Statistical awareness promoting a data culture

2.1Revolutions, evolutions, mega-trends

As a prerequisite for promoting improved data practices, there needs to be an analysis and understanding of what role data plays in contemporary society. In this context, it is by no means sufficient to devote attention to the technical processes, the infrastructure, the economics of data or similar. Likewise, it is not enough to focus the ethical and legal questions on the protection of individual privacy. With a proliferation of data, new dynamics have been triggered that have led to the replacement of the previous well-ordered (infra)structures of statistical production based in the public sector by other forms of (decentralised and individual) use and production of data. This may be accompanied by a great promise, as a result of which ubiquitous amounts of data would also include more equality and less dependence on authoritative and technocratic institutions. In a direct sense, the fulfilment of such a promise would depend on everyone actually having a sufficient level of competence to be able to draw good statistical analyses from the data sets. That is however less the point here. Rather, the indirect and accumulated risks associated with such an approach that relies significantly on decentralised competences and processes are of interest, i.e. when there is no longer a common statistical language based on standardised methodologies, classifications, etc. [9, 10].

In a traditional division of roles, there are those who produce data, facts, indicators, etc. and those who use such results, producers vs. consumers, to put it simply. Today, the situation is a little more complicated. The traditionally divided roles have become mixed, consumers can also become producers using widely available data and evaluation tools, everyone is a potential ‘prosumer’. In principle, this development is to be welcomed; it opens up diverse opportunities for participation and emancipation. At the same time, however, this development is also accompanied by substantial dangers:

These trends and phenomena can combine to form a cocktail that poses a threat to democratic societies. All too easily, so-called ’alternative facts’ are manufactured in the alchemy kitchens of populism to be held up (as equivalent) to scientifically based facts produced with high quality.

2.2Data worlds

Apparently, there is no single answer to the question of how statistics relates to democracy, and certainly not a simple one according to which more statistics would generally lead to more democracy. What is obvious, however, and has been described many times scientifically [11, 12, 13], is the interaction between the state in its very different forms on the one hand and statistics on the other. Statistics and the making of the state have been closely married for two centuries and since the time of the Enlightenment.

Today, the situation is more complex and difficult to assess insofar as there are different data worlds [14] in which diverging ideas and conventions have emerged regarding the quality of information and the role of statisticians in today’s society. “The concept of the world here emphasises objectification: the practice of action and the evaluative networking of things and people creates its own objectivity to which the practice of justification can refer, whereby objects, qualities and categories as well as their interlinking appear as given by nature and as a network like an ontology in social reality and thus appear situationally evident to the actors as a world.”22 Measurement, argues the Économie des Conventions [16, 17, 18], is necessarily grounded in conventions and norms, even if (or precisely because) they are not in fact visible or made to be transparent. “Boltanski and Thévenot [19] emphasise that there are a multitude of different conventions and associated forms of justification/orders of value/worlds. While these claim to be a complete basis for action in their own right, they can never completely eliminate the competing claim of the other value orders to be equally relevant as a basis for action. Social situations are therefore mostly characterised by a plurality of value orders. This plurality can take the form of a conflict between conventions. Conventions then enable each other to question the other ontologies, i.e. to criticise (denaturalise) their qualities/values. On the other hand, forms of compromise between the conventions can occur, as they become possible especially in organisations.”33 The fact that communities differ in terms of their values, convictions, quality concepts and norms rarely plays a role when it comes to defining or contrasting their respective professional profiles. Rather, the focus is usually directed towards a distinction regarding the methods, technologies and tools used. In this article, however, we will be concentrating on the other characteristics that distinguish data worlds from each other.

3.1The data value circle: From data to knowledge

“The data value chain describes the evolution of data from collection to analysis, dissemination, and the final impact of data on decision making.” [21] The process of statistics, from their preparation, through the collection of data and their processing, to their dissemination and evaluation, is regularly presented as a linear sequence of steps, e.g. in the Generic Statistical Business Process Model [22]. Although these seem to be rather details, it is essential that statistics firstly do not start with the collection of data and secondly should not be understood as a single and unique process. Rather, the beginning of a statistic is characterised by a question, a need for information or a model from a theory that calls for quantification (unemployment, inflation, need for schools, etc.). In the design phase of a statistic, in which the theoretical terms are translated into quantifiable constructs, one often does not start from scratch. Instead, one orients on the existing and tries to learn from it and add improvements or extensions.66 Statistics are not typically individual gems, but are part of a larger informational collection or jewellery of diverse, related facts.

