Self-similarity and synthetic biology: a possible fractal anticipation

ABSTRACT

In the field of synthetic biology, a promissory technoscience, researchers use risk-based and speculative scenarios to anticipate synthetic biology futures. Yet, there is a sense in the field of history repeating as synthetic biologists repeatedly deal with similar challenges. Perhaps a focus on contemplating the unexpected and attempts to deal with complex futures comes with a failure, paradoxically, to prepare sufficiently for the expected. This study develops a conceptual framework derived from fractals and analyses data from ethnographic involvement in the field synthetic biology to reimagine how patterns in synthetic biology practices repeat at different orders of scale. This study begins with the ‘pressure testing’ of a biofoundry and shows how synthetic biology repeats at different scales of organisation, interaction and time. The case is made for a possible fractal anticipation which would have the capacity to identify patterns and support innovators, social scientists and researchers to be better prepared for encountering similar developments at different orders of magnitude.

KEYWORDS:

• Engineering biology
• fractal
• anticipation
• similarity
• repetition

Introduction

Synthetic biology emerged around the start of the millennium as a design-led approach to biotechnology. Collaborations of molecular biologists, computer scientists, engineers, mathematicians, and others, have sought to find ways to make biology more predictable, more modular, and to abstract design from manufacture with an aim of making biology easier to engineer (Silver 2009). Synthetic biology uses concepts like ‘parts, devices and systems', ‘biobricks’, and ‘circuits’ as metaphors to apprehend designed biological matter (Calvert 2013; Frow 2013; O'Malley et al. 2008). Synthetic biology practices are orientated to the future and the promises of making biology easier and faster to engineer (Delgado 2016; Frow 2013) and are concerned with delivering on the promise of accelerating engineering of biology (Hilgartner 2015). Synthetic biologists repeatedly revisit the challenges that seem to continually prevent the realisation of promise, including the inherent variability of living matter and concerns of industrial translation and societal impact (e.g. Brooks and Alper 2021; Hanson and de Lorenzo 2023; Kwok 2010).

The emergence of synthetic biology has coincided with that of responsible (research and) innovation (RRI), and concepts including anticipation, reflexivity, integration and responsiveness (Barben et al., 2008; Stilgoe, Owen, and Macnaghten 2013). Anticipation in responsible innovation is conceived of as dealing with futures (Stilgoe, Owen, and Macnaghten 2013) and there are many attempts at considering the futures of synthetic biology in governance and policy, social science, and scientific literature. Synthetic biologists deploy concepts of risk to deal with futures (Kallergi, Asin-Garcia, and Martins dos Santos 2021; McLeod, de Saille, and Nerlich 2018; Taylor and Woods 2020). Speculations on what may come to pass can begin with provocations of, or working up, future scenarios (Betten et al. 2018; Frow and Calvert 2013b). Synthetic biology futures can also be related to personal circumstances or the fate of science rather than anticipation about future societal impact (McLeod, de Saille, and Nerlich 2018). At the same time, synthetic biologists and social scientists comment on earlier issues and problems reappearing (Balmer et al. 2015; Casini et al. 2018). Such apparent repetitions occur with implications for the development of fields of research, as well as having affective connotations (Guggenheim and Nowotny 2003) and in my personal experience, scientists and social researchers act surprised when concerns about integration, collaboration, commercialisation or publics reappear again and again. Perhaps, a focus on contemplating the unexpected and attempting to deal with complex futures may come with a failure, paradoxically, to prepare sufficiently for the expected.

This argument deals with recursiveness. It is a sense of history repeating in science, innovation and collaboration that this analysis tackles by developing a conceptual framework derived from fractals that may support forms of anticipation in innovation more broadly. It concerns forms and patterns that seem similar, familiar, or reminiscent of other patterns, but at different scales. Scaling is important to synthetic biology innovation – through concepts like ‘scale-up’ (Li 2023; Zhang et al. 2021) – partly because doing things on smaller scales implies lower stakes in terms of resources. However, those stakes would only be lower if lessons are taken forwards and integrated through anticipatory processes at the next iteration. In what follows I argue that learning from a fractal analysis may help to identify patterns of repeats that happen at different scales (Abbott 2001; Jensen 2007) and is thus explicitly attentive to iteration. I seek to emphasise a descriptive sense of similarity where the qualities of forms and patterns are foregrounded. A fractal imagination of this kind offers the potential to approach anticipatory work in responsible innovation through the metaphor of self-similarity, which assumes that it would be possible to identify patterns in innovation processes and extrapolate those, rather than imagining the future is somehow different and special in the sense that solving problems at one scale and time would mean they would not re-emerge at another. In this way, fractals reprioritise what should be learnt from projects: it is not just research findings, but it is the sociotechnical patterns, problems and matters of concern that may be reproduced at other scales.

The argument proceeds as follows: first, I discuss anticipatory work in synthetic biology and touch on RRI more broadly; second, I develop an analytical framework by highlighting some pertinent uses of fractal figures in social science analyses; third, I introduce more details regarding synthetic biology and biofoundry facilities; fourth, I explore the emergence of ‘pressure testing’ synthetic biology infrastructures and the self-similarities that appear using the idea of fractals. To close, I discuss how the preceding observations and analysis may prove useful to innovators, policymakers, scientists, and social scientists.

