Validating what counts as knowledge: The algorithmic gaze and nested multifocality in the biosciences

Validating what counts as knowledge: The algorithmic gaze and nested multifocality in the biosciences | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Validating what counts as knowledge: The algorithmic gaze and nested multifocality in the biosciences Lorenzo Beltrame, Fabio Gasparini, Erik Hernaamt This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7246270/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 12 You are reading this latest preprint version Abstract In this article, we introduce the notions of algorithmic gaze and nested multifocality as analytical categories to investigate the production of evidence in contemporary biosciences. We take a critical stance against certain rhetorics of data-driven science that suggest what counts as evidence is increasingly less the result of experimental procedures and more the outcome of computational methods and pattern recognition algorithms. Drawing on an ethnographic study of a translational medicine project involving the clinic, the biological lab, and bioinformatic work – as well as interviews with bioinformaticians and computational biologists – this paper shows how evidence emerges through the negotiation of different gazes and professional visions. We define the algorithmic gaze as the correlate of a computational style of reasoning, whose output is validated insofar as it incorporates both the clinical and the molecular gaze. The concept of nested multifocality accounts for an epistemic condition in which evidence is not only relative to epistemic cultures and specific research situations, but also emerges through a broader multifocal vision that accommodates different gazes and professional visions. Algorithmic gaze Nested multifocality Validation Bioinformatics Computational biology 1. Introduction In his history of bioinformatics, historian of science Hallam Stevens (2013), reflecting on the profound impact of the computerization of biological work, emphasizes that the biosciences [1] sup> are undergoing a deep transformation in how they produce knowledge. This change has been characterized as a new scientific paradigm, commonly referred to as “data-driven” science. Complex information and communication technologies, digital infrastructures, and computational methods are deployed on large volumes of data to answer biological questions by finding patterns and correlations within these big data sets. This novel data-driven approach has been contrasted with traditional conceptions of science as hypothesis-driven, leading some critics to speak of an “end of theory” (Anderson 2008) and the marginalization of experimental work in knowledge production (Gilbert 1991; Stevens 2013), in favor of an exploratory research that is “not driven by hypothesis and … as model-independent as possible” (Brown and Botstein 1999, 3). Critics warn that such an approach produces “neither knowledge nor understanding” (Allen 2001, 107), since it proceeds through “convenience experimentation” in a mere “gathering mode” (Krohs 2012). Even if “a general characterisation of data-driven methods is hard to achieve” (Leonelli 2012a, 1), theories and rhetoric about the computational turn in the biosciences have been proliferating, fueling a debate about what now counts as evidence in the production of biological knowledge. Stevens (2013), for instance, argues that data-driven biology by investigating “general problems answered by searching big patterns and correlations” (p. 60), entails “new criteria for evaluating knowledge claims, based on statistical, rather than direct experimental, evidence” (p. 63). In particular, he contends that databases make it possible “to investigate biology without doing lab experiments” (p. 148) and, more radically, that “computers have created new ways of making authorized and valuable knowledge through careful accounting and management of data” (p. 102, emphasis added). In this article, we aim to introduce the notions of algorithmic gaze and nested multifocality to show that the search for correlations and hidden patterns in data has not replaced older, traditional “styles of reasoning” (Hacking 1992) in the biosciences. Conversely, we argue that what counts as evidence is the outcome of social-epistemic processes in which different gazes are accommodated to produce a sight deemed objective by the relevant scientific communities. The notion of nested multifocality , in particular, accounts for both the mutual accommodation of gazes on biological entities and phenomena, and the complex assemblages of practices, methods, and styles of reasoning through which bioscientists produce knowledge and decide what counts as evidence. In order to show how pattern recognition — i.e., the algorithmic gaze — has not replaced older, traditional experimental ways of validating claims about phenomena, we will develop an analytical framework that combines the notions of “data-centrism” and the “relational approach” to data developed by Sabina Leonelli (2015; 2016) with the Foucauldian concept of the “clinical gaze,” Hacking’s (1992) “styles of scientific reasoning,” and Goodwin’s (1994) “professional vision.” This analytical framework will be applied to our empirical material, which consists of an ethnographic analysis of a translational oncology research project conducted in Northern Italy, as well as 22 semi-structured interviews with bioinformaticians and computational biologists working in Italy. This project was selected because it involves different gazes, as it is articulated through the bedside, the wet lab (the bench), and the so-called dry lab – that is, the bioinformatic space in which data are processed, interrogated, and visualized. Fieldwork and interviews were conducted by author 2 and author 3, with particular attention to the interactions between the wet and dry labs in validating evidence, and to how the outputs of laboratory-based experimental practices and the computational results of bioinformatic tools are arranged and negotiated with other epistemic communities in the bioscientific field. The paper is organized as follows. Section 2 discusses relevant literature in Science and Technology Studies (STS), as well as in the history and philosophy of science, focusing on the computational turn in contemporary biosciences in order to position our analytical framework. Section 3 further develops this framework by connecting the notion of gazes (Foucault 2003 [1963]) with those of styles of reasoning (Hacking 1992) and professional vision (Goodwin 1994). Section 4 introduces and discusses the concepts of the algorithmic gaze and nested multifocality. After a note on data and methods, the following sections apply our analytical framework to the empirical material. 2. What counts as evidence in data-intensive biosciences Over the last decades, the irruption of bioinformatics, computational methods, the availability of large amounts of data, and the related digital infrastructures has profoundly affected knowledge production in the biosciences. There is no doubt that we are witnessing “a qualitative shift in how scientific research is carried out,” with significant implications for “what counts as scientific knowledge” (Leonelli 2012 b, 47). Today, discoveries can be made by extracting inferences from online datasets, identifying patterns or correlations through data mining, and triangulating evidence across multiple databases. Critics have denounced the automated exploration and analysis of large quantities of data aimed at discovering meaningful patterns as producing a science that is no longer hypothesis-driven (Allen 2001 a; 2001 b). Chris Anderson popularized this criticism in a well-known article published in Wired , titled “The End of Theory” (Anderson 2008 ). Stevens ( 2013 ) cites the concerns of Nobel laureate and DNA sequencing pioneer Walter Gilbert, who envisioned a novel paradigm in biology in which biological entities would no longer be known through traditional laboratory-based experimental procedures, but rather by “being resident in databases available electronically,” containing “enough information to affect the interpretations of almost every sequence” (Gilbert 1991 , 99). Historians and philosophers of science have strongly criticized the rhetoric of the "end of theory," as well as the idea that data ‘speak for themselves’ and the oversimplified binary framing of data-driven versus hypothesis-driven science. Bruno Strasser ( 2012 ) argues that the biological sciences have always operated with data in an exploratory manner, yet theoretical and ontological assumptions have consistently underpinned data management, analysis, and interpretation. Philosopher Sabina Leonelli ( 2012 b; 2016) has shown how data are organized in databases according to what she refers to as “classificatory theories.” More precisely, the system of bio-ontologies used by database curators to organize data constitutes a formal representation of entities that serves as “a form of scientific theorizing that has the potential to affect the direction and practice of experimental biology” and “to gather and express consensus on what constitutes established knowledge" (Leonelli 2016 , 121–122). Similarly, in creating metadata for datasets, curators are involved in translating embodied experimental knowledge and information about the experimental conditions under which the data were produced. Metadata, in her view, demonstrate how such experimental knowledge is used to “evaluate the potential meaning of data” and to determine “the value of data as evidence” (Leonelli 2016 , 30). Her conclusion, therefore, is: While there is no doubt that research grounded on database mining is playing an increasingly important role in complementing and supporting experimental work, the interplay between these two approaches remains crucial to obtaining valid and significant knowledge about the natural world… Consequently, I contest the idea that discovery through the analysis of large datasets can ever be fully automated and/or used as a substitute for experimental intervention in vivo" (Leonelli 2016 , 94–95) A different position is taken up by Hallam Stevens ( 2011 ; 2013 ), for whom, while data are “something other than knowledge” (2013, 6), the introduction of computing has changed what counts as “satisfactory or validated solutions in biology” (p. 10). Stevens does not claim that what he calls bioinformatic biology is theory-free. On the contrary, in his historical analysis of the development of genomic databases, Stevens ( 2013 ) elaborates a complex co-evolution of database structures and theoretical assumptions in biology – for example, early “flat-file” databases were congruent with the one gene-one enzyme hypothesis, while later “relational” databases emphasized and reflected the more recent conception of the interconnectedness of biological elements (p. 138). However, Stevens argues that bioinformatic biology is characterized by distinct theoretical aims (namely, general biological questions) and methodological approaches (specifically, the algorithmic search for patterns and correlations), to the extent that it introduces “specific epistemologies, practices, and modalities of knowing that were and are embedded in the transistors and chips of computing machines” (Stevens 2013 , p. 67). Positions congruent with that of Stevens can be found among STS scholars. Adrian Mackenzie ( 2003 ) defines bioinformatics as an enterprise that deprives the body of its organismic character and transforms it into “a somewhat abstract relational entity” (p. 317), amenable to calculation. In general, STS scholars look primarily at the theoretical and epistemic transformations related to the post-genomic turn, in which notions of the gene, as well as of how the genome functions, have changed (Kay 2000 ; Keller 2000 ; Rheinberger and Müller-Wille 2017 ). In particular, it is the post-genomic networked view of biological organisms (see Keller 1995 ; Landecker 2016 ) that has led scholars to focus on the enactment of biological entities through computers’ “queries and calculations resulting from them, rather than being just accessible through them” (van Baren-Nawrocka et al. 2020 , p. 100). Accordingly, knowledge and understanding of the biological would be constructed through statistical correlation, pattern recognition, and “computer vision research,” rather than through the experimental elucidation of causal relationships or processes (ibid., p. 105). We therefore have two views about how evidence is established and objective knowledge is validated and authorized: one in which the computational analysis of data has gained primacy over traditional practices of experimentation (e.g. Stevens 2013 ); and one that contends that computational methods produce results that are not substitutive of experimental validation (e.g. Leonelli 2016 ). This friction can be investigated by examining the various organizational arrangements that involve relationships, negotiations, and epistemic conflicts among different professional communities (Bourret et al. 2021 ). In purely sociological terms, this means mobilizing the analytical framework on professional dynamics developed by Andrew Abbott ( 1988 ), or Gieryn’s ( 1983 ; 1999 ) notion of “boundary work,” in which such epistemic frictions are linked to jurisdictional disputes and claims. For example, Lewis and Bartlett ( 2013 , p. 247) draw on Gieryn in discussing how bioinformatics engages in disputes with experimental biology in order to characterize itself as a discipline rather than merely a research service. Here, the cultural power derived from being “the legitimate interpreters of the biological world” is at stake (Bartlett et al. 2016 , p. 202) in determining who holds the epistemic authority to define what counts as evidence. This sociological approach can be incorporated into an epistemological reflection that allows situating jurisdictional claims within epistemic practices. To this end, we draw on the notion of data-centrism and the relational framework developed by Sabina Leonelli ( 2015 ; 2016 ). First, the notion of data-centrism highlights the relevance of “data handling and dissemination practices” (Leonelli 2016 , 1), and the growing awareness that data generation must be carried out in ways that allow data to travel beyond the boundaries of local investigation. This awareness is an essential feature of knowledge production that, according to Leonelli ( 2015 , 816), is “built into” experimental and epistemic practices. Data-centrism, therefore, recognizes the relevance of data practices as genuine research activities (Tempini and Leonelli 2018 ; Tempini 2021 ), without dismissing the importance of theorization and experimental work in the lab, and thus rejects some exaggerated claims about data-driven science and the rhetoric of the end of theory. Second, the relational approach to data is consistent with an epistemology that emphasizes the study of the practices and instruments through which research is carried out, with particular attention to the institutional and social dimensions involved (Leonelli 2016 , 69). Considering data as relational categories means that they “do not have a fixed scientific value in and of themselves” (Leonelli 2016 , 70), but rather they are “defined in terms of their function within specific processes of inquiry” (Leonelli 2015 , 818). Consequently, the evidential value attributed to the data also depends on “the range of claims for which data can be considered as evidence" (Leonelli 2016 , 70). So, data do not ‘speak for themselves’, but the evaluation of evidential value is left to researchers and influenced by the “epistemic cultures” (Knorr-Cetina 1999) of the different research communities involved. According to Knorr-Cetina (1999, 1), epistemic cultures are “amalgams of arrangements and mechanisms” which “make up how we know what we know .” Accordingly, the decision about evidential value is shaped within an epistemic culture, but also influenced by the more fluid and pragmatic research situation (Leonelli 2016 , 184), where local interests and the shifting goals of the enquirers are at play. However, any given research situation is always embedded in historical trajectories “in which specific ways of reasoning and knowing have been cultivated and established” (ibid., 185). Data centrism and the relational approach to data allow us to investigate computational methods and “data practices” (Leonelli and Tempini 2020 ) within specific research settings, exploring the role played by pattern recognition algorithms not as overarching dominants but in their interplay with other, more traditional experimental procedures and forms of theorization. 