In this week’s article I’m going to go from the general to particular with respect to the vital role of ethics as an integral feature of data and AI governance. Rather than focusing on how and why the analytic tools of ethics are necessary in any model of good governance, I’m going to focus on a specific technology and what the role of ethics is for that in terms of optimising governance. I’ve done this on a few previous occasions, for example looking at contexts as diverse as forensic psychiatry, brain-computer interfaces, the Post Office / Fujitsu Horizon scandal, and so on. In this article I’m going unpack the importance of ethics in the governance of what are known as ‘digital twins’, and how we at IGS can help with both compliance and ethics-related dimensions of the governance of such systems.
What is a digital twin?
A ‘digital twin’ refers to the creation of a virtual model that accurately mirrors a physical object, system, or process. This means that the twin continuously receives real-time data from sensors and other sources, which enable it to replicate the operational status, behaviour, and performance of its physical counterpart.
Digital twins were originally pioneered in manufacturing and aerospace, but now find applications across numerous industries and sectors, including but not limited to healthcare, urban planning, and energy. An example of one from the latter might be a digital twin of a wind turbine that collects data on vibrations, temperatures, and wind speeds, and so providing the means by which the turbine’s operators can simulate different scenarios, improve their ability to predict failures, optimise maintenance, and so on.
As such, the central advantage of a digital twin of a physical system is its ability to simulate ‘what-if?’ scenarios and forecast outcomes without disrupting the real-world system. And this means that having a digital twin enables organisations to experiment, to test, and to innovate safely and cost-effectively. Unsurprisingly, given that component of the advantage of a digital twin related to prediction, these systems often use AI and machine learning-driven data analytics in the interests of continuous improvement of the real-world system which the twin is replicating. So, as an increasing range of industries adopt digital twins, this technology will influence how we design, operate, maintain and govern complex systems, through a bridging of the physical and digital environments.
Data governance in digital twin technology
Digital twin technology, in its ability to create dynamic, data-driven replicas of physical systems might conceivably have significant impacts across numerous industries. Given that potential significance at the societal level, it will be vital to ensure that the data governance processes, policies, frameworks, and so on that organisations use are not only robust in a regulatory sense, but also reflective of the ethical challenges that digital twins create.
I’m aware that my articles have a distinct theme; namely, explaining repeatedly why ethics is an integral rather than optional feature of good data governance, and increasingly so given the trajectory of technological development that we’re on. One of the points I sometimes make within this is that ethical governance, beyond compliance, is not a checklist. The reason for raising this is that, by now, we are all probably familiar with the laundry list of ethical issues that arise in data and AI governance, such as privacy, bias, transparency, accountability, and so on. Undoubtedly, these are all central ethical priorities, but it’s vital that anticipation of these challenges does not reduce to being merely formulaic. Focusing only on these obviously salient concerns can risk occluding deeper and more structural, philosophical questions that underlie the landscape in which a technology such as digital twins will operate.
So it’s this which I’d like to address in this article, going beyond the familiar list of ethical issues of which we’re all aware, to ask how we might build governance conditions that not only manage the obvious ethical risks but also ensure that digital twin systems operate in a way that is just and democratic.
The importance of data sovereignty
Of course, digital twins require data, but decisions about the particular data used and why are not necessarily always neutral, especially in those models which will include data about not only technological systems but those that involve humans. Often, governance approaches data stewardship in terms of risk management and compliance; but if it only does this, it overlooks values-driven choices inherent in the collection of the data itself. A relevant example here to illustrate why this matters could be if local government decide to build a digital twin of a city’s transportation network in the interests of increasing its efficiency. This would be an entirely sensible, civic-minded project, but of course, such a twin would not only be modelling roads and buses and so on, as in doing this it will also, implicitly, be mapping the movement and choices of its citizens.
So, given that there are people, and therefore moral agents with rights and interests, involved, it becomes necessary to ask other questions. For example, it matters who controls this data; it matters who decides what does and doesn’t get modelled and for what reasons; and so on. I use this example to remind us that appropriately robust data governance in such an example will need to grapple with the ethics of data sovereignty. Ultimately, in examples such as the one I’ve given here, this could mean ensuring that the communities whose movement is tracked have some kind of a voice in how their data is used, shared, and, potentially monetised. As such, building the means to achieve this into the governance arrangements would be an important ethical priority beyond the usual list of issues with which we’re all already familiar.