Risk-based anticipation and speculative futures

Synthetic biology's future orientation can be discerned in practice (Delgado 2016) and in the array of policy documents such as roadmaps (Le Feuvre and Scrutton 2018; Patwari, Pruckner, and Fabris 2023; Si and Zhao 2016; Synthetic Biology Roadmap Coordination Group 2012) that describe and aim to govern future states (Marris and Calvert 2020). Synthetic biology innovations can raise a range of complex anticipatory concerns for governance and publics (Meckin and Balmer 2018; Pansera et al. 2020; Ribeiro et al. 2023; Ribeiro and Shapira 2019). However, synthetic biologists can frame RRI concerns in terms of risk, which can come in different forms including risks to society, risks to science, and personal risks such as careers and mental health (Kallergi, Asin-Garcia, and Martins dos Santos 2021; McLeod, de Saille, and Nerlich 2018; Taylor and Woods 2020). Such a view tends to enact a separation between synthetic biology and society, where the former will impact the latter, and is evident in synthetic biologists’ own accounts of ‘societal views’ (Kallergi, Asin-Garcia, and Martins dos Santos 2021, 21). Risk framings, while tackling particular potential concerns, may limit the range of possibilities for considering futures because the risk is often approached in terms of safety and may normalise RRI in existing ethical frameworks (Taylor and Woods 2020). Alternatively, after anticipating future-orientated concerns and framing those concerns as to do with risk and safety, synthetic biologists propose technical fixes. For instance, synthetic biologists plan to create ‘kill switches’ in genetic code (Wright, Stan, and Ellis 2013), produce other ‘genetic safeguards’ (Kallergi, Asin-Garcia, and Martins dos Santos 2021), or may attempt to genetically engineer away concerns at the molecular level (Meckin 2016). While these are noble attempts at acting responsibly, taking into account ‘societal views’ with a particular set of resources, they are self-vindicating because synthetic biology is proposed as the fix to synthetic biology. Furthermore, since such fixes rely on reducing concerns about safety and risk and focusing on technical aspects, they are unlikely to address socio-technical complexity of futures in synthetic biology.

Social researchers have worked with synthetic biologists in different contexts to actively think through futures, ‘opening up’ (Stirling 2008) possibilities by using methods such as scenario-building (Betten et al. 2018; Stemerding et al. 2019), reflexive ethical engagement (Balmer and Bulpin 2013), and discursive methods using provocative ideas (such as the idea that the main global synthetic biology competition would be shut down) (Frow and Calvert 2013b). The latter authors say:

Such reflexivity is not oriented towards predicting the future, and certainly cannot eliminate uncertainty about the future, but it can assist with identifying areas of uncertainty and maintaining a degree of flexibility in response to unanticipated developments. (Frow and Calvert 2013b, 34)

These future-orientated initiatives help synthetic biologists explore potentialities and aim to support recognition of the ways that problems and possibilities are framed. Similarly, scenario-building using technological choice and societal imperatives can help ‘to get the science right’ and ‘to get the right science’, respectively (Stemerding et al. 2019, 222). However, it is not clear if synthetic biologists themselves recognise an issue of framing, or understand themselves as having choices about those framings, or their capacity to act.

RRI focuses on the process of responsibility, as much as the product, so cannot be a ‘tick-box’ exercise and instead should be designed in context (Smith et al. 2021). Changes due to RRI are hard to evidence in the field of synthetic biology because they are often concerned with everyday scientific practices (Pansera et al. 2020). At the same time, the methodological language of horizons, visions, foresight, and science fiction literacy may make these imagined, hypothetical and exploratory endeavours seem removed from present experience. They can set up a problematic of deliberating a future changed by an emerging technology, often at a grand, distributed scale where a technology has ‘rolled out’ and changed the future, creating a speculative discussion (Nordmann 2014). Considering the future can also mean that responsibility is passed around, such as whether the responsibility for the future of antibiotics lies with innovators or policy stewardship (Stemerding et al. 2019). In other words, learning and action are deferred to other groups in the future. Perhaps, because it is difficult to learn from anticipatory practices, project leaders may ‘find it difficult to include such future-oriented focus in their proposals, particularly when technological applications appear to be unrealistic’ (Delgado and Åm 2018, 4). Anticipatory work therefore presents a challenge due to ways it can be incorporated in action and evidenced in practice and, it is the contention here, contributes to apparent repetitions in emerging technologies. Much of the anticipatory work in synthetic biology is speculative or risk-based and focuses on safety concerns, technical aspects, or abstract futures. Approaches are ‘what if … ?’ style approaches to anticipation, where creative and imaginative capacities are needed to provoke and engage in discursive activities (Stilgoe, Owen, and Macnaghten 2013). However, because they may be divorced from synthetic biologists’ everyday experiences of scientific work it may be difficult to recognise the relevance and outcomes of RRI interventions. This, in turn, would make learning difficult to incorporate and potentially result in forms of repetition (Stilgoe 2018).

A twenty-year-old critique of initiatives of social sciences’ efforts regarding emerging technologies is that social science is itself repetitive.

The stock social science response to technical controversy appears to consist of a reiteration of the precautionary principle and a call for greater public participation in decision making. Until we can learn more about the nature of technological controversy that enables us to offer a richer range of responses, our ability to contribute to wider social learning will inevitably remain restricted. (Rayner 2004, 354)

This view can be read as applying to elements of RRI, such as a commitment to more inclusion in innovation (Stilgoe, Owen, and Macnaghten 2013), despite other emphases on creativity and non-prescriptive process. Thus, the exasperation at repeating patterns of generating emerging technologies includes the relevant social science approaches and distributes the onus of learning to the social sciences as well.

In the next section, I begin to develop the idea of fractals as a way to understand a form of anticipation that draws on repetition and recursiveness so that synthetic biologists and social scientists can think about what does happen and generate the resources to understand sociotechnical futures. As such, Rayner’s (2004) point about contributing to social learning will be revisited towards the end of this paper.

