security, for example, knowledge about the rights of access, rectification and erasure or right to be forgotten.

In the category analysing, we find examples that require a deeper understanding of the underlying processes of data, with examples emphasising the different aspects of the role of AI in personalisation (example 4) and large amounts of data collection with sensors and applications such as wearables and virtual assistants (example 32). The examples we coded in this category seem to refer to the working of AI and increasing integration of AI into various aspects of digital technology and its implications for users.

In the DLCM, the category evaluating looks at whether consequences are harmful or constructive. This weighing of consequences is present in 16 examples. Example 14 notes the ability to weigh the benefits and disadvantages of using AI-driven search engines, and example 117 notes that users should be able to identify positive and negative implications of data use, especially when it comes to personal data use for AI-driven technology. A lot of the examples in this category refer to competences necessary to assess bias, accuracy, unrecognised processes of AI-systems, which are indeed important competences in relation to AI.

Reflecting in the DLCM refers to the minimising of harmful consequences and the societal impact, referring to the ethical considerations (example 79) and lifelong learning in the context of AI (example 225), highlighting the importance of ethical awareness and continuous learning related to all things AI, although the latter is a competence only few people will actually achieve. Example 79 probably reflects this the best: Readiness to contemplate ethical questions related to AI systems (e.g. in which contexts, such as sentencing criminals, should AI recommendations not be used without human intervention) (AI)? From this analysis it becomes clear that most examples in the observing category relate to insights in the way data is being used (e.g. as a currency for free services) and the rights people have related to their data, for example the right to be forgotten. The second biggest category of examples focus on evaluating, which also takes a more normative approach (e.g. evaluating whether something can be harmful).

Analysing DigComp 2.2 thematically

In the thematic analysis, we made an overall distinction between data and AI (Tables 3 and 4). In all, 49 examples refer to data and 35 to artificial intelligence. For the examples coded under the thematic category of data, three are very general in terms of understanding data. Examples 154 and example 162 highlight the ability to recognise and work with input and output data in simple programs. A second thematic category is the one on problem-solving. Only four examples fit this category and zoom in on understanding how data is utilised in various domains, especially in problem-solving and decision-making processes. Example 40 states: Can use data tools […] designed to manage and organise complex information, to support decision-making and solving problems. Example 40 is a direct illustration of the open data field approach to media literacy.

The rest of the examples in the overall category data are coded in the categories security (7), identity, privacy and ethics (12), representation of data (14) transparency (5), environmental impact (2) and citizenship (2). An example representative for security is example 167: Knows about measures to protect devices (e.g. password, fingerprints and encryption) and prevent others (e.g. a thief, commercial organisation and government agency) from having access to all data.

Here, we highlight interesting insights from four categories, namely, representation of data, transparency, environmental impact and citizenship. The examples coded in representation of data have to do with being able to create and present data and being able to read data in its different forms. Example 124: Knows how to create digital content to support one’s own ideas and opinions (e.g. to produce data representations such as interactive visualisations using basic data sets such as open government data). Example 124 can also be seen as an example of the open data field approach to data literacy. In fact, in the category of representation of data, more entries could be interpreted in this way. The category of transparency is closely related to the former, revolving around the importance of verifying information (sources) from the internet and recognising possible biases in data, particularly on platforms like social media (example 18). Recognising that data can be shared in a biased way and data handlers can have a biased point of view, strongly relates to the concepts of data feminism and data justice, requiring a critical interpretation of data.

The category on environmental impact only has two examples that stress the positive and negative impacts of technology. These examples focus on the intersection of digital technology and the environment, emphasising the potential for digital technologies to contribute to energy efficiency (example 208) and underscoring the importance of understanding and mitigating the environmental impact of digital practices, for example, carbon emissions (example 203). The category on citizenship has examples that again are in line with the open data field view, such as, example 75: Knows how to monitor public spending of local and national government (e.g. through open data on the government’s website and open data portals).

For the 35 examples coded on the overall theme of artificial intelligence, 5 are very general in nature, for example, awareness about the uncertain impact of AI. Example 251 notes: “Aware that AI is a constantly-evolving field, whose development and impact is still very unclear (AI).” Again, the theme of identity, privacy and ethics has a lot of examples, 12 in total. The examples are very much in line with the examples related to data. In relation to the theme of AI, examples state AI awareness and transparency, highlighting different aspects of understanding and being aware of artificial intelligence (AI) systems, for example, user profiling or black box decision-making (examples 5 and 105). The next three themes recommender systems, search and bias in algorithms refer to the working of AI and warns us for possible biases related to AI and underlying algorithms, which relates to the abovementioned category transparency. The examples in these themes underscore the importance of being critical, informed and discerning when using AI-driven tools and platforms, like search engines (example 14). This especially when it comes to AI bias, for example, AI algorithms embodying a political message, thus having certain negative consequences like perpetuating stereotypes (example 21).