3. Between styles and gazes: pattern recognition algorithms In this section, we aim to conceptualize computational methods based on pattern recognition algorithms as one of what Hacking (1992) calls “styles of scientific reasoning,” a style that involves a particular gaze on biological phenomena. In this sense, we argue that computational methods have introduced a novel “algorithmic gaze” that joins the “clinical gaze” (Foucault 2003) and the “molecular gaze” (Rose 2006) in the lineage of scientific ways of seeing. Using the concept of professional visions developed by Charles Goodwin (1994), we will show how the notion of styles of reasoning can be associated with the Foucauldian notion of the gaze. Hacking borrowed the notion of styles of scientific reasoning from the work of the historian of science Alistair Crombie, who identified six methodological approaches in the European scientific tradition; they are: The simple method of postulation exemplified by the Greek mathematical sciences. The deployment of experiment both to control postulation and to explore by observation and measurement. Hypothetical construction of analogical models. Ordering of variety by comparison and taxonomy. Statistical analysis of regularities of populations, and the calculus of probabilities. The historical derivation of genetic development. (Hacking 1992, 4) These styles are neither specific methods, nor theoretical orientations. They are, according to Hacking, “what we need to understand what we call objectivity,” not because they are themselves objective, but because “they have settled what it is to be objective (truths of certain sorts are just what we obtain by conducting certain sorts of investigations, answering to certain standards)” (Hacking 1992, 4). According to Leonelli (2016, 177), it is impossible to equate data-centrism with a single style of reasoning, since several styles are implicated. Conversely, we do not see this as a problem. Hacking indeed explains that Crombie did not intend the list to be exhaustive or mutually exclusive. Accordingly, Hacking (1992, 5-6) argues that scientific inquiry can employ several styles, styles may evolve, new styles may emerge, and/or two or more styles may merge into an emerging one. This is the case with what Hacking called the “laboratory style,” which uses experimentation (b) “to produce phenomena” to test hypothetical models (c). In the case of pattern recognition through algorithms, we suggest that it could be considered an emergent style that constructs models by performing statistical analysis on available observations. It could also be considered an evolution of the “statistical analysis of regularities of populations, and the calculus of probabilities,” but using a form of statistics different from the classical inferential one (Stevens 2013, 69–70). Keating and Cambrosio (2012) discussing microarray data analysis, have emphasized the hybridization of exploratory techniques (based on algorithms for cluster analysis) with classical statistical hypothesis testing. Therefore, we can consider computational methods like pattern recognition as a novel style of reasoning, which does not exert a hegemonic monopoly over other experimental and statistical analysis techniques, but which is instead interpolated with other practices and styles of reasoning to generate “a standard or model of what it is to be reasonable about this or that type of subject matter” (Hacking 1992, 10). In data-centric biosciences, the relational nature of what counts as evidence implies that algorithm-based pattern recognition constitutes a style of reasoning that should be situated within the network of other styles through which, in specific research endeavors, “objectivity comes into being” (Hacking 1992, 10). Associating styles of reasoning with the Foucauldian notion of gaze can be theoretically fraught. While Hacking sees styles as concerning what it is possible to say and as introducing “new types of objects, evidence, sentences, [and] new ways of being a candidate for truth or falsehood” (p. 11), for Foucault this productive function cannot be located at the level of individual methodological procedures. For him, objects are constituted by the series of rules that make discourse possible, but he excludes “the constitution of a unique horizon of objectivity” (Foucault 1998, 313). According to Foucault, knowledge gains its positivity through the ordering of a set of unfolded enunciations—“which are far from … having the same exigencies of proof … and from having the same operational function” (p. 315). Rather than equating styles of reasoning with Foucault’s episteme or discursive formations, we suggest that what counts as evidence is relative to historically specific articulations of styles of reasoning, which establish variable criteria of objectivity. What characterizes medicine, according to Foucault is the reorganization of the structure of seeing, which is “at once perceptual and epistemological” (Foucault 2003, 165). The clinical gaze is a mode of seeing that “prescribes its norm and epistemological structure” (p. 122) and its access to the body is premised on “a recasting at the level of epistemic knowledge ( savoir ) itself” (p. 137). It is not the simple sensorial act of perceiving, but an epistemic operation that: bears jointly on the type of objects to be known, on the grid that makes it appear, isolates it, and carves up the elements relevant to a possible epistemic knowledge ( savoir ), on the position that the subject must occupy in order to map them, on the instrumental mediations that enable it to grasp them, on the modalities of registration and memory that it must put into operation, and on the forms of conceptualization that it must practice and that qualify it as a subject of legitimate knowledge (Foucault 2003, 137). The clinical gaze is productive: it provides the medical discourse with the visible, as it “took up once again the structures of visibility that it had itself deposited in its field of perception” (p. 117). The fact that visualization is an epistemic act of conceptualization and object construction is consistent with classical reflections in Science and Technology Studies (STS). Bruno Latour emphasized the importance of image production, defining the scientific revolution as a rationalization “of the sight” (Latour 1990, 27). Similarly, Lynch (1988) argued that “reasoning and vision are intimately associated from the beginning” (p. 216) in producing images that are “eidetic,” that is, icons of “the theoretical domain of pure structure and universal laws” (p. 210). Visualization techniques are thus described as an “externalized retina” that constitutes “the sensible, palpable, tangible, and appreciable properties of data” (Lynch 1985, 59). In this way, objects are made “docile,” scientifically knowable by assuming a material form that is “sensible, analyzable, measurable, examinable, manipulable, and ‘intelligible’” (ibid., 43). A useful notion is that of Goodwin’s professional visions , defined as socially organized ways of seeing and understanding events, shaped by the discursive practices of professional groups and their “theories, artifacts and bodies of expertise” (Goodwin 1994, 606). For Goodwin, objects of knowledge emerge within a domain of scrutiny through three practices: (1) coding – which “transforms phenomena observed in a specific setting into the object of knowledge” specific to a given professional discourse; (2) highlighting – which “makes specific phenomena in a complex perceptual field salient by marking them in some fashion”; and (3) the production and articulation of material representations (ibid.). A relevant object of knowledge is thus an event being seen, but through the interplay between a professional domain of scrutiny and the practices of visualization (coding, highlighting, representation), which respond to “a structure of intentionality” within the organizational system of a professional group and are mediated through specific technical artifacts (ibid., p. 609). The fact that visions are perspectival and “lodged” within professional communities means that the power “to authoritatively see,” to produce phenomena, and to “constitute and articulate alternative kinds of events” becomes a potential terrain of contestation among professions. On the one hand, the framework developed by Goodwin allows us to conceive styles of reasoning as productive of peculiar visions and gazes. On the other hand, it suggests that the sociological analysis of jurisdictional disputes among different professional gazes and styles of reasoning can be incorporated into epistemological reflection on the production of evidential value. Now the question is about what kind of peculiar, emergent gaze computational methods like pattern recognition have introduced in the lineage of medical gazes. Indeed, alongside the clinical gaze, contemporary biomedicine has developed a novel gaze that Nikolas Rose (2006, 12) calls the molecular gaze. Biomedicine, he argues, reasons in terms of functional properties, molecular mechanisms of regulation, expression, transcription, and “mechanical and biological properties” (ibid.). Accordingly, the molecular gaze entails understanding and acting upon life at the molecular level. According to Rose, the clinical gaze “has been supplemented, if not supplanted, by this molecular gaze” (ibid.). The very question is whether the irruption of computational methods, and pattern recognition in particular, has pushed specific biological entities into the background, just as statistical computer analysis has overshadowed the study of gene function (Stevens 2013, 66), thereby producing a new gaze that supplants the clinical and molecular ones. We disagree with the idea of supplantation and instead propose a more nuanced understanding in terms of supplementation . 4. The algorithmic gaze and nested multifocality Scholars agree that data visualization is not simply the end result of data analysis; it is constitutive of biological objects and their understanding. Visualization generates “new and often unexpected relationships between biological objects” (Stevens 2013 , 171) by revealing “patterns that would not be spotted unless data are adequately displayed” (Leonelli 2016 , 88). Accordingly, visualization is considered what transforms data into knowledge and is therefore epistemically relevant to the evidential value of data. While the identification of patterns through data visualization techniques has a long history (Müller-Wille and Charmantier 2012 ), scholars recognize that patterns generated by visualization tools are fundamental for data interpretation and dissemination (Bechtel 2020 ), for transforming datasets into targets for investigation (Griesemer 2020 ; Leonelli 2020 ), and, more generally, for organizing knowledge (Burgio and Raffaetà 2024 ). The computational turn connected with post-, epi-, and meta-genomics has led to a informational way of thinking in which biological relationships are conceived as complex informational computer network (Landecker 2016 , 91; Raffaetà 2022 ; Fasel and Chiapperino 2023 ; Chiapperino 2024 ). Burgio and Raffaetà ( 2024 , 823) draw attention to the encounter between two visual and epistemic cultures (i.e., evolutionary biology and informatics) that coalesced in the process of digitalization, which, through algorithms, enables the detection of biological entities. The metagenomic approach applied in the study of the microbiome (Raffaetà 2022 ), for instance, tends to view the human body and its functioning as networked, understandable by digitally reconstructed and visualized through algorithms as a novel “ homo-algorithmicus ” (Kotliar and Grosglik 2023 , 94). Post-genomic approaches in general embed biological entities within a “digital logic of networks and patterns” shaped by computer-mediated observation, in which digital logic is entrenched in practices of observation, as it “underlies the theoretical assumptions on which the use of these observation technologies is based” (van Baren-Nawrocka et al. 2020 , 107). As the authors note, “equating images” with biological entities is predicated upon “a digitalised image on which calculations are possible” (ibid.). We borrow the notion of the algorithmic gaze to refer to a new and emerging epistemological mode of seeing and conceiving biological phenomena, that claims to assert its objectivity. The expression algorithmic gaze was first introduced by Graham ( 2010 ) within a body of literature discussing new forms of power and subjectivation in what is called surveillance or platform capitalism (Zuboff 2019 ). Within this body of literature, user profiling algorithms are considered “a dominant means of organizing – and governing – people’s action,” as individuals’ choices are increasingly “mediated, restricted, and afforded by algorithmic systems” (Kotliar 2020 , 920). Kotliar ( 2020 ), reflecting on how tech companies use analytical algorithms, argues that these tools “increasingly affect how we come to see the world” (p. 921, emphasis added). He further notes that “the ‘algorithmic gaze’ … constitutes the ways in which such companies design, construct, and tweak their algorithms to better ‘see’, conceptualize, and influence people” (ibid., footnote 1). Similarly, pattern recognition algorithms in the biosciences define how to see what to see , thereby constructing objects and disciplining the ways in which they can be known. Secondly, Kotliar’s notion of “data colonialism” proves especially insightful. In particular, Kotliar’s ( 2020 , 922) claim that “data colonialism simultaneously seeks new territories to set its algorithmic eyes on,” can be equated to the movement through which the algorithmic gaze of data-intensive biosciences aims to colonize other epistemic gazes, imposing its own mode of interpreting biological reality as hegemonic. In fact, Ricaurte ( 2019 , 381) defined data colonialism as an imposition of ways of thinking that “denies the existence of alternative worlds and epistemologies.” Of course, we do not take this colonization as a given; rather, we aim to problematize the relationships among different gazes. Kotliar himself offers conceptual tools for undertaking this avenue. First, he argues that, while the colonial logic of expansion renders things knowable, the algorithmic gaze does not necessarily rely on stable and bounded categorizations. Instead, it tends to favor “far from completed” and “allegedly more fine-grained” categories (Kotliar 2020 , 928). The algorithmic gaze does not displace existing classifications but rather operates upon them – as demonstrated by Leonelli’s work on bio-ontologies (2012b; 2016). At the same time, it opens up possibilities for exploring correlations and patterns that rigid categorizations would be unable to capture. Second, Kotliar ( 2020 , 934) characterizes the algorithmic gaze as inherently multifocal: “a gaze that stems from a complex combination of very different types of lenses.” This multifocality provides a theoretical advantage: it allows us to argue that the algorithmic gaze has not supplanted the molecular gaze or the earlier clinical gaze. Rather than adopting metaphors such as “hybridization” (Keating and Cambrosio 2012 ) or “integration” (O’Malley and Soyer 2012 ), we propose the concept of nested multifocality . The algorithmic gaze is multifocal in that it draws on various lenses (moving across theories, epistemologies, and epistemic cultures), and, at the same time, it is embedded within a broader, multifocal vision that must accommodate both the medical and molecular gazes. Nested multifocality means that all these gazes operate together on multifocal platforms, where the vision must be trained to shift across different focal points in order to achieve a knowing sight on biological reality. Only when the sight align itself with these varying focal layers can it support claims to objectivity. What counts as evidence, then, is negotiated among different styles of reasoning (clinical, experimental, algorithmic) – a negotiation that unfolds through the nested multifocality of the algorithmic gaze and its predecessors. In what follows, we empirically examine the algorithmic gaze and its nested multifocality within a translational oncology project that employs a bioinformatic platform. 