Fractal imagination

Fractals are geometric shapes that result from recursive mathematics where a ‘simple’ formula feeds back into itself over and over creating fascinating patterns. They are, theoretically, endless. A basic example: take a line and remove the middle third; repeat for the resulting two shorter lines; repeat for the resulting four lines; repeat ad infinitum (Strathern 2004, 3). This is Cantor's dust. Since the same formula repeats, any partial element contains similar elements in similar relations. Strathern (2004, 9) puts it beautifully when she says:

The whorls and involutions of these self-similar shapes that repeat motifs through any scale of magnification produce the most seductive visuals. It is the glistening details of the forms that remain – whether one is looking at the convolutions of clouds or the bifurcations of tree trunks, branches and twigs. These fractal graphics could describe the patterning of maps or genealogies, but they would be maps without centers and genealogies without generations. It is the repetition, the not-quite replication, to which the viewer is compelled to attend. (emphasis added)

Fractal imaginations legitimise a focus on detail – finding repeating motifs at smaller and larger levels of magnitude. Fractals are useful in helping understand apparently chaotic ‘natural’ systems, such as the shape of coastlines, the surface area of mammalian lungs, and the shapes of leaves. Statistical fractals are those like a coastline where the overall shapes of bays, the rock formations, and the grains of sand all exhibit similar levels of complexity. While mathematical fractals can be infinite, physical fractals are often bounded, with fractalness occurring between particular scales or limits.

Represented as lines, fractals exhibit similar relationships and complexity at every level of magnification or scale. Such a pattern develops some curious qualities. For example:

[a simple, Euclidean] one-dimensional line fills no space at all. But the outline of the Koch curve, with infinite length crowding into finite area, does fill space. It is more than a line, yet less than a plane. It is greater than one-dimensional, yet less than a two-dimensional form. (Gleick 1988, 102)

Whereas lines are one-dimensional, and planes have two dimensions, fractal lines occupy space more efficiently than a line, yet do not create a plane, so they have a fractal dimensionality that is more than one and less than two. Used figuratively, this idea offers a route other than the binary choice of selecting between singularity or multiplicity (Law 2004) – there are entities which are not conventionally singular or plural.

Social science has invoked fractal figures in different ways. Perhaps, most notable for my arguments, are those in anthropology and sociology: see: (Abbott 1990; Abbott 2001; Abraham 1993; Haraway 1991; Strathern 2004; Strathern 2011). Almost concurrently, in the 1980s, Andrew Abbott was working on his idea of ‘chaotic disciplines’ focusing on qualitative-quantitative divisions in US sociology while Marilyn Strathern was developing Haraway's cyborg through her Partial Connections (2004) – unconnected yet self-similar developments in sociology and social anthropology (Strathern 2012). Both Abbott's and Strathern's fractal arguments draw on recursive division. Abbott argues that bifurcations in sociology into qualitative/interpretivist and quantitative/positivist worldviews, each produce worldviews that can bifurcate again, with each worldview position capable of a division into a more interpretivist or more positivist position (Abbott 2001). In this way, whatever worldview position emerges can be further divided using the same formula. His arguments indicate continual and chaotic evolution of theoretical positions infinitely dividing, yet always repeating the central formula. Partial Connections deals with ethnographic comparison. It is structurally like Cantor's dust with each phase of the argument divided into two, divided into two, divided into two, but each leaving a remainder for further exploration. Strathern argues that empirical comparison and theoretical articulation never fully explain a phenomenon: they are selective; always leave elements unexplained; are always partial.

One of the key elements of fractals is that ‘irregularity’ or complexity is similar at different orders of magnitude. Fractality means that scale and complexity do not necessarily correlate. Instead, whatever scale one chooses to analyse (through method/ological choice) one can produce the same level of complexity (Jensen 2007; Strathern 2004). For example, rolling out an electronic patient record system produces complex sociotechnical issues whether at the level of national policy, at a given hospital, or within the practices of a single nurse:

Whereas the head of department invokes standards, agencies, hospitals, governments, and regions as his relevant associates, the nurse invokes software developers, clinicians, patients, and cardboard mock-ups. Regardless of the presumed scale of these interactions they all seem obsessed with the relationships which make their current interest and activities coherent and meaningful for themselves and others.

A second point follows, which is that in the making explicit of relationships (which I am trying to facilitate in my interviews) the quantity of information seems, indeed, to remain constant in each instance … . (Jensen 2007, 839)

In spite of the importance of such grand scales, they do not determine practical outcomes. Instead, they are interfered with by many other scales, which are deployed by multiple other actors often with the result of frustrating and delaying development efforts. (Jensen 2007, 848)

Problems of introducing a technology into a large infrastructure occur with similar complexity at different scales. The knowledge of how, say, management and organisational practice affect nursing will be useful, but will likely be differently practised at individual sites so resolving issues at one site may not resolve those at another. Knowledge of the complexity of nurses’ practices is crucial to understand changes and challenges emerging through the introduction of a particular technology. This implies that scales affect each other.

Manufacturing biology

Synthetic biology is partly premised on an imagination of another technological revolution (Hilgartner 2015) and the potential for manufacturing is deeply connected to the rhetoric of some parts of the field (Mackenzie 2013a). Industrial production of spider silk and antimalarial drugs have been high-profile examples (Paddon and Keasling 2014; Poddar, Breitling, and Takano 2020). Groups committed to engineering and economic gain have been prominent in lobbying for policies to support synthetic biology (Schyfter and Calvert 2015). An important synthetic biology motif is the design-build-test-learn (DBTL) cycle, which has been used to organise research, present research papers, apply for funding, and demarcate synthetic biology from other approaches to biology by being more rational and amenable to improvement through automation, see examples (Carbonell et al. 2018; Kitano et al. 2023; Opgenorth et al. 2019). Synthetic biology has a large annual event, the international Genetically Engineered Machine competition (iGEM), where interdisciplinary teams of university students learn about synthetic biology and contend for the best project, simultaneously being inducted in the potential future of designing biology.

