With the increased digitisation of society comes an increase in the role of data. Business analytics, statistics-based AI, the development of digital twins, etc, are typical examples of “data hungry” applications. Such, “data hungry” applications not only need data in different shapes and forms, they also need data from a wide variety of sources. The systems involved in gathering, storing, processing, analysing, and visualising data, have evolved to be complex systems themselves, involving many actors of widely differing nature. We argue that, as such, these complex systems can be best thought of as 'data ecosystems', which we see as involving the entire complex of social / physical / digital actors which provide, own, sell, buy, exchange, manipulate, store, and use, data. Within these data ecosystems, one needs to deal with technical concerns regarding reliability, performance, interoperability, semantics, etc, as well as social concerns, such as value of data, privacy, trust, ownership, ethics, risk, etc. In line with this, we argue that there is a need to define / study 'data ecosystems' more closely, where we see a potential future role for the VMBO community.
Our society is transitioning from the industrial age to the digital age. With the increasing digitisation of society comes an increase in the role of data. Data is gathered from sensors, consequently stored, processed, analysed and visualised, and is eventually consumed by (human and / or digital) actors to enable them to gain insight and / or make informed decisions.
Business analytics, statistics-based AI, the development of digital twins, etc, are
typical examples of modern-day “data hungry” applications. For example, data is
essential for the training of statistics-based AI and the development of digital twins [
The, “data hungry” applications need to be “fuelled” with a wide variety of data resources. For example, ranging from: raw observations from different sensors / informants, processed and / or enriched artefacts in terms of e.g. predictive models, representations of intentions (e.g. plans, strategy documents, designs, etc), specifications
(source code, work procedures, etc), or norms (regulations, principles, policies, etc). Next to that, such applications also need data from a wide variety of sources, requiring the need to transfer ownership of data, or at least a transfer of the right to use the data.
We specifically use the term data, as, in line with e.g. [
As a result of the growing role of data as a key underlying resource, the systems
involved in gathering, storing, processing, analysing, and visualising data have evolved
to become complex systems themselves, involving different actors with their own
interests. We argue that, as such, these complex systems can be best thought of as ‘data
ecosystems’, which we see as involving the entire complex web of social / physical /
digital actors (i.e. an ActorWeb [
Within these data ecosystems, one needs to deal with technical concerns
regarding reliability, performance, interoperability, semantics, etc, as well as social concerns,
such as value of data, privacy, trust, ownership, ethics, risk, etc. For instance, as the data
involved may pertain to (the behaviour of) humans, privacy and ethical considerations
may clearly play a role. Furthermore, as the data has some correspondence to
“something” in the social, economical, or physical world, it is important to consider quality
of this correspondence. At the same time, some actors may have an interest in
maliciously changing the data, thus distorting this correspondence. Data also comes with
the question of ownership. Data may be of strategic value to some actors, leading them
to want to control / sell the access for others. For instance, [
A data ecosystem can also be regarded as a “data-management enterprise”, i.e. a networked enterprise with “data-management” as its primary business, where datamanagement refers to all data related activities (gathering, exchanging, manipulating, storing, using, etc.). Such a “data-management enterprise” will typically be embedded in a larger enterprise, where the latter focuses on a “regular” products / services.
The development of data ecosystems, as “data-management enterprises”, can clearly benefit from the use of enterprise modelling approaches. As such, the above considerations directly apply, while at the same time suggesting the need to more specifically capture data ownership, data lineage, value of data (to specific stakeholders), access control, data regulations, etc. 3
We conclude this discussion paper with some some possible research challenges in relation to data ecosystems. They are certainly not intended as a complete list of challenges, but should rather provide a starting point for a broader discussion at the VMBO workshop.
Data as a key resource – It is clear that data is a key resource in a data ecosystem. As
such, it generates several important questions:
What is the (potential) value of data? How to assess / express this?
What does ownership of data mean, also in relationship to “the original” (e.g.
behaviour / properties of a human being), and associated privacy concerns.
How to model the ownership, access to, the (potential!) value of data, etc, as well as
associated risks?
How to take these elements into due consideration when designing / developing /
evolving data ecosystems?
Trust at the core – Exchanging data requires trust between the (human) actors involved,
regarding (1) the way they handle the data and / or access to the data, (2) on how the
data is gathered (quality of data), and (3) the way data is used (ethics and privacy). This
results in several challenges:
What is “trust” in the context of data ecosystems, and what can threaten such trust?
How to conduct a risk analysis on how data is handled?
How to nurture / increase trust between different stakeholders?
Does the notion of “privacy by design” work in an (open and evolving) data ecosystem?
How to identify system risks for data ecosystem, and how to manage these?
Regulation of data ecosystems – Regulators are likely to have a need to regulate the
risks (see above), privacy concerns of data ecoststems, as well as possibly other
properties. This results in challenges such as:
To what extent can data ecosystems be regulated at all, given their open, and evolving,
nature?
How to express, and enforce, regulations on data ecosystems?
What are the possible risks that need regulation?
Data needs semantics – With the large amounts of data available to us, it is important
to also capture its informational semantics. Both to enable re-use and relating
(interoperability between) different data sources. Of course this takes us back to semantic
modelling [