Presentation at Aalto University on 12 May 2022 on how information systems discipline needs to study digital data as fully fledged artifacts. The presentation is based on a paper coauthored with Marta Stelmaszak.

1. 12 May 2022
Aalto University
Beyond the Facts:
Data as Digital-Semantic Artifacts
Aleksi Aaltonen (a paper co-authored with Marta Stelmaszak)
aleksi@temple.edu
@aleksi_aaltonen

2. Data are the lifeblood of all digital systems and devices that cannot operate without
the steady flow of data between their interfaces and algorithmic processing units.
A standard view in IS considers data narrowly as raw, unorganized facts – the closer
examination of data as factual representations of reality has mainly been left to
computer scientists, engineers, and data modelers.
The role that data assume in business and society has expanded dramatically over
the last two decades, calling for a re-assessment of how we think about the digital
stuff some have described as the new ‘oil’.

3. The Factual (Standard) View of Data
The factual view of data has its roots in engineering and objectivist philosophies that consider data
as more or less truthful representations of events or entities in a domain of reality.
In this view, data are considered external to the domain of reality that they represent, offering a
sort of freeze-frame representation of the phenomenon of interest.
Problems with data are mainly seen as a matter of correspondence between the domain of reality
and its representations in data, that is, how accurately and reliably data represent external things.
However, the factual view engenders assumptions that limit our capacity to answer important
questions that emerge in increasingly datafied settings.

4. The Factual View Creates Blind Spots
Contrary to the established Data-Information-Knowledge-Wisdom (DIKW) hierarchy (Ackoff, 1989),
Tuomi (1999) argued already twenty years ago that knowledge is as much the foundation of data as
data are the foundation of knowledge.
More recent critiques show that data are of human creation, embed decisions, and require
interpretive agency to have a capacity to meaningfully re-present anything from a domain of reality
(e.g., Aaltonen & Penttinen, 2021; boyd & Crawford, 2012; Jones, 2019; Gitelman, 2013; Kitchin,
2014) .
Digital data tokens are fundamentally different from earlier carriers of facts. Datums are inscribed
on a computational medium whose attributes affect how the data operate in different contexts and
how they become entangled with organizational practices.

5. Data Are Not Just Facts!
For instance,
1. Economics has framed digital data as intangible goods described by the attributes of non-rivalry
and zero marginal cost of reproduction (Llewellyn et al., forthcoming).
2. Science and technology studies has investigated how data relate to reality, facts, and knowing
in practice (Bowker, 2005; Gitelman, 2013; Knorr Cetina, 1997; Latour, 1999; Latour & Woolgar,
1979; Leonelli, 2020).
3. Data justice and communication scholars have drawn attention to how people are made visible
and treated through data (Dencik et al., 2019; Leonardi & Treem, 2020; Taylor, 2017).
The field of information systems has until now largely stuck to its narrow view of data despite IS
scholars being in a perfect position to study data.

6. Data as Artifacts
Perceiving data first and foremost as factual inputs to algorithmic systems leaves
several important questions open.
The factual view makes it difficult to answer questions that concern: i) the genesis
of data, ii) the emergence of novel data-based entities and, for instance, iii) how
data become impregnated with different worldviews and human interests.
To help answer these and other emerging questions, we attempt to articulate what
we call the artifactual view of data and its promise to the field.

7. The Artifactual View of Data
We build on earlier agenda papers by Jones (2019), Jarvenpaa and Markus (2020),
and Parmiggiani and colleagues (2022) as well as an emerging body of studies (e.g.
Aaltonen & Tempini, 2014; Alaimo & Kallinikos, 2017; Alaimo & Kallinikos, 2020;
Bechmann & Bowker, 2019; Hron et al., 2021; Jarvenpaa & Markus, 2018; 2020;
Kallinikos, 1995; 2006; Monteiro & Parmiggiani, 2019; Parmiggiani et al., 2022;
Østerlie & Monteiro, 2020; Østerlund et al., 2020).
Central to these papers and, more generally, to works that go beyond the factual
view, is that they acknowledge in one way or the other the existence of digital data
as human-made artifacts (cf. Orlikowski and Iacono 2001).

8. Two Emerging Sub-streams of Research
1. Digital inscriptions become effective data only by becoming entangled in and by
being performed through organizational practices. Data artifacts are involved
in performing ‘artificial facts’, that is, the re-presentations of events and entities
as they are actively produced at every stage of data modeling, capture,
circulation, analytics, use, and reuse in a particular setting.
2. At the same time, data artifacts could hardly be performed and thus support
the enactment of facts unless they have some enduring attributes that people
engage through their practices.