Science and technology studies researchers and responsible innovation scholars have generated a large literature investigating: The dynamic ‘internal differentiation’ of synthetic biology's practices, objectives, values and methods (Balmer, Bulpin, and Molyneux-Hodgson 2016a; Calvert 2008; Calvert 2012; O'Malley et al. 2008; Raimbault, Cointet, and Joly 2016; Raimbault and Joly 2021); the emergence and practices of iGEM (Balmer and Bulpin 2013; Frow and Calvert 2013a); attempts at translating synthetic biology into industrial contexts (Balmer and Molyneux-Hodgson 2013; Mackenzie 2013a; Molyneux-Hodgson and Balmer 2014); the construction of different publics (Frow 2020; Mackenzie 2013b; Marris 2015); social science involvement in synthetic biology (Balmer et al. 2015; Balmer et al. 2016b; Calvert and Martin 2009); analysing and realising the futures and governance of the field (Frow 2013; Frow and Calvert 2013b; Mackenzie 2013c; Ribeiro and Shapira 2019; Schyfter and Calvert 2015; Wiek et al. 2012); and the practices, implications and extent of design in synthetic biology (Calvert 2013; Frow and Calvert 2013a; Kastenhofer 2013; Schyfter, Frow, and Calvert 2013). From cultural practices to heterogeneous engineering, anticipatory governance to infrastructure, analyses of synthetic biology consistently generate new (partial) understandings of knowledge and engineering practice as the field of synthetic biology continues to evolve and mutate. In other words, synthetic biology contains elements similar to those found in other emerging technosciences and well as iterations within these analyses.

Synthetic biologists are concerned with making facilities in terms of both abilities and physical apparatus needed for their practices (Bennett 2015). In addition, in order to work towards the manufacturing promise of the field synthetic biology groups have begun using machine learning, artificial intelligence, and robotics with different implications for scientific work and the DBTL cycle (Hammang 2023; Meckin 2019; Ribeiro et al. 2023). In this context of revolutionary promise and automation, an organisational form has emerged of a ‘biofoundry’:

A biofoundry is an integrated molecular biology facility that includes robotic liquid-handling equipment, high-throughput analytical equipment, and the software, personnel and data management systems required to run the equipment and broader biofoundry capabilities. Biofoundries marry synthetic biology with automation engineering … . (Holowko et al. 2021, 1)

An increasing number of biofoundries are springing up and joining the ‘global biofoundry alliance’ in which various automations are being developed (Dixon, Curach, and Pretorius 2020; Hillson et al. 2019; Holowko et al. 2021). These efforts embed a key promise for synthetic biology: to speed up biology. This means the rapid construction of biological parts or organisms for specified purposes is an important contribution for biofoundries to make. The biofoundry agenda is, perhaps, creating sites of automation, organisational experimentation, and restructuring in synthetic biology that provide a crucial site for synthetic biology to demonstrate its achievements (Zhang et al. 2021), particularly given the high levels of funding they have had. This means the credibility of synthetic biology as trustworthy actors may be on the line (Delgado 2016) as they attempt to demonstrate the value of these investments and deliver on their promises.

Pressure testing

A significant moment of inspiration for this argument came when, after I was invited to join a grant proposal, I looked back through years of data, particularly fieldnotes from ethnographic participation in UK Synthetic Biology Research Centres (SBRC).¹ Some of my data had been developed as one SBRC had conducted a ‘pressure test’ to try and ascertain how it would manufacture prototypes of engineered bioproduction pathways in an idealised scenario. The test aimed to demonstrate synthetic biology approaches could make a series of molecules that could be relevant to materials sciences in a limited amount of time. After the SBRC management team announced the pressure test, one of the experimental scientists exclaimed something to the effect of, ‘it's a grown-up iGEM!’ [the international Genetically Engineering Machine competition] (Fieldnotes 8th Oct 2018). iGEM is a crucial organisation in the field of synthetic biology for capacity building. Although the exact formats, requirements, scope and structure of iGEM have seen changes since its inception in 2004, iGEM has, in principle, remained similar: student teams supervised by groups of more senior researchers create a project that they run for 3 months over the summer and present their project and findings at an autumn conference (Jamboree), which includes various competitions for different prizes.² Thus, in my field notes, I was surprised to find synthetic biologists pointing out self-similar patterns repeating at different scales.

A pressure (or stress) test of biofoundry capabilities was pioneered at The Foundry at MIT. The Foundry test involved DARPA (Defense Advanced Research Projects Agency), which had funded numerous ambitious synthetic biology projects, and would identify the targets for the test (Casini et al. 2018). The Foundry would not have prior knowledge of the targets and would not have (knowingly) worked on them in their own biofoundry or those of their partners. The Foundry would be given a fixed period of time. DARPA's targets could have been anything: from materials to living cells with particular functions. DARPA nominated 10 small molecules with a 3-month deadline for delivery. The time limit meant that all work would be conducted in 3 months.³ The list of small molecules spanned those with pharmaceutical and medical properties, with one industrial chemical. The test began to evaluate the preparedness of the centralised facility to address a rapid response need for a molecule, but focused on the first phases of delivery from target identification to the initial measurement of a product. For the ‘pressure test’ authors, the second stage of scaling up is to optimise strains, the third stage is to pilot production processes, and the fourth stage is to scale up to a full fermentation process (Casini et al. 2018). So, the pressure test is a long way upstream in the synthetic biologists’ projection of innovation.