5. Methodological note In order to explore the dynamics of nested multifocality among gazes (clinical, molecular, and algorithmic), we draw primarily on the ethnographic research conducted by author two within the SMAP project (acronym invented). SMAP is a precision oncology project conducted in Northern Italy aimed at ascertaining the role of genetic diversity and genomic aberrations in the evolution and progression of prostate cancer. One of the objectives of SMAP is to develop a clinically viable test for liquid biopsy, a technique that facilitates the collection of molecular elements contained in the blood, in order to expand its clinical use in molecular stratification, treatment selection, and monitoring of treatment resistance. The SMAP project is, therefore, structured around a network involving close collaboration between the research group and four public hospitals in Northern Italy. The hospitals employ clinical oncologists who are responsible for enrolling patients in the study and collecting blood samples (the bedside and the clinical gaze). Patient data and clinical records are entered into a database by a data manager. Blood samples are processed and sequenced in a designated facility by laboratory biologists (the bench/wet lab and the molecular gaze). Sequence data are processed and analyzed, together with patient data and data from public repositories, by bioinformaticians to identify biomarkers (the dry lab and the algorithmic gaze). It is, therefore, a project that combines different areas of expertise, in which clinicians, experimental biologists, and bioinformaticians negotiate the validation of results and define a shared view of what counts as evidence. The ethnographic research has been complemented by 22 semi-structured interviews with bioinformaticians and computational biologists working in Italy, conducted by authors two and three. The transcripts were analyzed using Atlas.ti for thematic analysis by author one, who also outlined the theoretical and analytical framework adopted in this paper. The distinction between bioinformaticians and computational biologists is not always straightforward, so we have chosen to rely on the self-identification provided by the interviewees. We use the acronyms JCB and SCB to refer to junior and senior Computational Biologists, and JBI and SBI for junior and senior Bioinformaticians, respectively. In total, we interviewed 15 bioinformaticians (6 senior, 9 junior) and 7 computational biologists (5 senior, 2 junior). Ethical approval was granted by the Ethics Committee at the University of X. Interviewees signed a consent form prior to participating in the interviews. To ensure anonymity and prevent identification, any geographical or institutional references have been removed from the interview transcripts. 6. The persistence of gazes from the bedside to the bench and the computer SMAP is fundamentally a project of translational and precision medicine in cancer genomics and, as such, represents a pivotal site for investigating the complex relationships among the heterogeneous professional visions involved: those of the clinical setting, the biology laboratory, and the bioinformatic computational infrastructure. Translational medicine has a long history and a complex genealogy (Harrington and Hauskeller 2014 ), and attempts within STS to grasp its specificities have generated several analytical concepts. Translation has been placed at the core of contemporary biomedicine’s very essence. The notion of biomedical platforms (Keating and Cambrosio 2000 , 359) has been introduced to capture the articulation of clinical routines and biomedical innovation in ways that connect “the biological or normal” with “the medical or pathological” (see also Cambrosio et al. 2006 ). In this sense, translational medicine has been described as a hybrid domain that configures and reconfigures both “persons and tools” and “biology, genomics, and medicine” (Kohli-Laven et al. 2011 , 488). Within this reconfiguration, the role of bioinformatics positions the dry lab somewhere within the two-way flow “from the bench to the bedside” (e.g., Douglas 2014 ; Levin 2014 ). Whether the focus is placed on the structural conditions, norms, and conventions of knowledge production and treatment (Cambrosio et al. 2009 ; 2018 ), or instead on the situated practices through which clinical staff and biomedical investigators mobilize their expertise “to pursue experimental protocols and fabricate clinically actionable knowledge” (Crabu 2021 , 60), the question of “what counts as ‘precise knowledge’ when practising cancer precision medicine” (ibid., 59) remains relevant. In what follows, we will investigate this issue not by directly examining situated practices, but rather by employing the analytical framework outlined above, by focusing on gazes, professional visions, and how different forms of expertise seek to manage the nested multifocality of the involved gazes. 6.1. The clinical gaze The SMAP project began by providing the participating hospitals with equipment and instruments to perform blood sampling in a standardized manner. This was done in order to enable hospital staff to follow the standard operating procedures (SOPs) defined by biologists, ensuring a flow of comparable samples produced through codified protocols using the same instrumentation. SOPs and the standardization of instruments can be understood as a way of unifying the domain of scrutiny by channeling the practices of coding and highlighting (Goodwin 1994 ) through the materiality of technological artefacts and the regulatory device of protocols. This does not mean that the professional visions of nurses, physicians, and oncologists are completely flattened into the styles of reasoning of molecular biologists and bioinformaticians, where standardization of instrument and protocols is seen as warranting the production of evidential value. Nor does it mean that hospital staff are confined within the boundaries of service tasks, such as enrolling new patients and scheduling appointments for the various blood sampling sessions. On the contrary, clinicians protect their professional domain (Gieryn 1999 , 16) by framing the work of the wet and dry labs as “a fundamental analysis for the patient,” as explained by Maurizio, an oncologist, who emphasized that the samples sent to the lab allow clinical oncologists to “decide on the treatment plan.” Moreover, oncologists do not always incorporate the molecular gaze and the clinical gaze into their practices of highlighting . According to Maurizio, in fact, “there are aspects that aren’t taken into consideration” by these gazes, elements that are instead consubstantial to clinicians’ “socially organized ways of seeing” (Goodwin 1994 , 606) their objects of relevance: the patients. As Maurizio explains: The oncologist chooses based on the situation, based on the patient... The data we receive from the lab (from bioinformaticians) don’t always represent the best strategy. They tell me that the strategy is to use drug X, which causes hand tremors after taking it, perhaps for life. If the patient is a pianist... it’s true that it’s the best strategy, it’s true that it would give him a better chance of survival, but he would no longer be able to play the piano. And from a biological standpoint, this isn’t considered; they say, “This one does what’s most beneficial for him.” (Maurizio, oncologist) This is also confirmed by our interviewees outside the SMAP project. Davide, an SCB working in pharmacogenomics, in describing his cooperation with clinicians, clearly states that “the doctor is the patient,” meaning that “if the patient needs a certain drug, even if it’s not the one you want to study, [clinicians] give the patient the drug the patient needs.” This approach applies to data collection as well: even when protocols call for samples to be taken at specific intervals, clinicians “can’t engineer it, so to speak.” Therefore, the regularity of measurement is often missed (e.g., “patients stopped coming”), and this misalignment from codified protocols (i.e., coding ) is seen by other experts as “not [clinicians’] fault.” As Davide puts it: “I can’t say that the clinicians don’t know how to do their job, but their job is different.” 6.2. The molecular gaze Before the wet lab enters into play, samples are handled by the data manager, who is responsible to control both the flow of biological materials (that must be processed and sequenced) and the flow of bio-information: patient clinical records must be anonymized and entered into a database shared with biologists and bioinformaticians. Processing and sequencing samples are two distinct tasks, requiring different expertise, over which biologists assert their professional jurisdiction (Abbott 1988 ) and draw boundaries with the work of bioinformaticians. Linda, a junior biologist assigned to processing samples for sequencing, explains: When you do molecular biology on nucleic acids, you have to be much more careful with everything you do. Making a mistake can mean something as simple as setting the wrong volume on a pipette, so you have to be very precise and very tidy, otherwise, you’ll have to throw everything away… If you make a mistake in bioinformatics, you can just go back with a button, but not here. (Linda, junior biologists) At the same time, the wet lab intertwines with the bedside, seeking to expand the boundaries of its jurisdictional control (Gieryn 1999 , 16) in order to shape the domain of scrutiny according to the molecular gaze. For instance, the evaluation of cancer severity is no longer delegated solely to computed tomography scans (within the jurisdiction of the clinical gaze), but also to biological assays, through which the oncologist is expected to “decide which path to take” based on molecular features (Carla, junior biologist). In this encounter between different platforms (Keating and Cambrosio 2000 ), the wet lab asserts that its biomedical platform will provide clinicians with “the right choice” and “the optimal strategies for their situation” (Yuri, junior biologist). As noted by Nelson and colleagues ( 2014 , 75), in precision oncology clinical trials shift their nature: from testing machines for a drug to becoming clinical experimental systems grounded in biological hypotheses and molecular studies – namely, “devices for materializing new questions about cancer biology and treatment.” However, the clinical gaze cannot be dismissed or entirely colonized by the molecular one. To understand cancer biology and its progression, molecular biologists need to interface with the clinic. “None of us are medical doctors, and as much as we know biology... medicine is another matter entirely,” says Mirko, a senior biologist. Indeed, an understanding of cancer biology cannot be achieved solely through the experimental style of reasoning, as Marta explains: When we don’t understand a piece of data, medical doctors can provide an explanation. Maybe it doesn’t make sense to us, and then they tell us, “No, look, that patient has this particular medical history, and therefore…,” then it makes sense to us too. (Marta, senior biologist) Crabu ( 2016 ) introduced the notion of technomimicry to make sense of the reframing of clinical care and biomedical research in translational medicine, highlighting how the clinic and the laboratory can maintain distinct institutional identities while reconfiguring themselves to mimic one another. We instead propose the notion of multifocality to capture how professional visions, along with the clinical and molecular gaze, seek to develop a multilayered optic to grasp a composite sight of biological phenomena, in order to ground claims about what counts as evidence. Multifocality, as we will discuss later, represents a novel rationalization “of the sight” (Latour 1990 , 27) within a scientific field where different styles of reasoning are assembled to produce evidential knowledge. 6.3. The dry lab: toward the algorithmic gaze Processed samples are then sent to the sequencing center, where Next-Generation Sequencing (NGS) technicians “take care of getting out from [biological samples] all the computer data that bioinformaticians need” (Martina, NGS technician). After that, the dry lab handles the data to identify biomarkers. This “industrialization” of biology, this “politics of productivity” based on automation and informatics (Stevens 2011 , 240), provides bioinformaticians with an alleged positional advantage, advancing claims of epistemic authority over other professional visions: Many biologists… have no background in quantitative fields like computer science or statistics. However, none of them would deny the importance of bioinformatic analysis in research. … That is, there would be no bioinformatics if there were no data and no biological samples from which to collect that data. Likewise, there would be no wet lab as we know it today if massively parallel DNA sequencing did not exist. Without bioinformatics , and all that concerns the analysis of biological data, biology as we know it today would not exist . (Julien, SBI, emphasis added) However, in accordance with the relational approach to data (Leonelli 2016 ) that we adopt, what counts as data, and its related evidential value, is always determined within specific research situations. This means that bioinformatics does not exert hegemonic dominance over other styles of reasoning. Instead, within the SMAP project, other gazes must be incorporated into the algorithmic gaze of computational procedures in order to gain a clearer view of the biology of cancer under investigation. Salvatore, an SBI working on the SMAP project, explains how working with oncologists allows bioinformaticians to understand “what we’re looking at, what questions we need to ask or answer.” Silvio, a clinical oncologist, argues that “clinical parameters are just numbers” to bioinformaticians, who “don’t know what they mean” and therefore “do things that don’t make sense from a clinical point of view.” Similarly, the molecular gaze must also be brought into the perceptual field (Goodwin 1994 ) when the object of relevance is fashioned: the biomarker identified through the algorithm. Luciano, a JBI working on the SMAP project, states that in “processing data… analyzing them… and in explaining the data you obtain,” you must “think at every moment about the biology behind” the data, always proceeding “in light of their biological relevance.” Otherwise, as Martina explains: … the bioinformatician runs their analyses based on their preferred algorithms… but sometimes it happens that the bioinformatician, not knowing how the whole upstream part of the experiment was done on the wet lab side, comes here and asks for a meeting to try to make sense of things that don’t quite add up. (Martina, NGS technician) This once again brings the theme of multifocality into focus, along with its role in warranting evidence. However, we now need to explore more closely how the algorithmic gaze is constituted. 