More on stewardship: dynamism and embedding ethics
I mentioned stewardship in the section above, and I want to push that point a little further here. As I’ve indicated, it is a risk of traditional data governance that in focusing on compliance it can be defined in terms of rules, checklists, audits and so on. As necessary as they are, all of these measures are relatively static; however, part of the value of digital twin systems is that they are, in a sense, living systems, insofar as they constantly evolve as new data streams in and the model is updated.
There is, then, a risk that an approach to governance which is too static will struggle adequately to keep pace with this dynamism. The risk of this is precisely what the role of ethics as an integral feature obviates, namely that if problems occur they will only be identified retrospectively; and this, in turn can lead to governance as fire fighting rather than fire prevention, where it’s the latter of the two which is clearly preferable.
I make this point to argue for a model of dynamic stewardship when it comes to the governance of these systems, rather than a model which relies on compliance alone. This means building processes that not only monitor data flows but also anticipate and adapt to emerging risks, such that governance can be iterative, adaptive and anticipatory. In practical terms, then, this means embedding data stewardship, and therefore the stewards themselves, within digital twin teams, rather than being external auditors who arrive at regular intervals but are not always present.
Digital twins and power
I mentioned earlier the need to dig beneath the surface of the standard ethical risks with which we’re all familiar, and look for more subtle, structural underlying ethical challenges when it comes to the governance of systems such as digital twins. To develop that point a little, perhaps the most overlooked ethical dimension of digital twins is their role as instruments of power.
For instance, and to return to the example of a digital twin which models civic infrastructure, a digital twin of a city can, with bad actors and the absence of good governance, become a means of wielding control. In that context, a digital twin could be used to optimise resources, but, given that it would track individual movement and choices, it could also be used for surveillance in ways that could be exploited for malign ends. So, when we think about power specifically, it once again becomes relevant to ask who decides what features of civic life do and do not get optimised, and on what basis, and who benefits from those optimisation choices.
It matters that data governance confronts deep structural challenges such as this, so as to ensure that digital twins don’t simply become ‘evidence-based’ levers to be used by power structures already in place. This means that embedding democratic accountability into digital twin systems, for example by participatory design processes or, more generally, adopting an ‘embedded ethics’ approach to system development is vital for being able to exert appropriate scrutiny over digital twin outputs and how they might influence policy development in the real world.
Responsibility as an ethos of governance
In stitching together these various points, I’m trying to argue that, fundamentally, good data governance in the context of digital twins is a matter of expansive responsibility, not only regulators or anyone with financial skin in the game of the system in question, but to the communities and long-term societal outcomes that we produce through such models as the scope for their application widens. To return to a point I made earlier, this responsibility cannot be outsourced or reduced to an ethics checklist; rather, it demands foundational engagement with decisions about how we should and should not build, deploy, and sustain digital twin technologies, and for what reasons.
So, by way of a summary before making some conclusions, I suggest that we should move beyond only asking the question of ‘is it ethical?’ to also asking questions about whose interests digital twin systems will serve, especially when the data includes data about people, and about the values that are embedded in the structural parameters of the twin. If we want to ensure that widespread use of digital twins conduces to human flourishing rather than only operational efficiency, then we need to start asking questions like this.
Conclusion
Digital twins offer significant promise, but they require a sophisticated and creative approach to data governance, of the kind that I’ve sketched out here. If we don’t adopt an approach like this, we risk building systems that are definitely efficient but might be unjust and are definitely powerful but might have terminal gaps in accountability. As I keep saying in these articles (and I’m sorry if the message is getting boring if so), the challenges and opportunities here turn on thinking about data governance not as a bureaucratic hurdle, or a way to default to saying ‘no’, but as an essential feature of the architecture required for ensuring the trustworthiness, and therefore ethical legitimacy, of digital twin technology.
Given the approach that I’ve set out for thinking about optimal governance of the rapidly expanding field of digital twins, if it resonates with your concerns or you think it can meet your needs as an organisation using digital twin technology, get in touch with us at IGS, as we are here to help.