Casini et al. (2018) refer to other ‘third-party technical evaluations’ in aeronautics, robotics and finance. They justified developing a stress test to assess the progress and capacity of the engineering approach to biology in the following way:

During periods of rapid technology development, it can be challenging to compare disparate approaches, particularly when failures are unpublished and work at some facilities is proprietary. To this end, third-party time-limited assessments can provide valuable perspectives on technology readiness and reveal bottlenecks. (Casini et al. 2018, 4303)

Their references are all about different kinds of evaluations – responding to the robotics challenges in the wake of the Fukushima nuclear accident; developing long-running innovation competitions; and regular stress-testing of financial institutions in anticipation of financial turmoil. Their main biological example is the ‘Critical Assessment of Protein Structure Prediction’ (CASP), where computer labs are given 3 months to predict the structure of a particular protein sequence in order to assess the current methods and help steer technology development in the field.⁴

The references to other competitions and challenges have similarities to the pressure test in synthetic biology but, as mentioned above, synthetic biology already has a self-similar structural pattern – the iGEM competition. Evaluation in iGEM is currently done through a range of adjudicated prizes, with teams submitting various advances for biological ‘parts’, but teams can also opt-in for prizes including education, entrepreneurship, integrated human practices, measurement, inclusivity, and so on (Judging-Committee-and-iGEM-headquarters 2023, 46). In other words, community appraisal, assessment and evaluation through a challenge-led format is built into synthetic biology and many researchers are inducted into these forms, either through supervision of iGEM teams or as student team members.

I have been involved in iGEM, supervising teams in 2014, the 10th anniversary year, and in 2017, and was struck by the scale, effort and industriousness of teams’ projects and presentations as well as the overarching organisation. iGEM, alongside being understood as a social movement for inducting future researchers, is also a way the field can prove its potential. In a previous project,⁵ a senior professor commented:

Several thousand, maybe even ten thousand, students have gone through, most of them have dropped out of synthetic biology but a minority have now done PhDs and some are even starting their own labs. And of course that means that the profile of the field has been bootstrapped from this competition. To the point that now, the BBSRC [Biotechnology and Biological Sciences Research Council] in the UK, synthetic biology is actually a research priority and you can get grant funding for it. (Academic researcher 14 interview, 19th July 2014)

The structure of a time-limited challenge-led project is replicated by hundreds of teams year after year and, by sharing these intensive knowledge-making efforts, is hoped to be a way to both demonstrate and advance synthetic biology, and for the field to pull itself up through working on interdisciplinary projects. Thus, as an organisational form, the competitive test format repeats across hundreds of student teams year on year, as well as the fewer but significantly more highly funded biofoundries, all to advance the technologies and evidence the applied benefits of synthetic biology.

Partial enrolments

The SBRC was set up to be ‘agnostic’ to different chemical requests and so should, in principle, be able to use the same broad methodology to address whatever challenge was proposed. The SBRC was already focused on the production of fine and speciality chemicals (Ribeiro and Shapira 2019). However, during discussions planning and scoping out the pressure test, it was agreed that the test should concern the production of products that would be interesting to materials scientists. What would give the test the greatest credibility, the leadership team felt, would be to enrol materials scientists and have their input in the selection of chemical targets. The concepts of interessment and enrolment have been important in actor-network theory since the 1980s. They are processes by which actors seek ‘to lock in the other actors into roles that have been proposed for them’ and to ‘define and interrelate the various roles they have allocated to others’ (Callon 1984). This could equally be attempts at ‘enmeshment’ in building a network of associations.

The centre leadership approached materials scientists, invited them to meetings, and offered to add their names to the planned pressure test publication. Although a materials scientist representative attended a discussion where possible needs, ideas and amounts were explored, they proved difficult to keep engaged, despite regular follow-ups. After repeated attempts, the centre leadership reported they had received no useful replies to emails and the materials scientists were regarded as no longer being interested enough to keep pursuing. This meant there were open concerns about the credibility of the process for selecting test targets, but the centre continued with a test aimed at manufacturing materially relevant molecules.

I want to go back to a previous project I was involved with in a different institution, with none of the same scientists. This previous experience had a biomedical application at heart: to support improved cell adhesion following skin grafts and stem cell treatments. In the project, synthetic biologists were trying to generate a ‘bioglue’ using E. coli flagella synthesis ‘machinery’ to produce a collagen-binding protein (see Meckin 2016). This project was on different scales to the biofoundries. It was earlier in my career and significantly smaller in terms of funding and resources. The project cost tens of thousands of GBP rather than millions and involved far fewer experimental researchers than the team of ten or more at the SBRC. The reason I find it striking is because there were similar difficulties with enrolling materials scientists which had implications for choosing targets and designing tests.

The project had originally been conceived with materials scientists (tissue engineers) and they were engaged, at least partially, in the project and named on the funding. However, the flagella system had a known sticking point in the collaboration – it was immunogenic. The synthetic biologists saw immunogenicity as the next logical target for their work because they felt they could reengineer the protein to reduce its inflammatory response. Immunogenicity, was something that, even though it was a known issue all along, was a reason a tissue engineer gave for being difficult to contact as the project was coming to an end and only making minimal resources available. The synthetic biologists, with the tissue engineer, ended up designing a brief and somewhat unsatisfactory, inconclusive and biomedically unconvincing assay that showed cells seemed to ‘clump’ with the addition of the bioglue product. Timing was so tight in the end the report of the experiment happened on the last day of the project and the test so cursory it was not possible to convincingly disseminate the findings at the time.

The SBRC's experimental group consisted predominantly of molecular biologists, chemists, and computer scientists in teams, along with technical support for the different teams. There were no materials experts. So the team proceeded by finding different materials precursors in the biotechnological literature. They initially scoped out targets by gathering various lists of potential materials targets or known targets. For instance, one collection of chemical targets was clustered around styrene and its analogues. Styrene is well known for its polymeric applications in (poly)styrene. Another set of compounds was identified from a UK biochemistry list of interesting materials targets. Drawing together these different collections and lists helped narrow down the potential categories of target. These included a range of organic molecules that could be used in polymers, resins and adhesives, as well as volatile compounds used in fragrances and perfumes. The origins of all the chemicals were not always obvious, however, and there were discussions about where particular suggestions had come from. Had chavicol, for instance, come from one of the projects, a list, or been suggested by a member of the management team?