The last two themes are in line with the societal impact we already saw under data, that refer to environmental impact and citizenship. The examples given are highly similar. In relation to environmental impact, example 209 warns that we should be: Aware that certain activities […] are resource intensive processes in terms of data and computing power. Therefore, energy consumption can be high which can also have a high environmental impact (AI).

Discussion

We observe that within social and educational sciences, the prevailing focus in the literature is on proficiency in using data. When articles focus on using data, the primary areas of focus are typically statistics, numerical proficiency and data handling. However, there is a growing trend towards enhancing the critical understanding of data. When the focus is on understanding data, we see links with equity and inclusion, ethics and citizenship. This is in line with the DigComp framework. Furthermore, there is an emphasis on the interconnectedness with media literacy, mis- and disinformation and data journalism. There is much less attention on personal data protection and privacy in the literature, aspects of notable prominence in the DigComp.

Other DLCMs and frameworks, however, often focus on data usage skills. A well-known model is that of Risdale et al. (2015) that identifies 22 competences for three proficiency levels. The model is an interesting start for data literacy, but is unbalanced because it mainly focusses on using data and, thus, looks less at participation, empowerment or citizenship. Existing models tend to follow the steps and logic we see in research and, consequently, start with searching for and collecting data. In other words, they often assume a research question aimed at actively searching/collecting appropriate data and using data to improve services. An example of this is the framework developed by Gummer and Madinach (2015). This model provides a useful overview for teachers on what competences they need to effectively use data to improve education, while also taking into account data ethics in later publications (Mandinach and Jimerson, 2022). This trend toward a more critical data literacy is also noticeable in recent literature, such as the data citizenship framework put forward by Carmi et al. (2020).

In contrast to older models, the DLCM starts with being able to interpret data. From a didactic point of view and in terms of necessary knowledge acquisition, this is more logical in structure (Seymoens et al., 2020) and considers not just researchers, librarians or students, but all citizens. As mentioned above, policymakers and educators in Europe and worldwide direct their attention towards specific topics and competences put forward by the DigComp framework, thereby influencing the capabilities of children, young people, but also other citizens. This affects how young people are and will be able to actively engage as future voters, members of the workforce and parents and partners. And it not only bears significance in the context of participation and engagement, but also has implications for emerging jobs, such as bias auditors, and the growing demand for professions like data scientists and engineers. It will also have an impact on and reshape existing occupations. For instance, teachers will need proficiency in working with educational data to enhance their teaching, or health professions will need to adapt to use AI- and e-health tools more effectively and safely.

The importance of this critical data literacy education is closely tied to the discussions surrounding data justice. As Dencik and Sanchez-Monedero (2022) state, this concept highlights the importance of considering data as an integral part of social justice, rather than something technical and purely numerical. In an increasingly data-driven society where data has become a valuable commodity (Zuboff, 2015), the perspectives of data justice, data feminism and critical data literacy education, which emphasise understanding data and datafication, gain even more relevance. While DigComp does not extensively address these aspects, there is a noticeable shift within the framework towards a more comprehensive and critical understanding of data literacy, albeit primarily at the level of individual data protection and privacy.

Conclusion

Reflections on the data literacy competence model

The goal of the DLCM is to provide a high-level model to orient a common understanding of data literacy competences within Flanders-Belgium. In this article, we used it to analyse the examples of the DigComp 2.2 Framework. This worked rather well as the DLCM clearly highlights the difference between on the one hand using data and understanding data and on the other hand differences in attention within the using data category with a clear focus on presenting data. As noted, while coding, we missed the specific category of processing data which covered many examples in the DigComp 2.2 Framework.

It forces us to reconsider and reconceptualise our own model. The model places significant emphasis on fostering critical thinking, but it is imperative that we conduct a more in-depth assessment of the understanding data component. We intend to thoroughly examine the model to ensure it accommodates competences related to data justice and data fairness. We will also need to test and validate the adapted model by mapping educational practices focussed on both using and understanding data.