9. Some Implications of the Artifactual View
1. Data are real and often exist as distinct artifacts in the domain of reality that is represented by
the data and, consequently, data artifacts may come to shape that which they represent.
2. To perform their referential function, that is, represent facts, data need to adhere to a system
of semantics that cannot derive from the represented phenomenon itself but is a feature of the
data production arrangements.
3. The artifactual view argues that data do not ‘have’ a structure but data artifacts are made
possible by (layers of) rules, whether called as languages, codes, structures, or grammars that
together allow to make sense of digitally inscribed distinctions and, among other things, give
data the capacity to represent external events and entities.

10. Why a New Perspective Is Needed Now?
1. Institutional frameworks and traditional expertise play today a substantially diminished role in governing
the production of diverse data than before (Alaimo & Kallinikos, 2020; Kitchin, 2014). There are simply not
enough experts to act as gatekeepers and the guardians of the semantics of data!
2. Attempts to capture faithful representations of events and entities run into unresolvable ambiguities if the
very categories used to describe reality and the processes of how entities are assigned to them become
fluid and subject to renegotiation.
3. Data are increasingly circulated, used, combined, repackaged, and reused across organizational and even
industry settings, which can make their relationship to an original referent reality ambiguous (Alaimo et al.
2020; Jarvenpaa & Markus, 2020; Llewellyn et al., forthcoming).
4. The (re)combination of data tokens is governed by semantic rather than functional rules, which makes
data different from other types of modular components discussed in the digital innovation literature.

11. Toward a Research Agenda
The relevance of the artifactual view of data emerges against the current data
revolution, which suggests that there is not just much more data available than
before, but the data are also produced and used differently.
We outline a research agenda:
1. Comparison of the factual and the artifactual views
2. Emerging research questions in datafied settings
3. Data as digital-semantic artifacts

12. FACTUAL VIEW ARTIFACTUAL VIEW
The purpose of data Data provide an accurate representation of
reality.
Data provide a semantic foundation for
representation.
Relationship to a
referent reality
Data exist as disembodied facts outside the
domain of studied reality providing a sort of
freeze-frame representation of reality.
Data exist as artifacts embedded in the domain of
studied reality, enabling actors to enact facts that
may become part of the represented reality.
The materiality of
data
Data are immaterial facts that do not have
materiality.
Data are digital artifacts whose materiality is defined
by their computational makeup and lack of physical
dimensions.
Relationship to
knowledge
Data are a foundation for knowledge about a
phenomenon of interest.
Data production entails knowledge about a
phenomenon of interest.
The semantics of data The factual view is agnostic about the specific
semantics of data-based representation.
Interacting layers of semantics define data as the
foundation of representation.
The role of data
structure
Data exist in an unstructured or structured form,
which defines the kinds of algorithmic operations
that can be applied to the data.
Data are made possible by a capacity to structure
observations meaningfully, which defines how reality
can be captured in the data.
Typical methods Econometrics
Behavioral experiments
Case study
Ethnography

13. Types of Emerging Research Questions
1. How are new kinds of data created?
2. How do data give rise to novel socioeconomic entities?
3. How do employees cope with data in their environment?
4. How data become impregnated with different worldviews and
interests?

14. Data as Digital-Semantic Artifacts
The artifactual view articulates a much-needed perspective for studying how data work in practice,
yet its theoretical potential lies in the analysis of the convergence of the digital and semantic
character of data artifacts.
1. Problematizing the structure of data as digital objects. Data are digital objects whose
granularity, malleability and the low cost of production combine with the diminishing role of
institutional frameworks and traditional expertise in governing production of data to offer a
substantially more fine-grained, multifaceted, and volatile basis for representing the reality
than before.
2. Understanding data as semantic objects. Data are made of several layers of compositional
rules not unlike digital innovations that emerge from the layered modular architecture (Yoo et
al. 2010), yet the composition of data is governed by semantic rather than functional rules.

15. Other recent takes on data…
1. A cognitive view of data is being developed by Jarvenpaa
2. Conceptual modeling as mediator between digital and physical
reality (Recker et al. 2021)
3. Data as infrastructure (Monteiro)
4. McKinney & Yoos theorizing about information

16. Thank you!