6.4. The algorithmic gaze Davide is a senior researcher, who defines himself a computational biologists and declares to have “an algorithmic background” and therefore he tends to see “every… biological problem” from what he calls an “algorithmic lens [2] .” Asked to specify the point he explains: "Algorithmic thinking,” in my opinion... uh... it’s like, you reason in terms of properties. You have an input, and you want a certain result at the end, but you need to figure out what happens in the middle, like, what kind of properties this “middle” needs to have. And usually, in the middle, there’s some kind of representation, I use graphs, basically, like, there are people who use strings... If you don’t have algorithmic experience, you only think in terms of start and end, but you can’t really figure out what should be in between. If you do have experience, then you can say something like: “Okay, this problem can be tackled as a graph theory problem” … I map everything into some n-dimensional space, then I look where the little point clouds end up, I do stuff. But that feeling , of what’s supposed to be in the middle, only comes from algorithmic experience. If you’re a biologist, you just don’t think that a biological problem might actually be a graph problem …So “algorithmic thinking” means understanding what could be in between the problem and the solution, and who can actually work on that. Like, if it’s a graph, we know how to work on graphs; if it’s strings, we have algorithms for strings; if it’s geometry, we’ve got geometric algorithms, etc. And once you see that, it kind of opens up a whole world, which might be a world you’re familiar with, more or less, but that the biologist maybe has no idea even exists (Davide, SCB) In this long excerpt, Davide outlines some of the main features of what the algorithmic gaze is. First, how this gaze translates biological mechanisms into informational ones. Chiara (SBI) says that what a bioinformatician does is to “translate” biology “into an interpretation closer to how an information engineer would think.” For her, cellular mechanisms are signals and pathways (see Landecker 2016 ), which can be reconstructed and understood by “reasoning in terms of data.” Second, it highlights the theme of complexity. Bioinformaticians and computational biologists conceive their work as essential to dealing with the vast amount of data – not simply in terms of managing a ‘data deluge’, but in enabling an understanding of biological phenomena that would be impossible without an algorithmic way of reasoning. They model “a systemic view of complex biological systems by involving integrative computational methods,” which provides “a tool to test the hypothesis, to understand whether you’re really grasping how the process works” (Flavio, SCB, emphasis added). They are developing computational methods “to make sense” (Emanuele, SCB) of the biological complexity that can only be addressed by “someone who works in information theory,” because “if you apply information theory to a DNA string, and extract some notion of how chaotic or not chaotic that language is, then you can get a measure of this complexity” (Riccardo, SBI). Algorithms step in when biological complexity exceeds the limits of human perception. This is the case with n-dimensionality, as mentioned by Davide, which, according to Chiara: Then there’s the reasoning part of the algorithm, precisely because I’m dealing with complex data, complex patterns, complex relationships... I need to abstract a kind of reasoning that I’d normally use with just a few variables, in low-dimensional spaces… But when I’m talking about machine learning, identifying patterns, I’m already on another level… I can extend to higher-dimensional spaces, and I do that through an algorithm. (Chiara, SBI) Complexity is closely tied to the third theme: the difference between experimental biologists and bioinformaticians in the tools they use to explore biological phenomena. Bioinformaticians draw a boundary between the experimental and the algorithmic that aligns with the point made by Stevens ( 2013 , 45), who argues that bioinformatics poses and answers “general questions” such as “what are the overall rules or patterns for how genomes or diseases behave?” (Stevens 2011 , 236). Alberto (JBI), for instance, points out that “at the wet-lab level you can’t study every single element of [the DNA] sequence in detail,” whereas “computational analysis and bioinformatics allow you to explore this world in depth.” Similarly, Riccardo (SBI) notes that geneticists using “classical methods” face serious difficulties when dealing with complex diseases, while algorithms can detect “what happens with 10, 20, or even 100 genes at once.” By stressing that bioinformaticians have tools that enable them “to ask questions that biologists and clinicians cannot ask themselves” (Riccardo, SBI), they claim professional jurisdiction (Abbott 1988 ) and reject the idea that their work is merely a service (Lewis and Bartlett 2013 ). Is the algorithmic gaze therefore establishing its sovereignty over other biomedical gazes, as “the [algorithmic] eye that knows and decides, the [algorithmic] eye that governs” (Foucault 2003 , 89)? Is it supplanting the clinical and molecular gazes? Is data colonialism seeking “new territories to set its algorithmic eyes on” (Kotliar 2020 , 922) and denying “the existence of alternative worlds and epistemologies” (Ricaurte 2019 , 381)? Rather, our definition of the algorithmic gaze suggests that, even if it does not rely on stable and bounded categorizations, it does not displace existing ones either (Kotliar 2020 , 928). In this sense, Chiara (SBI) explains that “where there is prior knowledge … of a physiological process,” this must be integrated into the algorithm. Moreover, data-driven algorithms are tested on data collected through wet-lab experiments, designed according to parameters defined by existing biological knowledge – which, in turn, “has allowed us to forbid certain relationships between variables or make them mandatory.” So, while the algorithmic gaze detects “nonlinearities in the data that might escape the biotechnologist or biologist, but do not escape these algorithms” (Tommaso, JBI), the clinical and molecular gazes are not supplanted, because any “candidate pathway … has to be investigated further in the wet lab to validate the result” (Marta, JBI). This leads to the question of validation and, consequently, to what counts as evidence: what is considered authoritative, objective knowledge – even when it is produced through an algorithmic gaze. 7. What counts as evidence: the problem of validation While we agree with Stevens ( 2013 , 67) that the algorithmic gaze involves “specific epistemologies, practices, and modalities of knowing,” this does not suffice to constitute a colonizing style of scientific reasoning that eliminates experimental practices. According to Hacking, a style of reasoning is “a standard or model of what it is to be reasonable about this or that type of subject matter” (Hacking, 1992 , p. 10). The question, then, becomes where the boundaries of this “subject matter” lie. Our interviewees, while defending their work as epistemically robust, do not seek to ‘extend the boundaries’ (Gieryn 1999 ) of their subject matter to encompass all biological and biomedical knowledge. When we asked our interviewees about validation processes, their answers stand in contrast to the tenet made by Stevens that “computers have created new ways of making authorized and valuable knowledge” (Stevens 2013 , 102 emphasis added). They explained how they statistically validate their models and algorithms, how they split samples into training and validation sets, or how they use datasets retrieved from public databases to check whether the algorithm can generalize correctly. However, they consider these “sanity checks” (Stevens 2013 , 64) insufficient to produce authorized evidence. The first point to consider is that evidential value is ascribed to the specific moment of inquiry (Leonelli 2016 , 70), and it is shaped by the epistemic cultures involved—since they “create and warrant knowledge” (Knorr-Cetina 1999). For example, in Genome-Wide Association Studies (GWAS), “validation is purely statistical… there, wet-lab validation doesn’t even exist” (Elisa, SBI). Similarly, in metagenomics, the standard validation “is to replicate the study in another population” (Emanuele, SCB). In contrast, when we examine computational biology and bioinformatics applied to cancer genomics or other clinical investigations, the evidential value ascribed to statistical validation changes dramatically. According to Chiara (SBI), in biomarker identification, algorithm-based models are scarcely reproducible, making them only “great exploratory tools.” However, in order “to verify that the biomarker is actually valid and that I can really use it as such,” the wet lab, with its “more standardized and standardizable techniques,” becomes “essential.” Similarly, Elisa (SBI) states that when “it comes to gene function or biological mechanisms, it’s more important, or anyway fundamental, to have the wet lab component.” Among our interviewees, there is a shared belief that experimentation (whether clinical or laboratory-based) holds an epistemic power that in silico validation does not possess. This belief is sometimes attributed to epistemic cultures; for example, the biomedical field “still expects the lab to be the final step” (Matteo, SBI), and consequently, biomedical journals are “more oriented toward experimental validation” (Davide, SCB). Other times, it is simply seen as part and parcel of the research endeavor: since “biology is very complicated” and one “will never exactly reproduce biology,” the outcomes of any computational model “need to be taken with a pinch of salt” (Antonia, JBI). In fact, “going from saying that something is statistically significant… to saying that it’s valid, there’s a whole world in between” (Tommaso, JBI). This does not mean that our interviewees devalue their algorithmic models; rather, they are aware that the algorithmic style of reasoning must be situated within a complex ecology of what counts as evidence in the biosciences. In this ecology, the algorithmic gaze plays an important role, but it is interwoven with other gazes and styles of reasoning in order to gain a more layered understanding of biological phenomena. The algorithmic gaze is linked to an emergent style of reasoning that adopts specific statistical procedures for pattern detection (Keating and Cambrosio 2012 ), merging them with other, more traditional styles – such as model construction, statistical analysis, and the ordering of variety (see Hacking 1992 ). However, “objectivity comes into being” (Hacking 1992 , 10) not solely in silico, but rather through the interplay of a plurality of styles and a network of theories, practices, tools, and technologies that together constitute epistemic cultures (Knorr-Cetina 1999). 8. Nested multifocality or the meeting of multiple gazes How is the algorithmic gaze multifocal? A gaze is multifocal when it “stems from a complex combination of very different types of lenses” (Kotliar 2020 , 934). The algorithmic gaze is already multifocal when an algorithm explores data to find hidden patterns. Not only because it embeds a system of classification rooted in the construction of databases – the classificatory theorization discussed by Leonelli ( 2012 ) – or because data interpretation is influenced by metadata that, according to Leonelli ( 2016 ), express the embodied knowledge of the experimental craftwork that produced the data. Rather, it is the very construction of algorithms that is carried out from a multifocal perspective. Flavio (SCB), for example, explains that when his team has to develop a novel pipeline, they have “periodical meetings with [biomedical] experts” in order to “understand the details of the data that they are going to share, [and] of the technique that is used to collect the data.” At the same time, computational people also need “feedback from them on the quality of the output that we have to produce or the quality of the results, etc.” Indeed, Davide (SCB), who has solid experience in graph theory and graph analysis, told us that applying algorithmic “tricks” used on other kinds of graphs doesn’t “work at all on biological ones,” requiring “new algorithmic challenges and solutions that [are] more specific to biological graphs.” Using the framework developed by Charles Goodwin ( 1994 ), multifocality does not mean warping the practices of coding and highlighting toward a single, unified biomedical vision, but rather constructing algorithms and tools in such a way as to produce objects of knowledge understandable to, and consistent with, the professional visions of biologists and clinicians. This is well explained by Elisa: In my personal experience, this approach works when the other side [the clinical and the experimental ones] has contributed not to the concrete, implementational development of the code, but to the design phase of the method itself. ... This helps me understand whether the expected outputs align with the experience and knowledge of the clinician, geneticist, or biologist … a geneticist, in the case of genetics, or a molecular biologist if lab experiments are involved, have, in a way, some a priori knowledge of a specific disease that I don’t have. They may also have seen the same problem from many different angles , so it becomes important to understand whether my perspective is consistent or at odds with theirs. … through this exchange and comparison, it becomes possible to refine the method, the algorithm, and consequently the tool itself, making it into something that can actually be used by clinicians. (Elisa, SBI, emphasis added) Algorithms must be free to explore data and detect patterns, but in ways that account for different focal layers. The algorithmic gaze must be multifocal not only to be validated, but also because biomedicine operates within multifocal platforms, where different epistemic cultures and professional visions meet and seek to establish alignment on what counts as evidence. Without multifocal lenses, the algorithmic gaze “may not be able to capture a difference that, instead, the wet lab identified because they were working at a different, downstream level” (Alberto, JBI). Nested multifocality seeks to capture how a multifocal gaze operates across different optical layers, where evidential value emerges from the convergence of multiple gazes. This meeting, however, is challenging, always open to potential contestation and friction, because different gazes are not always easily accommodated. Each gaze is entrenched in a professional vision, with its own consolidated styles of reasoning through which “they have settled what it is to be objective” (Hacking 1992 , 4). The algorithmic gaze must be multifocal in order to prevent different professional visions from clashing in jurisdictional disputes over biological phenomena, rather than constituting a shared – though multilayered – domain of scrutiny. Or, to use Abbott’s terms (1988, 41), to prevent the process of colligation – that is, the act of establishing “rules declaring what kinds of evidence are relevant and irrelevant, valid and invalid”—from turning into an irreconcilable wrangle. Conclusions In this paper, we have explored how what counts as evidence in data-intensive biosciences can be rethought through the conceptual lenses of the algorithmic gaze and nested multifocality. Drawing on the idea of data as relational categories (Leonelli 2016 ), we argued that evidence does not simply emerge from the technical output of computational methods. Rather, the production of evidential value is a situated epistemic achievement, conditioned by the alignment of an emergent computational style of reasoning (Hacking 1992 ) with other, coexisting professional visions (Goodwin 1994 ). In this sense, the algorithmic gaze does not assert epistemic sovereignty over clinical or molecular gazes. Instead, what counts as objective knowledge emerges through the negotiation of multiple styles of reasoning within historically specific epistemic cultures (Knorr-Cetina 1999) and contingent research situations (Leonelli 2016 ). The complexity of contemporary biosciences, marked by large heterogeneous collaboration networks (Vermeulen 2009; Leonelli 2013 ), requires the careful articulation of multiple epistemic perspectives. It is in this space of encounter that we situate the notion of multifocality : a metaphor for the epistemic condition in which evidence is no longer anchored to a singular mode of seeing, but emerges from the dynamic interplay of heterogeneous gazes. The knowing eye , must learn to shift across layered focal planes in order to achieve a comprehensive sight. By exploring a translational medicine project involving the clinic, the wet lab, and the dry lab – and by interviewing bioinformaticians and computational biologists – we have defined a particular epistemology, which we term nested multifocality . This notion captures both the multifocal nature of algorithms, designed to traverse and include multiple socio-epistemically organized perceptual frameworks, and the normative requirement that validated knowledge is what is visible through the various professional visions involved. The algorithmic gaze, in this view, is not an autonomous epistemic force, but a situated way of seeing that gains epistemic legitimacy only when it contributes to a shared – though multilayered – domain of scrutiny. Declarations Author Contribution F.G. conducted the ethnographic fieldwork and 12 interviews under the supervision of L.B. E.H. conducted 10 interviews under the supervision of L.B. L.B. analysed the interviews transcripts and worked on the theoretical and analytical framework.L.B. wrote the main manuscript text (section 1-2-3-4-8)L.B. and F.G. wrote together the remaining sections (5-6-7)All authors reviewed the manuscript References Abbott, A. 1988. The System of Professions . Chicago: University of Chicago Press. Allen, J.F. (2001) Bioinformatics and discovery: induction beckons again. BioEssays 23(1): 104–107. Anderson, C. (2008). The End of Theory: The Data Deluge Makes the Scientific Method Obsolete. Wired Jun 23, 2008 12:00 PM https://www.wired.com/2008/06/pb-theory/ Bartlett, A., J. Lewis, and M.L. Williams (2016) Generations of interdisciplinarity in bioinformatics. New Genetics and Society 35(2): 186–209. Bechtel, W. (2020) Data Journeys Beyond Databases in Systems Biology: Cytoscape and NDEx. In: S. Leonelli and N. Tempini (eds.) Data Journeys in the Sciences . Springer International Publishing, pp. 121-144 Bourret, P., P. Castel, H. Bergeron, and A. Cambrosio (2021) Organizing precision oncology: introduction to the special issue. New Genetics and Society 40(1): 