Today’s society requires citizens to possess competences for using data and understanding data. However, it is important to note that using data does not inherently foster a critical understanding (Van Audenhove et al., 2020; Carlson and Stowell Bracke, 2015). Citizens need to be aware of and understand how data works and the potential risks that it can introduce (Polman et al., 2022). Understanding data then supports decision-making, participation and adaptability in an ever-changing society.

Reflections on DigComp 2.2

Reading the examples of DigComp 2.2 provides one with the impression that data literacy is well covered in the framework. All aspects of data literacy and the different fields of data literacy are covered. Our coding and analysis against the DLCM, however, clearly tells another story. DigComp 2.2 is not neutral, as codes are not distributed evenly among the different DCLM competences. DigComp 2.2 emphasises different themes, and it is clearly inspired by one of the data literacy fields.

A first observation to make is that DigComp 2.2 emphasises understanding data much more than being able to use data. In this respect, the DigComp 2.2 interpretation of data literacy is more inspired by the sociological field and less from an open data field or STEM field, which build more on using data. The more active and strategic use of data involving collecting data or actively using existing data, does not receive much attention. These competences are present but are only referred to in a limited number of examples. As indicated, only one example is coded within the specific category of collecting data. This is the reason why we split up collecting data in collecting and processing data. DigComp 2.2 has a lot of examples that can be coded under processing data.

Thematically, DigComp 2.2 puts a lot of emphasis on security and privacy, a theme less present in literature on data literacy education. Combining Tables 3 and 4, 31 of the 84 examples fit this category. It is logical that a competence framework strives to protect the user. However, in terms of balance, we could ask whether more than one-third of examples should focus on a protectionist approach to data literacy. Especially as other thematic coding such as representation and bias also have a certain focus on protecting the user. Topics such as environmental impact and citizenship which figure high in some of the discussions on data literacy are present in the examples, but only with a few observations. Competences related to these themes are essential, as citizens will be able to recognise the limitations of data.

This imbalance will affect the models’ users and affect their level of active participation in society. It will also have societal implications. European member states, their policymakers and teachers use the framework to set learning targets and adapt media education curricula. Therefore, citizens’ proficiency will be more in line with the competences and topics that DigComp puts forth, possibly resulting in a competence gap. It can also lead to an unbalanced curriculum and can further reinforce inequalities. With the deployment of AI and the continuous use of data, competences in recognising, reflecting on and responding to data- and AI-related injustices are required. When emphasising online privacy over, for example, citizenship-related competences, citizens will become proficient in safeguarding their personal data; however, they may lack the competences to consider ethical and equitable implications.

Seeing the importance of DigComp and its newest 2.2 version, we applaud the strong emphasis of the new framework on data and AI. By doing so, DigComp 2.2 has integrated a strong data literacy focus that will certainly have an impact on education in Europe. However, its lack of focus on actively using data might leave many teachers disappointed from a STEM perspective. We note that the framework accentuates “understanding data”. If we want to have a more balanced competence framework on data literacy, that includes a stronger focus on data research, statistics and mathematics for being able to use data, then that DigComp 2.2. will have to be complemented by another competence model.

The ongoing evaluation and validation of competence models and frameworks are essential to ensure that they are relevant to citizens. This is especially the case for competences related to the digitalisation and datafication of society, which evolves and brings new challenges with the continuous introduction of new software and technologies.

Figures

Figure 1.

Data literacy competence model (mediawiijs.be)

Figure 2.

DigComp 2.2: digital competence framework for citizens (Vuorikari et al., 2022)

Table 1.

DigComp 2.2 Coding on data literacy competence model: using data

Coding on data literacy competence model
Using data
Interpreting	Navigating	Collecting	Processing	Presenting
3	7	1	9	11
31

Source: Table by authors

Table 2.

DigComp 2.2 Coding on data literacy competence model: understanding data

Coding on data literacy competence model Understanding data
Observing	Analysing	Evaluating	Reflecting
18	9	16	10
53

Source: Table by authors

Table 3.

Thematic coding: data

Thematic coding
Data
Understanding data	Problem-solving	Security	Identity, privacy and ethics	Representation of data	Transparency	Environmental impact	Citizenship
3	4	7	12	14	5	2	2
		Security and privacy	Representation	Societal impact
3	4	19	19	4
				49

Source: Table by authors

Table 4.

Thematic coding: artificial intelligence

Thematic coding
Artificial intelligence
Understanding AI	Identity, privacy and ethics	Recommender systems	Search	Bias in algorithms	Environmental impact	Citizenship
5	12	2	2	8	3	3
	Security and privacy	Bias	Societal impact
5	12	12	6
		35

Source: Table by authors

Appendix