About two years ago, I was invited to a digital twin lab based at a university, which encompassed around 20 engineers and other types of modelers, and was asked to present some of my reflections on in silico medicine from a social science perspective during their weekly lab meeting. After my presentation, I was asked if I had any questions for them. I found myself debating whether to ask the question I truly wanted to ask or to play it safe and focus on building my ‘credibility’. Eventually, I used the occasion to raise something I had been struggling to understand since the beginning of my ‘digital twin journey’: when does an in silico model become a digital twin? The literature seemed to point out that this has to do with a model’s maturity level.
Looking around the room and noticing a mix of laughter and perplexed expressions, it quickly became clear that this was not a straightforward question. A lab member then said that they wished they could answer, but they themselves were uncertain. This honesty was quite unexpected. Others began to share anecdotes, including one about a recently submitted paper in which the authors were asked by a reviewer why they were using such an outdated term as ‘in silico model’. They were told to replace it with ‘digital twin’. Another researcher recounted how they had been asked by conference organisers to use the term ‘digital twin’ in their presentation because it appeared in the event’s title. Hence, the researcher had to reframe their model as a digital twin rather than presenting it as originally intended.
Both these instances make visible that there are socio-political drivers motivating the use of the term "digital twin", a factor that has been omitted in the literature, yet one that can help us better understand what we are actually talking about.
So, what is a digital twin supposed to be?
The idea of developing a digital twin, a virtual replica of a human body, has gained significant traction in biomedical research and is said to inspire a ‘new era of personalised medicine’ (Highfield & Coveney, 2023). While the vision of a ‘comprehensive digital twin’ remains fiction, albeit a guiding vision, progress is well underway to model things such as cells, biological and physiological processes and individual organs. These computational models, also referred to as in silico models, are part of a broader push to explore how simulations can inform medical decision-making, from testing treatments to understanding diseases. The most advanced digital twins tend to be hybrid models, meaning they are both knowledge-driven and data-driven.
Now, this is how such digital twins are usually introduced. The focus tends to lay on its promises and expectations, and what it ought to be. It is not difficult to portray digital twins as a charismatic technological idea considering how it plays with the fantasm of the digital double. Hearing about the idea of a ‘virtual you’ naturally sparks curiosity. As Ames (2015) argues, a ‘charismatic technology’ draws its power from two things, the experiences it enables and the promise of what it could achieve. The significance lies not in what it is but what it claims it can do and, through that, operates within the frame of technological progress as inevitable. In the in silico field, this sense of inevitability is often reinforced by pointing to the growing uptake of simulations and modeling across other sectors.
A (not so) simple question
There is no consensus on the definition of a ‘digital twin’ in medicine, despite efforts by institutions such as the leading Virtual Physiological Human Institute (VPHI) and the Avicenna Alliance to address this fragmentation, as well as academic articles that have attempted to clarify the distinctions between a model and a digital twin (see Wright & Davidson, 2020). The struggle to define and to enforce definitions is amplified by the more mature use of digital twins in other sectors, such as aerospace and manufacturing. However, one aspect that all definitions seem to agree on is that digital twins are some type of virtual representations of real-world systems and processes.
However, as an STS-inspired researcher, my focus here is not on proposing fixed definitions or contributing to definitional debates, as the boundaries of what constitutes a digital twin remain fluid, situated, temporal and multiple (Mol, 2002). The focus in this blog post, as shown through the fieldwork instance, is to highlight how they are enacted. In particular, the differences in performativity as collectives use different definitions for different purposes.
Not Magic, Not New
The takeaway is the following: digital twins are not straightforward technological breakthroughs but products of ongoing negotiations and reframing of the central concepts of the field, shaped by visions of desired futures and decades of progress in Computational Modeling & Simulation (CM&S). It would be a mistake to frame medical digital twins as something radically new just because the term has been buzzing (Bensaude-Vincent, 2015). We need to view them as fluid entities that are enacted through practices and that are part of a longer research trajectory. This is also of interest to those advancing the technology, as they deal with the tension between slow-moving research and the funding machinery that often fuels hype. As researchers in my fieldwork have pointed out, they become embedded in the hype, whether they want to or not, as these concepts inevitably land on their work.
Once digital twins are de-mystified, we can take a more grounded, un-hyped, perspective, one that resists the allure of promises. They did not appear out of nowhere: they are modelled, deployed, and used by humans who embed their ways of thinking (and values) into it - and the same goes for the term "digital twin" itself. They are, after all, the result of human practices. So, instead of getting caught up in definitional frictions and hype, the next question we should be asking is: Are digital twins designed for our healthcare systems? Or, are they just a new spin on the same promises of personalised healthcare, but packaged differently?
References
Mol, A. (2002). The body multiple: Ontology in medical practice. Duke University Press.
**
Elisa Elhadj is a Doctoral Researcher in the Reimagining Public Values in Algorithmic Futures project led by professor Minna Ruckenstein. She is currently working on a publication on the topic of this blog post.