1–6. Brown, P.O., and D. Botstein (1999) Exploring the new world of the genome with DNA microarrays. Nature Genetics 21(Suppl 1): 33–37. Burgio, V., and R. Raffaetà (2024) Organizing Microbial Diversity and Interspecies Relations through Diagrams: Trees, Maps, and the Visual Semiotics of the Living. Biosemiotics 17(3): 817–844. Cambrosio, A., P. Keating, T. Schlich, and G. Weisz (2006) Regulatory objectivity and the generation and management of evidence in medicine. Social Science and Medicine 63(1): 189–199. Cambrosio, A., P. Keating, T. Schlich, and G. Weisz (2009) Biomedical Conventions and Regulatory Objectivity: A Few Introductory Remarks. Social Studies of Science 39(5): 651–664. Cambrosio, A., P. Keating, E. Vignola-Gagné, S. Besle, and P. Bourret (2018) Extending experimentation: oncology’s fading boundary between research and care. New Genetics and Society 37(3): 207–226. Chiapperino, L. (2024) Enacting biosocial complexity: Stress, epigenetic biomarkers and the tools of postgenomics. Social Studies of Science 54 (4): 598-625. Crabu, S. (2016) Translational Biomedicine in Action: Constructing Biomarkers across Laboratory and Benchside. Social Theory & Health 14(3): 312–331. Crabu, S. (2021) Organizing the precision clinic: arranging expertise, knowledge and technologies in cancer precision medicine clinical trials. New Genetics and Society 40(1): 58–72. Douglas, C.M.W. (2014) The Role of Bioinformatics in Facilitating Translational Science and Medicine. Tecnoscienza – Italian Journal of Science & Technology Studies 5(1): 141–163. Fasel, C., and L. Chiapperino (2023) Between the genotype and the phenotype lies the microbiome: symbiosis and the making of ‘postgenomic’ knowledge. History and Philosophy of the Life Sciences 45(4): 43.. Foucault, M. (1998) On the Archaeology of the Sciences: Response to the Epistemology Circle. In: J. D. Faubion (ed.) Aesthetics, Method, And Epistemology. New York: The New Press, pp. 297-333. Foucault, M. (2003) The Birth of the Clinic. An Archaeology of Medical Perception . London: Routledge Gieryn, T. F. (1983) Boundary-Work and the Demarcation of Science from Non-Science: Strains and Interests in Professional Ideologies of Scientists. American Sociological Review 48 (6): 781– 795. Gieryn, T. F. (1999) Cultural Boundaries of Science. Credibility on the Line . Chicago: University of Chicago Press. Gilbert, W. (1991). Towards a Paradigm Shift in Biology. Nature 349 (6305): 99–99. Goodwin, C. (1994) Professional Vision. American Anthropologist 96(3): 606-633. Graham, S. D. (2010) Combat Zones that See: Urban Warfare and US Military Technology. In Observant States. Geopolitics and Visual Culture , ed. F. MacDonald, R. Hughes and K Dodds, 199-223. London: I.B. Tauris. Griesemer, J. (2020)A Data Journey Through Dataset-Centric Population Genomics. In: S. Leonelli and N. Tempini (eds.) Data Journeys in the Sciences . Springer International Publishing, pp. 145-167. Hacking, I. (1992) ‘Style’ for Historians and Philosophers. Studies in History and Philosophy of Science Part A 23 (1): 1–20.. Harrington, J., and C. Hauskeller (2014) Translational Research: An Imperative Shaping the Spaces in Biomedicine. Tecnoscienza – Italian Journal of Science & Technology Studies 5(1): 191–201. Kay, L.E. (2000) Who wrote the book of life? A history of the genetic code . Stanford, CA: Stanford University Press. Keating, P., and A. Cambrosio (2000) Biomedical Platforms. Configurations 8(3): 337–387. Keating, P., and A. Cambrosio (2012) Too many numbers: Microarrays in clinical cancer research. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 37–51. Keller, E.F. (1995) Refiguring life: Metaphors of twentieth-century biology . New York: Columbia University Press. Keller, E. F. (2000) The Century of the Gene . Cambridge, MA: Harvard University Press. Knorr Cetina, K. (1999) Epistemic Cultures. How the Sciences Make Knowledge , Cambridge, MA: Harvard University Press. Kohli-Laven, N., P. Bourret, P. Keating, and A. Cambrosio (2011) Cancer clinical trials in the era of genomic signatures: Biomedical innovation, clinical utility, and regulatory- scientific hybrids. Social Studies of Science 41(4): 487–513. Kotliar, D.M. (2020) Data orientalism: on the algorithmic construction of the non-Western other. Theory and Society 49(5–6): 919–939. Kotliar, D.M., and R. Grosglik (2023) On the Contesting Conceptualisation of the Human Body: Between ‘Homo-Microbis’ and ‘Homo-Algorithmicus.’ Body & Society 29(3): 81–108. Krohs, U. (2012) Convenience experimentation. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 52–57. Landecker, H. (2016) The Social as Signal in the Body of Chromatin. The Sociological Review 64(1_suppl): 79–99. Latour, B. (1990) Drawing Things Together. In: M. Lynch and S. Woolgar (eds.) Representation in Scientific Practice. Cambridge, MA: The MIT Press, pp. 19-68. Leonelli, S. (2012a) Introduction: Making sense of data-driven research in the biological and biomedical sciences. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 1–3. Leonelli, S. (2012) Classificatory Theory in Data-intensive Science: The Case of Open Biomedical Ontologies. International Studies in the Philosophy of Science 26(1): 47–65. Leonelli, S. (2013) Global data for local science: Assessing the scale of data infrastructures in biological and biomedical research. BioSocieties 8(4): 449–465. Leonelli, S. (2015) What Counts as Scientific Data? A Relational Framework. Philosophy of Science 82(5): 810–821. doi:10.1080/02698595.2012.653119. Leonelli, S. (2016) Data-centric Biology. A philosophical study. C hicago: University of Chicago Press Leonelli, S. (2020) Learning from Data Journeys. In Data Journeys in the Sciences , ed S. Leonelli and N. Tempini, 1-24. Springer International Publishing. Leonelli, S. and Tempini N. (eds.) (2020) Data Journeys in the Sciences . Springer International Publishing. Levin, N. (2014) What’s Being Translated in Translational Research? Making and Making Sense of Data between the Laboratory and the Clinic. Tecnoscienza – Italian Journal of Science & Technology Studies 5(1): 91–113. Lewis, J. and A. Bartlett (2013) Inscribing a discipline: tensions in the field of bioinformatics. New Genetics and Society 32(3): 243–263. Lynch, M. (1985) Discipline and the Material Form of Images: An Analysis of Scientific Visibility. Social Studies of Science 15(1): 37–66. Lynch, M. (1988) The externalized retina: Selection and mathematization in the visual documentation of objects in the life sciences. Human Studies 11(2–3): 201–234. Mackenzie, A. (2003) Bringing sequences to life: how bioinformatics corporealizes sequence data. New Genetics and Society 22(3): 315–332. Müller-Wille, S. and I. Charmantier (2012) Natural history and information overload: The case of Linnaeus. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 4–15. Nelson, N. C., P. Keating, A. Cambrosio, A. Aguilar-Mahecha, and M. Basik (2014) Testing devices or experimental systems? Cancer clinical trials take the genomic turn. Social Science & Medicine 111, 74–83. O’Malley, M.A. and O.S. Soyer (2012) The roles of integration in molecular systems biology. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 58–68. Raffaetà, R. (2022) Metagenomic Futures: How Microbiome Research is Reconfiguring Health and What it Means to Be Human . London: Routledge. Rheinberger, H.-J,. and S. Müller-Wille (2017) The Gene. From Genetics to Postgenomics . Chicago: University of Chicago press. Ricaurte, P. (2019). Data epistemologies, the Coloniality of power, and resistance. Television and New Media 20(4): 350–365. https://doi.org/10.1177/1527476419831640. Rose, N. (2006) The Politics of Life Itself : Biomedicine, Power, and Subjectivity in the Twenty-First Century, Princeton, NJ: Princeton University Press . Stevens, H. (2011) On the means of bio-production: Bioinformatics and how to make knowledge in a high-throughput genomics laboratory. BioSocieties 6(2): 217–242.. Stevens H. (2013) Life Out of Sequence: A Data-Driven History of Bioinformatics . Chicago: University of Chicago Press. Strasser, B.J. (2012) Data-driven sciences: From wonder cabinets to electronic databases. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 85–87.. Tempini, N. (2021) Data curation-research: practices of data standardization and exploration in a precision medicine database. New Genetics and Society 40(1): 73–94.. doi:10.1080/14636778.2020.1853513. Tempini, N. & Leonelli, S. (2018) Concealment and discovery: The role of information security in biomedical data re-use. Social Studies of Science 48(5): 663–690. van Baren-Nawrocka, J., Consoli, L. & Zwart, H. (2020) Calculable bodies: Analysing the enactment of bodies in bioinformatics. BioSocieties 15(1): 90–114. Vermeulen, N. (2016) Big Biology. NTM Zeitschrift für Geschichte der Wissenschaften, Technik und Medizin 24(2): 195–223. Zuboff, S. (2019) The age of surveillance capitalism: The fight for a human future at the new frontier of power . New York: Public Affairs. Footnotes We use the term bioscience and derivatives to include biology and biomedicine. The Italian word used by the interviewee was “ottica,” which is polysemous: in a literal sense, it means “optical” or “lens,” but in a more abstract sense, it can also mean “perspective,” “point of view,” or “approach.” We chose to preserve the literal sense in our translation. Tables Table 1: Professional profiles of ethnographic informants No. Pseudonym Role in SMAP project 1 Maurizio Senior oncologist 2 Linda Junior biologist 3 Carla Junior biologist 4 Yuri Junior biologist 5 Mirko Senior biologist 6 Martina Senior NGS technician 7 Adriano Junior bioinformatician 8 Julien Senior bioinformatician 9 Silvio Senior oncologist 10 Nicola Junior biologist 11 Luciano Junior bioinformatician 12 Edoardo Senior data manager Table 2: Professional profiles of interviewed participants No. Pseudonym Role Field and subfield 1 Chiara SBI Bioinformatics, omics data analysis 2 Matteo SBI Bioinformatics, pangenomics 3 Vittorio SBI Bioinformatics, meta-analysis 4 Giulia JBI Bioinformatics, cancer genomics 5 Davide SCB Computational biology, pharmacogenomics 6 Elisa SBI Bioinformatics, omics data analysis 7 Alberto JBI Bioinformatics, omics data analysis 8 Riccardo SBI Bioinformatics, omics data analysis 9 Tommaso JBI Bioinformatics, omics data analysis 10 Amedeo JBI Bioinformatics, cancer genomics 11 Flavio SCB Computational biology, pharmacogenomics 12 Gabriele SCB Computational biology, RNA regulatory networks 13 Stefano SCB Computational biology, cancer genomics 14 Emanuele SCB Computational biology, metagenomics 15 Marta JBI Bioinformatics, RNA networks 16 Francesco JCB Computational biology, transcriptomics 17 Valentina JBI Bioinformatics, cancer genomics 18 Antonia JBI Bioinformatics, pharmacokinetics 19 Serena JBI Bioinformatics, cancer genomics 20 Claudia JCB Computational biology, metagenomics 21 Diego JBI Bioinformatics, epigenetics 22 Renato SBI Bioinformatics, epigenetics Note: Abbreviations used in the table as follows: JBI = junior bioinformatician; SBI = senior bioinformatician; JCB = junior computational biologist; SCB =senior computational biologist Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 18 Dec, 2025 Reviews received at journal 28 Nov, 2025 Reviews received at journal 07 Nov, 2025 Reviewers agreed at journal 04 Nov, 2025 Reviewers agreed at journal 02 Nov, 2025 Reviews received at journal 09 Oct, 2025 Reviewers agreed at journal 08 Oct, 2025 Reviewers agreed at journal 03 Oct, 2025 Reviewers invited by journal 25 Aug, 2025 Editor assigned by journal 11 Aug, 2025 Submission checks completed at journal 11 Aug, 2025 First submitted to journal 29 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7246270","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":507794842,"identity":"2423caf1-6e65-48dd-b5e8-9ea4a20b91f8","order_by":0,"name":"Lorenzo Beltrame","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAklEQVRIiWNgGAWjYBACPhjDgIG5AczgB+IDQMyDSwsbQgtjI1iPJJA8ANTDg0sPphaDAwSsYWPvffi44I8Ngzl7Y/vjipptcsY3cg8e/riDQcYelxae48bGM3jSGCx7DjY2njl229jsRl7CgYNn8DhMIo1NmkfiMIPBjcTGxga224nbbuQYHDjYhlcL+28eA6CW+w+BWv7drt88g7AWNmaeBJAtjI2NjW23EwwkCGnhOcYszXMgjceyJ7FxZmPfbcMZZ94YHDjbJsHDcwC7Fn72NsbPPH9s5MzZDx/42PDttjx/e47xh8o2G3v2BhzWQAGGKyTwqx8Fo2AUjIJRgBcAAKzpWHYPl6m/AAAAAElFTkSuQmCC","orcid":"","institution":"University of Trento","correspondingAuthor":true,"prefix":"","firstName":"Lorenzo","middleName":"","lastName":"Beltrame","suffix":""},{"id":507794843,"identity":"7780b057-a556-4522-af4b-ea8def07d09a","order_by":1,"name":"Fabio Gasparini","email":"","orcid":"","institution":"National Research Council","correspondingAuthor":false,"prefix":"","firstName":"Fabio","middleName":"","lastName":"Gasparini","suffix":""},{"id":507794844,"identity":"5dc73ce8-0d89-4ae4-904f-5feb61244825","order_by":2,"name":"Erik Hernaamt","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Erik","middleName":"","lastName":"Hernaamt","suffix":""}],"badges":[],"createdAt":"2025-07-29 19:08:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7246270/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7246270/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90406468,"identity":"ca9efbb0-2caa-4664-af59-26a998896e72","added_by":"auto","created_at":"2025-09-02 11:12:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":802568,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7246270/v1/039ee85b-d4dd-4a39-ace9-706ccabf4220.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Validating what counts as knowledge: The algorithmic gaze and nested multifocality in the biosciences","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eIn his history of bioinformatics, historian of science Hallam Stevens (2013), reflecting on the profound impact of the computerization of biological work, emphasizes that the biosciences\u003csup\u003e[1]\u003c/sup\u003esup\u0026gt; are undergoing a deep transformation in how they produce knowledge. This change has been characterized as a new scientific paradigm, commonly referred to as “data-driven” science. Complex information and communication technologies, digital infrastructures, and computational methods are deployed on large volumes of data to answer biological questions by finding patterns and correlations within these big data sets. This novel data-driven approach has been contrasted with traditional conceptions of science as hypothesis-driven, leading some critics to speak of an “end of theory” (Anderson 2008) and the marginalization of experimental work in knowledge production (Gilbert 1991; Stevens 2013), in favor of an exploratory research that is “not driven by hypothesis and … as model-independent as possible” (Brown and Botstein 1999, 3). Critics warn that such an approach produces “neither knowledge nor understanding” (Allen 2001, 107), since it proceeds through “convenience experimentation” in a mere “gathering mode” (Krohs 2012).\u003c/p\u003e\n\u003cp\u003eEven if “a general characterisation of data-driven methods is hard to achieve” (Leonelli 2012a, 1), theories and rhetoric about the computational turn in the biosciences have been proliferating, fueling a debate about what now counts as evidence in the production of biological knowledge. Stevens (2013), for instance, argues that data-driven biology by investigating “general problems answered by searching big patterns and correlations” (p. 60), entails “new criteria for evaluating knowledge claims, based on statistical, rather than direct experimental, evidence” (p. 63). In particular, he contends that databases make it possible “to investigate biology without doing lab experiments” (p. 148) and, more radically, that “computers have created new ways of making \u003cem\u003eauthorized and valuable knowledge\u003c/em\u003e through careful accounting and management of data” (p. 102, emphasis added).\u003c/p\u003e\n\u003cp\u003eIn this article, we aim to introduce the notions of \u003cem\u003ealgorithmic gaze\u003c/em\u003e and \u003cem\u003enested multifocality\u003c/em\u003e to show that the search for correlations and hidden patterns in data has not replaced older, traditional “styles of reasoning” (Hacking 1992) in the biosciences. Conversely, we argue that what counts as evidence is the outcome of social-epistemic processes in which different gazes are accommodated to produce a \u003cem\u003esight\u003c/em\u003e deemed objective by the relevant scientific communities. The notion of \u003cem\u003enested multifocality\u003c/em\u003e, in particular, accounts for both the mutual accommodation of gazes on biological entities and phenomena, and the complex assemblages of practices, methods, and styles of reasoning through which bioscientists produce knowledge and decide what counts as evidence.