In the SBRC and bioglue cases, the initial interest of and the ongoing retention of materials scientists occurs at different scales, but with similar processes and effects: their uncertain enrolment disrupts the credibility of tests synthetic biologists seek to instantiate. At both scales, the loose enmeshment of materials scientists has knock-on effects on the capacity of synthetic biologists to enrol the material molecules – through identification or assay development – and so the synthetic biologists then need to use more of their own resources to continue their tests, with potential impact on the credibility of the test.

Intersecting scales of value

In discussions, the SBRC team found various biological and chemical reasons for rejecting targets. These included: they were unable to calculate a possible chain of reactions (metabolic pathway) (4 targets); the product, its intermediate, or feedstock was highly toxic to cells (2 targets); there were easier metabolic routes (1 target); there was known competition for the reaction (5 targets); the products or intermediates were too unstable (2 targets). Another reason for potential rejection at this stage was ‘academic novelty’. The contribution to the science of the pressure test was undefined and so the team were trying to orient to a range of potential goals with the various targets. The SBRC team presented to the management team their initial list of targets with a traffic light system showing which targets were possible, which precursors were possible, and other reasons for rejection. 7 targets were in green. There were discussions of potential problems. Styrenes, for example, can stick to plastic so can be difficult to assay. Some other targets were also discussed. Isobutyric acid had not been produced in E. coli and that would make an interesting target – there were two possible pathways. So, a set of molecular targets began emerging.

The team used a software program that could use existing knowledge of enzyme-catalysed reactions to reverse calculate metabolic reactions and thereby generate potential metabolic ‘pathways’ to particular products. If the calculator did not return a pathway, it meant it would not be possible, with current understanding, to synthesise that particular target in organisms. However, although some targets were not possible to make in this way, precursors were identified in their lieu. So, it could be possible to use biomanufacturing to produce a precursor of a final product and use conventional chemistry to create the target molecule. There is a precedent for this approach: before its industrial failure, a poster child for synthetic biology was the production of artemisinin, a useful antimalarial, where the final conversion from the precursor was done using photochemistry (Paddon and Keasling 2014). This meant it could be convincing to get ‘close’ to material molecules.

Throughout all their discussions, a range of criteria became evident: Previous experience of manufacturing (useful), academic novelty (not crucial), economic value (negligible importance), low hanging definite molecules (useful), testability (crucial), ‘route-ability’ (crucial). Through this, there was a subtle shift in the overall aim of the pressure test. In the first week, the management's aim was to demonstrate ‘fast predictive engineering’ for materials using synthetic biology. Indeed, before it started, the initial discussions with materials scientists had covered concerns like economic value, sustainability, and production volume (materials scientists need grams of material to work with, but synthetic biology experiments typically produce volumes a thousand or more times smaller). However, the explicit aim became ‘accessing chemical space’. This stance shifted the outcomes from producing final materials, to producing both material and material-precursor molecules. Perhaps more importantly, it temporarily side-lined industrial concerns of scaling up and focused the pressure test on the potential of synthetic biology to create a range of initial targets.

At one point, two members of the experimental team were leaning over a bench discussing the selection of one particular target. An important aspect was that chemicals should be ‘testable’ meaning they could be analysed in mass spectrometry systems in the SBRC. This involves using a known ‘standard’ and comparing this to a mass spec of the tested target. For one particular chemical, the standard was notably cheap – an impure version could be had at around £10/kg. They looked back at the catalogue prices of the pure versions (for the standard) and found it was substantially more. They then looked back down the pathways at other precursors. The initial feedstock would cost £286/g and after the SBRC's two enzymatic steps, the (ultrapure) product would be worth £86/g. They laughed a lot: two reactions for a threefold reduction in value! Synthetic biology has long wrestled with navigating epistemic and academic merits alongside its industrial and commercial values (Frow 2013). This shows how, as accessing chemical space became the main thrust of the test, industrial and commercial scales of value were affected.

The shape of the pressure test is then tied into purposes, which emerge over time, and the stress test becomes a demonstration of a particular version of manufacturing capabilities. Here, then, academic novelty in regards to individual compounds was less important and the main scientific contribution lay in the overall assessment of the SBRC's capabilities and progress towards automating synthetic biology. This meant targets needed a certain diversity to demonstrate the scope of the biofoundry, and issues like production volume or economic value were less important. The team were given the go-ahead to order the DNA and standards for the 10 targets (which later translated to 17 targets across 8 pathways). In this way, negotiations about the values of synthetic biology, biofoundries and particular targets seem to be self-similar and repeat at differing orders of scale and affect one another.

As with the Foundry pressure test, in the SBRC some molecules were already known to be produced, while some had no known pathway. DARPA and The Foundry had ranked the difficulty of manufacturing from 1 to 10. It is notable, perhaps, that the reporting focuses on numbers: firstly of targets (6/10). The 4 hardest judged by The Foundry were not produced and all those produced had been previously produced through biotechnological methods. Much of the top-line reporting focused on how many things were done or accomplished:

we constructed 1.2 Mb DNA, built 215 strains spanning five species (Saccharomyces cerevisiae, Escherichia coli, Streptomyces albidoflavus, Streptomyces coelicolor, and Streptomyces albovinaceus), established two cell-free systems, and performed 690 assays developed in-house for the molecules. (Casini et al. 2018, 4302)

The Foundry noted design and outsourced DNA synthesis being particularly slow steps (half of the allotted time). Rather than focusing on technical advances, the novelty in The Foundry's paper seems to regard the scale at which synthetic biology can be done and the main learning seems to be the rate at which biology can be engineered and prototyped. This ties value to the promise of accelerating synthetic biology and the importance of time.