What we call ‘data value circle’ corresponds to the statistical investigation or statistical enquiry or data investigation cycle and the long-established advocacy by both professional statisticians and statistical educators that teaching must reflect, and be built around, this. It is an essential criterion for assessing an educational structuring model in statistics whether it takes this broader view or whether it reduces itself to the narrower, purely technical process. Katharina Schüller’s Data Literacy Framework [23] approaches the question in this comprehensive sense. The data processes in the narrower sense are integrated into learning cycles that focus on evaluation and (re)design.

In a production process – and that is what the production of facts and evidence is – expertise and skills of several kinds are needed at the different stages and steps of the process. In later stages of processing, it is important to have a reasonable understanding of the upstream supply chain, even if one has to take its outcome more or less for granted. This roughly outlines what users of public statistics results should contribute at their stage: a reasonable understanding of what information product they are being provided with, what its quality profile is, what they can (or cannot) use it for, etc.

3.2Competences, competence levels

“Data literacy is the cluster of all efficient behaviours and attitudes for the effective execution of all process steps for the creation of value and/or decision making from data.” [23] With this broad decription of data literacy, it becomes very clear that it is about much more than just being familiar with techniques and methods.77 Rather, competence in a comprehensive sense involves a combination of knowledge, skills and attitudes. Furthermore, it is important at what level such competences must be available, which in turn – comparable to language competence – depends on the context in which and the purpose for which such competences are to be applied.

3.3Critical thinking skills

In the interest of empowering civil society with the aim of broad participation in evidence-based political discourse, it is not least a matter of being able to deal critically with statistics in addition to technical and methodological skills. This task can be approached in relation to statistics in general by analysing the characteristics of statistical literacy from the perspective of critical thinking [25]. With the eight critical thinking skills distinguished here, it should be possible to successfully develop a soundly based culture of data and to counter ‘crank’ science using sophisticated pseudo-statistics [26]. In the specific domain of civic, public and official data, the task of trustworthiness, transparency and institutionally based authority from the perspective of the users of statistics arises with particular urgency, far-reaching consequences for the democratic order and often without the existence of alternative data sources. What this means in concrete terms for statistics in the public sphere is explained in detail in the article “Literacy in statistics for the public discourse.” [27]

3.4Statistical and data literacy skills required in policy-making

As briefly described in the introduction, policy has several roles and tasks in the processes of evidence-based decision-making. Just as there are data and facts for policy, it is conversely important to emphasise the different policies regarding the framework for data on the one hand and for statistics on the other [28]. Accordingly, different competences are also required in the political sphere, depending on whether someone is preparing decisions on the analytical side of policy advice with the help of evidence, whether it is a matter of communication or whether the legal-political determinants for actors and action in informational value creation are to be designed.

For the one side, namely the use of evidence for policy-making, the competences and literacy required for doing so can largely be referred to what has already been elaborated. More interesting, however, is the opposite direction of activities, namely the politics for data and the politics for facts. In this respect, qualities and competencies that generally apply to good management are (additionally) important. Deming has established his criteria for profound knowledge, according to which successful management is based on the following four competences: “Knowledge about variation, Psychology (psychology of individuals, groups, society and change), Theory of knowledge, Appreciation for a system.” [29, 30] In the field considered here, the system for which ‘Appreciation’ is required is about a comprehensive and deep understanding of the processes and interrelationships related to statistics as

Towards a data culture

Several targets are required to achieve the abovementioned objectives. Some of the most important ones are listed here:

Overall, it is about actively promoting a data culture. This goal goes far beyond conveying ’digital skills’. In fact, in many cases, corresponding activities have already been initiated and have produced concrete results. Such activities should be built upon and further networked and strengthened. The following actions (with reference to exemplary examples) are recommended to proceed and achieve the goal of a data culture:

This is an extensive catalogue of demands, which presupposes the political impulses to be set in the corresponding direction. At the same time, however, it is also crucial to set the appropriate course in science, for example by promoting the necessary interdisciplinary research.2121 Finally, it will also be crucial to motivate the institutions of official statistics and enable them to play an active role in this programme.