\u003c/p\u003e\n\u003cp\u003eIn order to show how pattern recognition — i.e., the algorithmic gaze — has not replaced older, traditional experimental ways of validating claims about phenomena, we will develop an analytical framework that combines the notions of “data-centrism” and the “relational approach” to data developed by Sabina Leonelli (2015; 2016) with the Foucauldian concept of the “clinical gaze,” Hacking’s (1992) “styles of scientific reasoning,” and Goodwin’s (1994) “professional vision.” This analytical framework will be applied to our empirical material, which consists of an ethnographic analysis of a translational oncology research project conducted in Northern Italy, as well as 22 semi-structured interviews with bioinformaticians and computational biologists working in Italy. This project was selected because it involves different gazes, as it is articulated through the bedside, the wet lab (the bench), and the so-called dry lab – that is, the bioinformatic space in which data are processed, interrogated, and visualized. Fieldwork and interviews were conducted by author 2 and author 3, with particular attention to the interactions between the wet and dry labs in validating evidence, and to how the outputs of laboratory-based experimental practices and the computational results of bioinformatic tools are arranged and negotiated with other epistemic communities in the bioscientific field.\u003c/p\u003e\n\u003cp\u003eThe paper is organized as follows. Section 2 discusses relevant literature in Science and Technology Studies (STS), as well as in the history and philosophy of science, focusing on the computational turn in contemporary biosciences in order to position our analytical framework. Section 3 further develops this framework by connecting the notion of gazes (Foucault 2003 [1963]) with those of styles of reasoning (Hacking 1992) and professional vision (Goodwin 1994). Section 4 introduces and discusses the concepts of the algorithmic gaze and nested multifocality. After a note on data and methods, the following sections apply our analytical framework to the empirical material.\u003c/p\u003e"},{"header":"2. What counts as evidence in data-intensive biosciences","content":"\u003cp\u003eOver the last decades, the irruption of bioinformatics, computational methods, the availability of large amounts of data, and the related digital infrastructures has profoundly affected knowledge production in the biosciences. There is no doubt that we are witnessing \u0026ldquo;a qualitative shift in how scientific research is carried out,\u0026rdquo; with significant implications for \u0026ldquo;what counts as scientific knowledge\u0026rdquo; (Leonelli \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2012\u003c/span\u003eb, 47). Today, discoveries can be made by extracting inferences from online datasets, identifying patterns or correlations through data mining, and triangulating evidence across multiple databases.\u003c/p\u003e\u003cp\u003eCritics have denounced the automated exploration and analysis of large quantities of data aimed at discovering meaningful patterns as producing a science that is no longer hypothesis-driven (Allen \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2001\u003c/span\u003ea; \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2001\u003c/span\u003eb). Chris Anderson popularized this criticism in a well-known article published in \u003cem\u003eWired\u003c/em\u003e, titled \u0026ldquo;The End of Theory\u0026rdquo; (Anderson \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Stevens (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) cites the concerns of Nobel laureate and DNA sequencing pioneer Walter Gilbert, who envisioned a novel paradigm in biology in which biological entities would no longer be known through traditional laboratory-based experimental procedures, but rather by \u0026ldquo;being resident in databases available electronically,\u0026rdquo; containing \u0026ldquo;enough information to affect the interpretations of almost every sequence\u0026rdquo; (Gilbert \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e1991\u003c/span\u003e, 99).\u003c/p\u003e\u003cp\u003eHistorians and philosophers of science have strongly criticized the rhetoric of the \"end of theory,\" as well as the idea that data \u0026lsquo;speak for themselves\u0026rsquo; and the oversimplified binary framing of data-driven versus hypothesis-driven science. Bruno Strasser (\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) argues that the biological sciences have always operated with data in an exploratory manner, yet theoretical and ontological assumptions have consistently underpinned data management, analysis, and interpretation. Philosopher Sabina Leonelli (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2012\u003c/span\u003eb; 2016) has shown how data are organized in databases according to what she refers to as \u0026ldquo;classificatory theories.\u0026rdquo; More precisely, the system of bio-ontologies used by database curators to organize data constitutes a formal representation of entities that serves as \u0026ldquo;a form of scientific theorizing that has the potential to affect the direction and practice of experimental biology\u0026rdquo; and \u0026ldquo;to gather and express consensus on what constitutes established knowledge\" (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 121\u0026ndash;122). Similarly, in creating metadata for datasets, curators are involved in translating embodied experimental knowledge and information about the experimental conditions under which the data were produced. Metadata, in her view, demonstrate how such experimental knowledge is used to \u0026ldquo;evaluate the potential meaning of data\u0026rdquo; and to determine \u0026ldquo;the value of data as evidence\u0026rdquo; (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 30). Her conclusion, therefore, is:\u003c/p\u003e\u003cp\u003eWhile there is no doubt that research grounded on database mining is playing an increasingly important role in complementing and supporting experimental work, the interplay between these two approaches remains crucial to obtaining valid and significant knowledge about the natural world\u0026hellip; Consequently, I contest the idea that discovery through the analysis of large datasets can ever be fully automated and/or used as a substitute for experimental intervention in vivo\" (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 94\u0026ndash;95)\u003c/p\u003e\u003cp\u003eA different position is taken up by Hallam Stevens (\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), for whom, while data are \u0026ldquo;something other than knowledge\u0026rdquo; (2013, 6), the introduction of computing has changed what counts as \u0026ldquo;satisfactory or validated solutions in biology\u0026rdquo; (p. 10). Stevens does not claim that what he calls bioinformatic biology is theory-free. On the contrary, in his historical analysis of the development of genomic databases, Stevens (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) elaborates a complex co-evolution of database structures and theoretical assumptions in biology \u0026ndash; for example, early \u0026ldquo;flat-file\u0026rdquo; databases were congruent with the one gene-one enzyme hypothesis, while later \u0026ldquo;relational\u0026rdquo; databases emphasized and reflected the more recent conception of the interconnectedness of biological elements (p. 138). However, Stevens argues that bioinformatic biology is characterized by distinct theoretical aims (namely, general biological questions) and methodological approaches (specifically, the algorithmic search for patterns and correlations), to the extent that it introduces \u0026ldquo;specific epistemologies, practices, and modalities of knowing that were and are embedded in the transistors and chips of computing machines\u0026rdquo; (Stevens \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, p. 67).\u003c/p\u003e\u003cp\u003ePositions congruent with that of Stevens can be found among STS scholars. Adrian Mackenzie (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) defines bioinformatics as an enterprise that deprives the body of its organismic character and transforms it into \u0026ldquo;a somewhat abstract relational entity\u0026rdquo; (p. 317), amenable to calculation. In general, STS scholars look primarily at the theoretical and epistemic transformations related to the post-genomic turn, in which notions of the gene, as well as of how the genome functions, have changed (Kay \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Keller \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Rheinberger and M\u0026uuml;ller-Wille \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). In particular, it is the post-genomic networked view of biological organisms (see Keller \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1995\u003c/span\u003e; Landecker \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) that has led scholars to focus on the enactment of biological entities through computers\u0026rsquo; \u0026ldquo;queries and calculations resulting from them, rather than being just accessible through them\u0026rdquo; (van Baren-Nawrocka et al. \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, p. 100). Accordingly, knowledge and understanding of the biological would be constructed through statistical correlation, pattern recognition, and \u0026ldquo;computer vision research,\u0026rdquo; rather than through the experimental elucidation of causal relationships or processes (ibid., p. 105).\u003c/p\u003e\u003cp\u003eWe therefore have two views about how evidence is established and objective knowledge is validated and authorized: one in which the computational analysis of data has gained primacy over traditional practices of experimentation (e.g. Stevens \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e); and one that contends that computational methods produce results that are not substitutive of experimental validation (e.g. Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). This friction can be investigated by examining the various organizational arrangements that involve relationships, negotiations, and epistemic conflicts among different professional communities (Bourret et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In purely sociological terms, this means mobilizing the analytical framework on professional dynamics developed by Andrew Abbott (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1988\u003c/span\u003e), or Gieryn\u0026rsquo;s (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1983\u003c/span\u003e; \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1999\u003c/span\u003e) notion of \u0026ldquo;boundary work,\u0026rdquo; in which such epistemic frictions are linked to jurisdictional disputes and claims. For example, Lewis and Bartlett (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, p. 247) draw on Gieryn in discussing how bioinformatics engages in disputes with experimental biology in order to characterize itself as a discipline rather than merely a research service. Here, the cultural power derived from being \u0026ldquo;the legitimate interpreters of the biological world\u0026rdquo; is at stake (Bartlett et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, p. 202) in determining who holds the epistemic authority to define what counts as evidence. This sociological approach can be incorporated into an epistemological reflection that allows situating jurisdictional claims within epistemic practices. To this end, we draw on the notion of data-centrism and the relational framework developed by Sabina Leonelli (\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFirst, the notion of data-centrism highlights the relevance of \u0026ldquo;data handling and dissemination practices\u0026rdquo; (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 1), and the growing awareness that data generation must be carried out in ways that allow data to travel beyond the boundaries of local investigation. This awareness is an essential feature of knowledge production that, according to Leonelli (\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, 816), is \u0026ldquo;built into\u0026rdquo; experimental and epistemic practices. Data-centrism, therefore, recognizes the relevance of data practices as genuine research activities (Tempini and Leonelli \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Tempini \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), without dismissing the importance of theorization and experimental work in the lab, and thus rejects some exaggerated claims about data-driven science and the rhetoric of the end of theory.\u003c/p\u003e\u003cp\u003eSecond, the relational approach to data is consistent with an epistemology that emphasizes the study of the practices and instruments through which research is carried out, with particular attention to the institutional and social dimensions involved (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 69). Considering data as relational categories means that they \u0026ldquo;do not have a fixed scientific value in and of themselves\u0026rdquo; (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 70), but rather they are \u0026ldquo;defined in terms of their function within specific processes of inquiry\u0026rdquo; (Leonelli \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, 818). Consequently, the \u003cem\u003eevidential value\u003c/em\u003e attributed to the data also depends on \u0026ldquo;the range of claims for which data can be considered as evidence\" (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 70).\u003c/p\u003e\u003cp\u003eSo, data do not \u0026lsquo;speak for themselves\u0026rsquo;, but the evaluation of evidential value is left to researchers and influenced by the \u0026ldquo;epistemic cultures\u0026rdquo; (Knorr-Cetina 1999) of the different research communities involved. According to Knorr-Cetina (1999, 1), epistemic cultures are \u0026ldquo;amalgams of arrangements and mechanisms\u0026rdquo; which \u0026ldquo;make up \u003cem\u003ehow we know what we know\u003c/em\u003e.\u0026rdquo; Accordingly, the decision about evidential value is shaped within an epistemic culture, but also influenced by the more fluid and pragmatic research \u003cem\u003esituation\u003c/em\u003e (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 184), where local interests and the shifting goals of the enquirers are at play. However, any given research situation is always embedded in historical trajectories \u0026ldquo;in which specific ways of reasoning and knowing have been cultivated and established\u0026rdquo; (ibid., 185).\u003c/p\u003e\u003cp\u003eData centrism and the relational approach to data allow us to investigate computational methods and \u0026ldquo;data practices\u0026rdquo; (Leonelli and Tempini \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) within specific research settings, exploring the role played by pattern recognition algorithms not as overarching dominants but in their interplay with other, more traditional experimental procedures and forms of theorization.\u003c/p\u003e"},{"header":"3. Between styles and gazes: pattern recognition algorithms","content":"\u003cp\u003eIn this section, we aim to conceptualize computational methods based on pattern recognition algorithms as one of what Hacking (1992) calls \u0026ldquo;styles of scientific reasoning,\u0026rdquo; a style that involves a particular gaze on biological phenomena. In this sense, we argue that computational methods have introduced a novel \u0026ldquo;algorithmic gaze\u0026rdquo; that joins the \u0026ldquo;clinical gaze\u0026rdquo; (Foucault 2003) and the \u0026ldquo;molecular gaze\u0026rdquo; (Rose 2006) in the lineage of scientific ways of seeing. Using the concept of professional visions developed by Charles Goodwin (1994), we will show how the notion of styles of reasoning can be associated with the Foucauldian notion of the gaze.\u003c/p\u003e\n\u003cp\u003eHacking borrowed the notion of styles of scientific reasoning from the work of the historian of science Alistair Crombie, who identified six methodological approaches in the European scientific tradition; they are:\u0026nbsp;\u003c/p\u003e\n\u003col style=\"list-style-type: lower-alpha;\"\u003e\n \u003cli\u003eThe simple method of postulation exemplified by the Greek mathematical sciences.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eThe deployment of experiment both to control postulation and to explore by observation and measurement.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp;Hypothetical construction of analogical models.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp;Ordering of variety by comparison and taxonomy.