At the SBRC's October kick-off meeting the team calculated the test would end in early January. Someone joked, ‘Christmas is cancelled’. A month into the pressure test, the biofoundry and leadership teams discussed the topic: how long are three months? But what constitutes three months? The Foundry ran their test from ‘August 11 to November 11’ but, with the SBRC's test spanning Christmas, would holidays be counted? In fact, the building in which the biofoundry was located was shut down for a week annually during the Christmas holidays so no laboratory work could continue even if the time were included. The team decided not to count this period of shut down meaning Christmas was uncancelled. However, this opened up additional temporal considerations. If Christmas was uncancelled, and it was no longer three calendar months, what about staff holidays and leave? Or, was it 90 days from the start date? If 90 days, were weekends or other non-workdays included in the count? When the team were given the go-ahead to order the DNA and standards, the management team said the clock could stop for about 3 weeks while they waited for the DNA to arrive. Preparatory work continued to ensure the software, robotics, testing and so on were all ready for the arrival of the DNA and the following ‘build’ phase and so some of this time was counted in the end.

The SBRC pressure test ended at the start of February the following year. The working days were back counted to 85. The team wrote up the pressure test and sent it for publication. This was initially rejected from a prestigious synthetic biology journal on grounds of contribution – what new things did it tell synthetic biologists and the wider academic community? I returned several months later to find the team had just completed a second pressure test. This time around a 65-day test focused on strain optimisation, which could be regarded as focusing on Phase two (Casini et al. 2018) in the imagined pathway to prototyping in biomanufacturing. The second optimisation pressure test when combined with the first was enough to secure the publication of a journal article describing the two tests. In other words, the pressure test form needed to be partially replicated to demonstrate added value and novelty for the academic community, with the value being an additional step towards the manufacturing scale.

A possible anticipation

In this paper, I have been so far concerned with developing the idea of self-similarity in synthetic biology. I have identified qualitatively similar patterns in the organisation and structure of challenge-led projects in pressure tests and iGEM; the difficulties in engaging and retaining the interests of materials scientists; and the ways that different scales of value intersect as synthetic biologists negotiate their epistemic and innovation activities. Engineering expectations about aligning synthetic biology with chemical engineering and materials manufacturing are built into the pressure test format. Synthetic biology objects – biobricks and represillators – can swap from symbols of engineering success to infrastructures in the design of new parts (Mackenzie 2013c) and something similar seems true of a time-limited challenge project: it is a symbol of demonstrating capabilities and a repeating infrastructural form of organisation at different scales.

The notion of scale is important to biomanufacturing synthetic biology, in terms of operating at scales greater than a molecular scale, in terms of increasingly large and capable facilities, and in terms of doing more in smaller time scale (Li 2023; Synthetic Biology Roadmap Coordination Group 2012; Zhang et al. 2021). This attention to scales is why fractal reasoning might be relevant and useful in responsible innovation. Alfred Nordmann advocates for an anticipation that ‘is informed by historical experience, and that requires imagination for what might happen in the world as we know it – without anticipating impacts or requiring knowledge of what the future might hold’ (Nordmann 2014, 94). In their recursiveness, fractals concern themselves so we can think of self-similarity in reflexive ways and look for patterns across scales to and identify repeats emerging therein. This means fractality helps identify recursive problems faced over and over in synthetic biology or engineering biology. Identifying these self-similar repeats, which can appear at different orders of scale has the potential to alter how synthetic biology anticipates the future through learning from experience.

We think of history repeating. Stilgoe’s (2018) work on machine learning and self-driving cars is orientated to repetition and he includes a T. S. Eliot quotation about how little generations learn from one another (my lower brow brain always plays the opening lyrics to the Propellerheads featuring Shirley Bassey's ‘History repeating’). Jack Stilgoe is interested in what is not learnt and, following Charles Perrow, how framings of human deficit and operator error as causes of self-driving car crashes means that appropriate social learning cannot happen leaving problems destined to repeat. Rayner (2004), quoted earlier, also puts social learning at the heart of social sciences’ engagements with technological development, as do others in the context of RRI (Egeland, Forsberg, and Maximova-Mentzoni 2019; Smith et al. 2021). To supplement this point about what can be learned in RRI and anticipation, and how it can be learned, the term anticipate has two main meanings. One meaning of the verb is ‘to regard as probable or be aware of the future’. Existing RRI and anticipatory work in synthetic biology has focused on this meaning by tending to develop risk-based and speculative futures. The second meaning is ‘to act as precursor or forerunner’. The idea of self-similarity works in these terms: By identifying forerunners, we may be able to develop a more experientially proximal, reflexive and sociotechnically educational form of anticipation that complements existing futures work by drawing out the organisational, value-laden and interactional patterns that propagate in synthetic biology and potentially other emerging technosciences. This means that thinking of anticipation in terms of ‘what have we learnt from … ?’ would be a useful approach and is where the idea of fractals is intended to contribute.

Ethnographic work can be about finding the possible (Pandian 2019) and it is possible to create a fractal anticipation. Social science is weakly predictive and we know, following Abbott (2001), Abraham (1993), Strathern (2004) and others, that fractality can be a way to analyse data and show self-similar societal forms. After The Foundry's pressure test, the authors lamented how unexciting problems and their resolution took resources and refocused work time:

Across the [pressure test] process, the gaps were largely mundane and practical, and redirected our attention away from “fancier” research areas, such as artificial intelligence, droplet microfluidics, and robotic automation. (Casini et al. 2018, 4312)

Thinking with fractals, the disappointment of not generating more time for other interesting things may disappear. For example, the use of AI and automation in synthetic biology creates and proliferates mundane and practical problems (Ribeiro et al. 2023). Synthetic biologists are already reflexive about their experiences and careers (McLeod, de Saille, and Nerlich 2018). They could anticipate mundane sociotechnical change as part of their scaling-up work – practical issues may well fill the time planned for accelerating synthetic biology. Reflecting on the various observations presented in this article, it is possible that synthetic biologists could anticipate more time-limited attempts to demonstrate capability, more challenges negotiating academic and industrial value, and more issues maintaining the enrolment of other actors. From beyond the data in this paper, but from experience and other commentaries, scaled iterations of challenge-led projects may repeat the difficulties in collaborations, human-non-human interactions, and with imaginaries of publics. A fractal analysis could make it possible to recognise proximal pasts that are relevant to the experiences of synthetic biologists and better prepare for repetition.

Scientific researchers seem intrigued by the notion of fractals when I have mentioned them, and I have begun considering how to operationalise this framework as I think it provides a useful way to collaborate through a shared metaphor. This might be achieved by a collaborative ‘fractal mapping’ exercise that would use a graphic of a fractal, like a Kock snowflake (a repeating pattern of triangles), where researchers, innovators and policymakers use the six points of an initial ‘snowflake’ to identify issues at a particular scale e.g. a pilot project or student iGEM project. They would then develop those points, assuming self-similar patterns and problems, on the points of derivative triangles, anticipating projects involving larger funds and numbers of actors, or scaled into new arenas of development. Mapping such patterns as a project is being conceived, or at the outset of a project, would help prepare collaborators for kinds of patterns they can expect – where these could regard, as in the case developed above, interdisciplinary enrolments, negotiations of value, and organisational forms.

The use of fractals could be seen to be recursive and to simply remake the same patterns of history (for instance, recursiveness in the context of black futures (Ehlers 2023)). My understanding is that identifying the repeats, the self-similarities, facilitates reflexivity that means actors could open up the future by asking how ‘it could be otherwise?’, but not along the lines of radical otherness. Instead, this seems to me to be an opportunity to develop anticipatory tools that seek to identify patterns and extrapolate those to the next iteration in an innovation process. Changing the focus to repetition shifts the governance discourse from ‘solving problems in order to scale-up’, found in the stage-gating discourses (e.g. ‘proof of principle’), to ‘living with self-similar patterns at scales’. This would affect the criteria by which anticipatory learning from pilot projects may be judged. Reports would need to indicate how problems had been solved at a scale and then, rather than assuming its solution would scale unproblematically, anticipate how that problem would recur at a different scale or iteration. For instance, instead of classing mundane issues of automation as a time-consuming distraction, the fractal assumption is that since those problems were experienced, they should be likely re-emerge at a different scale such that they will again take time and resources to attend to. This is because, as I have shown, at a different scale a new range of actors could be introduced, or more actors of similar kinds could be introduced, or different values may play out, all taking time to articulate, embed, test, and resolve. Identifying and preparing for similar complexity at scale (Jensen 2007) is the purpose of a fractal component to anticipation.

I started the paper by highlighting the promissory, future orientation of synthetic biology and the affective effects of recognising repetitions (also noting that competitions and pressure tests put research teams under stress). There can be an element of fatigue and despondency in regard to experiencing similar events and collaborations, and taking similar roles over and over (Balmer et al. 2015). A disbelief can emerge, which can also manifest as surprise at the recurrences of challenges. Despondency and disbelief are affective implications for recognising repetition and anticipatory work should be able to learn from repetition. Deploying the concept of self-similarity practically has the potential to influence how researchers relate to repetition. Rather than dismay or disbelief, we may work collaboratively towards expecting and welcoming repetition. Ultimately, fractal anticipation may help researchers collaboratively identify patterns and better prepare for expecting their re-emergence at different scales.

Acknowledgements

I presented a short version of this paper at the 4S conference in Honolulu (November 2023) and had stimulating discussions with Erika Szymanski, Emma Frow, Jane Calvert, and Robert Smith, among others – thank you all. I would like to thank the two anonymous reviewers who engaged with this submission and whose comments were encouraging, critical, and supported its development. This is undoubtedly a better paper with their feedback and I have since borrowed parts of their approaches in my own peer reviews. I want to thank all the participants and interlocutors who have supported my empirical research and theoretical ideas, including members of the Responsible Research and Innovation group at the University of Manchester. And, thank you to the BBSRC and ESRC for their generous support over the years.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Biotechnology and Biological Sciences Research Council [grant number BB/M017702/1].

Notes on contributors

Robert Meckin

Robert Meckin is a lecturer in the School of Social Sciences, University of Manchester and is affiliated with the National Centre for Research Methods (NCRM) in the UK. His interests include biotechnological science and innovation, interdisciplinarity, everyday life, and sensory and creative social research methods. He recently co-authored Masking in the Pandemic: Materiality, Interaction and Moral Practice (2023), published by Palgrave, and co-edited Investigative Methods (2022), an NCRM collection about digital data and methods in contemporary social research in and outside the academe.

Notes

1 The project was granted ethical approval via the University of Manchester University Research Ethics committee, application no. 16051. Participants were given project information sheets and signed a consent form.

2 https://competition.igem.org/participation/navigating-igem.

3 For publication, additional assays were done to generate errors bars, Casini et al. (2018).

4 https://predictioncenter.org/.

5 Research ethics approval granted the University of Sheffield Research Ethics Committee, documentation available in (Meckin 2016, 313). Participants were given project information sheets and signed a consent form.