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp;Statistical analysis of regularities of populations, and the calculus of probabilities.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eThe historical derivation of genetic development. (Hacking 1992, 4)\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThese styles are neither specific methods, nor theoretical orientations. They are, according to Hacking, \u0026ldquo;what we need to understand what we call objectivity,\u0026rdquo; not because they are themselves objective, but because \u0026ldquo;they have settled what it is to be objective (truths of certain sorts are just what we obtain by conducting certain sorts of investigations, answering to certain standards)\u0026rdquo; (Hacking 1992, 4).\u003c/p\u003e\n\u003cp\u003eAccording to Leonelli (2016, 177), it is impossible to equate data-centrism with a single style of reasoning, since several styles are implicated. Conversely, we do not see this as a problem. Hacking indeed explains that Crombie did not intend the list to be exhaustive or mutually exclusive. Accordingly, Hacking (1992, 5-6) argues that scientific inquiry can employ several styles, styles may evolve, new styles may emerge, and/or two or more styles may merge into an emerging one. This is the case with what Hacking called the \u0026ldquo;laboratory style,\u0026rdquo; which uses experimentation (b) \u0026ldquo;to produce phenomena\u0026rdquo; to test hypothetical models (c). In the case of pattern recognition through algorithms, we suggest that it could be considered an emergent style that constructs models by performing statistical analysis on available observations. It could also be considered an evolution of the \u0026ldquo;statistical analysis of regularities of populations, and the calculus of probabilities,\u0026rdquo; but using a form of statistics different from the classical inferential one (Stevens 2013, 69\u0026ndash;70). Keating and Cambrosio (2012) discussing microarray data analysis, have emphasized the hybridization of exploratory techniques (based on algorithms for cluster analysis) with classical statistical hypothesis testing.\u003c/p\u003e\n\u003cp\u003eTherefore, we can consider computational methods like pattern recognition as a novel style of reasoning, which does not exert a hegemonic monopoly over other experimental and statistical analysis techniques, but which is instead interpolated with other practices and styles of reasoning to generate \u0026ldquo;a standard or model of what it is to be reasonable about this or that type of subject matter\u0026rdquo; (Hacking 1992, 10). In data-centric biosciences, the relational nature of what counts as evidence implies that algorithm-based pattern recognition constitutes a style of reasoning that should be situated within the network of other styles through which, in specific research endeavors, \u0026ldquo;objectivity comes into being\u0026rdquo; (Hacking 1992, 10).\u003c/p\u003e\n\u003cp\u003eAssociating styles of reasoning with the Foucauldian notion of gaze can be theoretically fraught. While Hacking sees styles as concerning what it is possible to say and as introducing \u0026ldquo;new types of objects, evidence, sentences, [and] new ways of being a candidate for truth or falsehood\u0026rdquo; (p. 11), for Foucault this productive function cannot be located at the level of individual methodological procedures. For him, objects are constituted by the series of rules that make discourse possible, but he excludes \u0026ldquo;the constitution of a unique horizon of objectivity\u0026rdquo; (Foucault 1998, 313). According to Foucault, knowledge gains its positivity through the ordering of a set of unfolded enunciations\u0026mdash;\u0026ldquo;which are far from \u0026hellip; having the same exigencies of proof \u0026hellip; and from having the same operational function\u0026rdquo; (p. 315). Rather than equating styles of reasoning with Foucault\u0026rsquo;s episteme or discursive formations, we suggest that what counts as evidence is relative to historically specific articulations of styles of reasoning, which establish variable criteria of objectivity.\u003c/p\u003e\n\u003cp\u003eWhat characterizes medicine, according to Foucault is the reorganization of the structure of seeing, which is \u0026ldquo;at once perceptual and epistemological\u0026rdquo; (Foucault 2003, 165). The clinical gaze is a mode of seeing that \u0026ldquo;prescribes its norm and epistemological structure\u0026rdquo; (p. 122) and its access to the body is premised on \u0026ldquo;a recasting at the level of epistemic knowledge (\u003cem\u003esavoir\u003c/em\u003e) itself\u0026rdquo; (p. 137). It is not the simple sensorial act of perceiving, but an epistemic operation that:\u003c/p\u003e\n\u003cp\u003ebears jointly on the type of objects to be known, on the grid that makes it appear, isolates it, and carves up the elements relevant to a possible epistemic knowledge (\u003cem\u003esavoir\u003c/em\u003e), on the position that the subject must occupy in order to map them, on the instrumental mediations that enable it to grasp them, on the modalities of registration and memory that it must put into operation, and on the forms of conceptualization that it must practice and that qualify it as a subject of legitimate knowledge (Foucault 2003, 137).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe clinical gaze is productive: it provides the medical discourse with the visible, as it \u0026ldquo;took up once again the structures of visibility that it had itself deposited in its field of perception\u0026rdquo; (p. 117).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe fact that visualization is an epistemic act of conceptualization and object construction is consistent with classical reflections in Science and Technology Studies (STS). Bruno Latour emphasized the importance of image production, defining the scientific revolution as a rationalization \u0026ldquo;of the sight\u0026rdquo; (Latour 1990, 27). Similarly, Lynch (1988) argued that \u0026ldquo;reasoning and vision are intimately associated from the beginning\u0026rdquo; (p. 216) in producing images that are \u0026ldquo;eidetic,\u0026rdquo; that is, icons of \u0026ldquo;the theoretical domain of pure structure and universal laws\u0026rdquo; (p. 210). Visualization techniques are thus described as an \u0026ldquo;externalized retina\u0026rdquo; that constitutes \u0026ldquo;the sensible, palpable, tangible, and appreciable properties of data\u0026rdquo; (Lynch 1985, 59). In this way, objects are made \u0026ldquo;docile,\u0026rdquo; scientifically knowable by assuming a material form that is \u0026ldquo;sensible, analyzable, measurable, examinable, manipulable, and \u0026lsquo;intelligible\u0026rsquo;\u0026rdquo; (ibid., 43).\u003c/p\u003e\n\u003cp\u003eA useful notion is that of Goodwin\u0026rsquo;s \u003cem\u003eprofessional visions\u003c/em\u003e, defined as socially organized ways of seeing and understanding events, shaped by the discursive practices of professional groups and their \u0026ldquo;theories, artifacts and bodies of expertise\u0026rdquo; (Goodwin 1994, 606). For Goodwin, objects of knowledge emerge within a \u003cem\u003edomain of scrutiny\u003c/em\u003e through three practices: (1) \u003cem\u003ecoding\u003c/em\u003e \u0026ndash; which \u0026ldquo;transforms phenomena observed in a specific setting into the object of knowledge\u0026rdquo; specific to a given professional discourse; (2) \u003cem\u003ehighlighting\u003c/em\u003e \u0026ndash; which \u0026ldquo;makes specific phenomena in a complex perceptual field salient by marking them in some fashion\u0026rdquo;; and (3) \u003cem\u003ethe production and articulation of material representations\u003c/em\u003e (ibid.). A relevant object of knowledge is thus an event being seen, but through the interplay between a professional domain of scrutiny and the practices of visualization (coding, highlighting, representation), which respond to \u0026ldquo;a structure of intentionality\u0026rdquo; within the organizational system of a professional group and are mediated through specific technical artifacts (ibid., p. 609). The fact that visions are perspectival and \u0026ldquo;lodged\u0026rdquo; within professional communities means that the power \u0026ldquo;to authoritatively see,\u0026rdquo; to produce phenomena, and to \u0026ldquo;constitute and articulate alternative kinds of events\u0026rdquo; becomes a potential terrain of contestation among professions. On the one hand, the framework developed by Goodwin allows us to conceive styles of reasoning as productive of peculiar visions and gazes. On the other hand, it suggests that the sociological analysis of jurisdictional disputes among different professional gazes and styles of reasoning can be incorporated into epistemological reflection on the production of evidential value.\u003c/p\u003e\n\u003cp\u003eNow the question is about what kind of peculiar, emergent gaze computational methods like pattern recognition have introduced in the lineage of medical gazes. Indeed, alongside the clinical gaze, contemporary biomedicine has developed a novel gaze that Nikolas Rose (2006, 12) calls the molecular gaze. Biomedicine, he argues, reasons in terms of functional properties, molecular mechanisms of regulation, expression, transcription, and \u0026ldquo;mechanical and biological properties\u0026rdquo; (ibid.). Accordingly, the molecular gaze entails understanding and acting upon life at the molecular level. According to Rose, the clinical gaze \u0026ldquo;has been supplemented, if not supplanted, by this molecular gaze\u0026rdquo; (ibid.). The very question is whether the irruption of computational methods, and pattern recognition in particular, has pushed specific biological entities into the background, just as statistical computer analysis has overshadowed the study of gene function (Stevens 2013, 66), thereby producing a new gaze that supplants the clinical and molecular ones. We disagree with the idea of \u003cem\u003esupplantation\u003c/em\u003e and instead propose a more nuanced understanding in terms of \u003cem\u003esupplementation\u003c/em\u003e.\u003c/p\u003e"},{"header":"4. The algorithmic gaze and nested multifocality","content":"\u003cp\u003eScholars agree that data visualization is not simply the end result of data analysis; it is constitutive of biological objects and their understanding. Visualization generates \u0026ldquo;new and often unexpected relationships between biological objects\u0026rdquo; (Stevens \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, 171) by revealing \u0026ldquo;patterns that would not be spotted unless data are adequately displayed\u0026rdquo; (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 88). Accordingly, visualization is considered what transforms data into knowledge and is therefore epistemically relevant to the evidential value of data. While the identification of patterns through data visualization techniques has a long history (M\u0026uuml;ller-Wille and Charmantier \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), scholars recognize that patterns generated by visualization tools are fundamental for data interpretation and dissemination (Bechtel \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), for transforming datasets into targets for investigation (Griesemer \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Leonelli \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), and, more generally, for organizing knowledge (Burgio and Raffaet\u0026agrave; \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe computational turn connected with post-, epi-, and meta-genomics has led to a informational way of thinking in which biological relationships are conceived as complex informational computer network (Landecker \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 91; Raffaet\u0026agrave; \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Fasel and Chiapperino \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Chiapperino \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Burgio and Raffaet\u0026agrave; (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e, 823) draw attention to the encounter between two visual and epistemic cultures (i.e., evolutionary biology and informatics) that coalesced in the process of digitalization, which, through algorithms, enables the detection of biological entities. The metagenomic approach applied in the study of the microbiome (Raffaet\u0026agrave; \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), for instance, tends to view the human body and its functioning as networked, understandable by digitally reconstructed and visualized through algorithms as a novel \u0026ldquo;\u003cem\u003ehomo-algorithmicus\u003c/em\u003e\u0026rdquo; (Kotliar and Grosglik \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2023\u003c/span\u003e, 94). Post-genomic approaches in general embed biological entities within a \u0026ldquo;digital logic of networks and patterns\u0026rdquo; shaped by computer-mediated observation, in which digital logic is entrenched in practices of observation, as it \u0026ldquo;underlies the theoretical assumptions on which the use of these observation technologies is based\u0026rdquo; (van Baren-Nawrocka et al. \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 107). As the authors note, \u0026ldquo;equating images\u0026rdquo; with biological entities is predicated upon \u0026ldquo;a digitalised image on which calculations are possible\u0026rdquo; (ibid.).\u003c/p\u003e\u003cp\u003eWe borrow the notion of the \u003cem\u003ealgorithmic gaze\u003c/em\u003e to refer to a new and emerging epistemological mode of seeing and conceiving biological phenomena, that claims to assert its objectivity. The expression \u003cem\u003ealgorithmic gaze was first\u003c/em\u003e introduced by Graham (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) within a body of literature discussing new forms of power and subjectivation in what is called surveillance or platform capitalism (Zuboff \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Within this body of literature, user profiling algorithms are considered \u0026ldquo;a dominant means of organizing \u0026ndash; and governing \u0026ndash; people\u0026rsquo;s action,\u0026rdquo; as individuals\u0026rsquo; choices are increasingly \u0026ldquo;mediated, restricted, and afforded by algorithmic systems\u0026rdquo; (Kotliar \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 920). Kotliar (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), reflecting on how tech companies use analytical algorithms, argues that these tools \u0026ldquo;increasingly affect how we come to \u003cem\u003esee\u003c/em\u003e the world\u0026rdquo; (p. 921, emphasis added). He further notes that \u0026ldquo;the \u0026lsquo;algorithmic gaze\u0026rsquo; \u0026hellip; constitutes the ways in which such companies design, construct, and tweak their algorithms to better \u0026lsquo;see\u0026rsquo;, conceptualize, and influence people\u0026rdquo; (ibid., footnote 1). Similarly, pattern recognition algorithms in the biosciences define \u003cem\u003ehow to see what to see\u003c/em\u003e, thereby constructing objects and disciplining the ways in which they can be known.\u003c/p\u003e\u003cp\u003eSecondly, Kotliar\u0026rsquo;s notion of \u0026ldquo;data colonialism\u0026rdquo; proves especially insightful. In particular, Kotliar\u0026rsquo;s (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 922) claim that \u0026ldquo;data colonialism simultaneously seeks new territories to set its algorithmic eyes on,\u0026rdquo; can be equated to the movement through which the algorithmic gaze of data-intensive biosciences aims to colonize other epistemic gazes, imposing its own mode of interpreting biological reality as hegemonic. In fact, Ricaurte (\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2019\u003c/span\u003e, 381) defined data colonialism as an imposition of ways of thinking that \u0026ldquo;denies the existence of alternative worlds and epistemologies.\u0026rdquo;\u003c/p\u003e\u003cp\u003eOf course, we do not take this colonization as a given; rather, we aim to problematize the relationships among different gazes. Kotliar himself offers conceptual tools for undertaking this avenue. First, he argues that, while the colonial logic of expansion renders things knowable, the algorithmic gaze does not necessarily rely on stable and bounded categorizations. Instead, it tends to favor \u0026ldquo;far from completed\u0026rdquo; and \u0026ldquo;allegedly more fine-grained\u0026rdquo; categories (Kotliar \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 928). The algorithmic gaze does not displace existing classifications but rather operates upon them \u0026ndash; as demonstrated by Leonelli\u0026rsquo;s work on bio-ontologies (2012b; 2016). At the same time, it opens up possibilities for exploring correlations and patterns that rigid categorizations would be unable to capture.\u003c/p\u003e\u003cp\u003eSecond, Kotliar (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 934) characterizes the algorithmic gaze as inherently multifocal: \u0026ldquo;a gaze that stems from a complex combination of very different types of lenses.\u0026rdquo; This multifocality provides a theoretical advantage: it allows us to argue that the algorithmic gaze has not supplanted the molecular gaze or the earlier clinical gaze. Rather than adopting metaphors such as \u0026ldquo;hybridization\u0026rdquo; (Keating and Cambrosio \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) or \u0026ldquo;integration\u0026rdquo; (O\u0026rsquo;Malley and Soyer \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), we propose the concept of \u003cem\u003enested multifocality\u003c/em\u003e. The algorithmic gaze is multifocal in that it draws on various lenses (moving across theories, epistemologies, and epistemic cultures), and, at the same time, it is embedded within a broader, multifocal vision that must accommodate both the medical and molecular gazes. Nested multifocality means that all these gazes operate together on multifocal platforms, where the vision must be trained to shift across different focal points in order to achieve a knowing sight on biological reality. Only when the sight align itself with these varying focal layers can it support claims to objectivity. What counts as evidence, then, is negotiated among different styles of reasoning (clinical, experimental, algorithmic) \u0026ndash; a negotiation that unfolds through the nested multifocality of the algorithmic gaze and its predecessors. In what follows, we empirically examine the algorithmic gaze and its nested multifocality within a translational oncology project that employs a bioinformatic platform.\u003c/p\u003e"},{"header":"5. Methodological note","content":"\u003cp\u003eIn order to explore the dynamics of nested multifocality among gazes (clinical, molecular, and algorithmic), we draw primarily on the ethnographic research conducted by author two within the SMAP project (acronym invented). SMAP is a precision oncology project conducted in Northern Italy aimed at ascertaining the role of genetic diversity and genomic aberrations in the evolution and progression of prostate cancer. One of the objectives of SMAP is to develop a clinically viable test for liquid biopsy, a technique that facilitates the collection of molecular elements contained in the blood, in order to expand its clinical use in molecular stratification, treatment selection, and monitoring of treatment resistance.\u003c/p\u003e\u003cp\u003eThe SMAP project is, therefore, structured around a network involving close collaboration between the research group and four public hospitals in Northern Italy. The hospitals employ clinical oncologists who are responsible for enrolling patients in the study and collecting blood samples (the bedside and the clinical gaze). Patient data and clinical records are entered into a database by a data manager. Blood samples are processed and sequenced in a designated facility by laboratory biologists (the bench/wet lab and the molecular gaze). Sequence data are processed and analyzed, together with patient data and data from public repositories, by bioinformaticians to identify biomarkers (the dry lab and the algorithmic gaze). It is, therefore, a project that combines different areas of expertise, in which clinicians, experimental biologists, and bioinformaticians negotiate the validation of results and define a shared view of what counts as evidence.\u003c/p\u003e\u003cp\u003eThe ethnographic research has been complemented by 22 semi-structured interviews with bioinformaticians and computational biologists working in Italy, conducted by authors two and three. The transcripts were analyzed using Atlas.ti for thematic analysis by author one, who also outlined the theoretical and analytical framework adopted in this paper.\u003c/p\u003e\u003cp\u003eThe distinction between bioinformaticians and computational biologists is not always straightforward, so we have chosen to rely on the self-identification provided by the interviewees. We use the acronyms JCB and SCB to refer to junior and senior Computational Biologists, and JBI and SBI for junior and senior Bioinformaticians, respectively. In total, we interviewed 15 bioinformaticians (6 senior, 9 junior) and 7 computational biologists (5 senior, 2 junior).\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eEthical approval\u003c/strong\u003e\u003cp\u003ewas granted by the Ethics Committee at the University of X. Interviewees signed a consent form prior to participating in the interviews. To ensure anonymity and prevent identification, any geographical or institutional references have been removed from the interview transcripts.\u003c/p\u003e\u003c/p\u003e"},{"header":"6. The persistence of gazes from the bedside to the bench and the computer","content":"\u003cp\u003eSMAP is fundamentally a project of translational and precision medicine in cancer genomics and, as such, represents a pivotal site for investigating the complex relationships among the heterogeneous professional visions involved: those of the clinical setting, the biology laboratory, and the bioinformatic computational infrastructure. Translational medicine has a long history and a complex genealogy (Harrington and Hauskeller \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), and attempts within STS to grasp its specificities have generated several analytical concepts. Translation has been placed at the core of contemporary biomedicine\u0026rsquo;s very essence. The notion of \u003cem\u003ebiomedical platforms\u003c/em\u003e (Keating and Cambrosio \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2000\u003c/span\u003e, 359) has been introduced to capture the articulation of clinical routines and biomedical innovation in ways that connect \u0026ldquo;the biological or normal\u0026rdquo; with \u0026ldquo;the medical or pathological\u0026rdquo; (see also Cambrosio et al. \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). In this sense, translational medicine has been described as a hybrid domain that configures and reconfigures both \u0026ldquo;persons and tools\u0026rdquo; and \u0026ldquo;biology, genomics, and medicine\u0026rdquo; (Kohli-Laven et al. \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2011\u003c/span\u003e, 488). Within this reconfiguration, the role of bioinformatics positions the dry lab somewhere within the two-way flow \u0026ldquo;from the bench to the bedside\u0026rdquo; (e.g., Douglas \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Levin \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWhether the focus is placed on the structural conditions, norms, and conventions of knowledge production and treatment (Cambrosio et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), or instead on the situated practices through which clinical staff and biomedical investigators mobilize their expertise \u0026ldquo;to pursue experimental protocols and fabricate clinically actionable knowledge\u0026rdquo; (Crabu \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e, 60), the question of \u0026ldquo;what counts as \u0026lsquo;precise knowledge\u0026rsquo; when practising cancer precision medicine\u0026rdquo; (ibid., 59) remains relevant. In what follows, we will investigate this issue not by directly examining situated practices, but rather by employing the analytical framework outlined above, by focusing on gazes, professional visions, and how different forms of expertise seek to manage the nested multifocality of the involved gazes.\u003c/p\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e6.1. The clinical gaze\u003c/h2\u003e\u003cp\u003eThe SMAP project began by providing the participating hospitals with equipment and instruments to perform blood sampling in a standardized manner. This was done in order to enable hospital staff to follow the standard operating procedures (SOPs) defined by biologists, ensuring a flow of comparable samples produced through codified protocols using the same instrumentation. SOPs and the standardization of instruments can be understood as a way of unifying the \u003cem\u003edomain of scrutiny\u003c/em\u003e by channeling the practices of \u003cem\u003ecoding\u003c/em\u003e and \u003cem\u003ehighlighting\u003c/em\u003e (Goodwin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1994\u003c/span\u003e) through the materiality of technological artefacts and the regulatory device of protocols.\u003c/p\u003e\u003cp\u003eThis does not mean that the professional visions of nurses, physicians, and oncologists are completely flattened into the styles of reasoning of molecular biologists and bioinformaticians, where standardization of instrument and protocols is seen as warranting the production of evidential value. Nor does it mean that hospital staff are confined within the boundaries of service tasks, such as enrolling new patients and scheduling appointments for the various blood sampling sessions. On the contrary, clinicians protect their professional domain (Gieryn \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1999\u003c/span\u003e, 16) by framing the work of the wet and dry labs as \u0026ldquo;a fundamental analysis for the patient,\u0026rdquo; as explained by Maurizio, an oncologist, who emphasized that the samples sent to the lab allow clinical oncologists to \u0026ldquo;decide on the treatment plan.\u0026rdquo; Moreover, oncologists do not always incorporate the molecular gaze and the clinical gaze into their practices of \u003cem\u003ehighlighting\u003c/em\u003e. According to Maurizio, in fact, \u0026ldquo;there are aspects that aren\u0026rsquo;t taken into consideration\u0026rdquo; by these gazes, elements that are instead consubstantial to clinicians\u0026rsquo; \u0026ldquo;socially organized ways of seeing\u0026rdquo; (Goodwin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1994\u003c/span\u003e, 606) their objects of relevance: the patients. As Maurizio explains:\u003c/p\u003e\u003cp\u003eThe oncologist chooses based on the situation, based on the patient... The data we receive from the lab (from bioinformaticians) don\u0026rsquo;t always represent the best strategy. They tell me that the strategy is to use drug X, which causes hand tremors after taking it, perhaps for life. If the patient is a pianist... it\u0026rsquo;s true that it\u0026rsquo;s the best strategy, it\u0026rsquo;s true that it would give him a better chance of survival, but he would no longer be able to play the piano. And from a biological standpoint, this isn\u0026rsquo;t considered; they say, \u0026ldquo;This one does what\u0026rsquo;s most beneficial for him.\u0026rdquo; (Maurizio, oncologist)\u003c/p\u003e\u003cp\u003eThis is also confirmed by our interviewees outside the SMAP project. Davide, an SCB working in pharmacogenomics, in describing his cooperation with clinicians, clearly states that \u0026ldquo;the doctor is the patient,\u0026rdquo; meaning that \u0026ldquo;if the patient needs a certain drug, even if it\u0026rsquo;s not the one you want to study, [clinicians] give the patient the drug the patient needs.\u0026rdquo; This approach applies to data collection as well: even when protocols call for samples to be taken at specific intervals, clinicians \u0026ldquo;can\u0026rsquo;t engineer it, so to speak.\u0026rdquo; Therefore, the regularity of measurement is often missed (e.g., \u0026ldquo;patients stopped coming\u0026rdquo;), and this misalignment from codified protocols (i.e., \u003cem\u003ecoding\u003c/em\u003e) is seen by other experts as \u0026ldquo;not [clinicians\u0026rsquo;] fault.\u0026rdquo; As Davide puts it: \u0026ldquo;I can\u0026rsquo;t say that the clinicians don\u0026rsquo;t know how to do their job, but their job is different.\u0026rdquo;\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e6.2. The molecular gaze\u003c/h2\u003e\u003cp\u003eBefore the wet lab enters into play, samples are handled by the data manager, who is responsible to control both the flow of biological materials (that must be processed and sequenced) and the flow of bio-information: patient clinical records must be anonymized and entered into a database shared with biologists and bioinformaticians.\u003c/p\u003e\u003cp\u003eProcessing and sequencing samples are two distinct tasks, requiring different expertise, over which biologists assert their \u003cem\u003eprofessional jurisdiction\u003c/em\u003e (Abbott \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1988\u003c/span\u003e) and draw boundaries with the work of bioinformaticians. Linda, a junior biologist assigned to processing samples for sequencing, explains:\u003c/p\u003e\u003cp\u003eWhen you do molecular biology on nucleic acids, you have to be much more careful with everything you do. Making a mistake can mean something as simple as setting the wrong volume on a pipette, so you have to be very precise and very tidy, otherwise, you\u0026rsquo;ll have to throw everything away\u0026hellip; If you make a mistake in bioinformatics, you can just go back with a button, but not here. (Linda, junior biologists)\u003c/p\u003e\u003cp\u003eAt the same time, the wet lab intertwines with the bedside, seeking to expand the boundaries of its jurisdictional control (Gieryn \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1999\u003c/span\u003e, 16) in order to shape the domain of scrutiny according to the molecular gaze. For instance, the evaluation of cancer severity is no longer delegated solely to computed tomography scans (within the jurisdiction of the clinical gaze), but also to biological assays, through which the oncologist is expected to \u0026ldquo;decide which path to take\u0026rdquo; based on molecular features (Carla, junior biologist). In this encounter between different platforms (Keating and Cambrosio \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), the wet lab asserts that its biomedical platform will provide clinicians with \u0026ldquo;the right choice\u0026rdquo; and \u0026ldquo;the optimal strategies for their situation\u0026rdquo; (Yuri, junior biologist). As noted by Nelson and colleagues (\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e, 75), in precision oncology clinical trials shift their nature: from \u003cem\u003etesting machines\u003c/em\u003e for a drug to becoming \u003cem\u003eclinical experimental systems\u003c/em\u003e grounded in biological hypotheses and molecular studies \u0026ndash; namely, \u0026ldquo;devices for materializing new questions about cancer biology and treatment.\u0026rdquo;\u003c/p\u003e\u003cp\u003eHowever, the clinical gaze cannot be dismissed or entirely colonized by the molecular one. To understand cancer biology and its progression, molecular biologists need to interface with the clinic. \u0026ldquo;None of us are medical doctors, and as much as we know biology... medicine is another matter entirely,\u0026rdquo; says Mirko, a senior biologist. Indeed, an understanding of cancer biology cannot be achieved solely through the experimental style of reasoning, as Marta explains:\u003c/p\u003e\u003cp\u003eWhen we don\u0026rsquo;t understand a piece of data, medical doctors can provide an explanation. Maybe it doesn\u0026rsquo;t make sense to us, and then they tell us, \u0026ldquo;No, look, that patient has this particular medical history, and therefore\u0026hellip;,\u0026rdquo; then it makes sense to us too. (Marta, senior biologist)\u003c/p\u003e\u003cp\u003eCrabu (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) introduced the notion of \u003cem\u003etechnomimicry\u003c/em\u003e to make sense of the reframing of clinical care and biomedical research in translational medicine, highlighting how the clinic and the laboratory can maintain distinct institutional identities while reconfiguring themselves to mimic one another. We instead propose the notion of \u003cem\u003emultifocality\u003c/em\u003e to capture how professional visions, along with the clinical and molecular gaze, seek to develop a multilayered optic to grasp a composite sight of biological phenomena, in order to ground claims about what counts as evidence. Multifocality, as we will discuss later, represents a novel rationalization \u0026ldquo;of the sight\u0026rdquo; (Latour \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e1990\u003c/span\u003e, 27) within a scientific field where different \u003cem\u003estyles of reasoning\u003c/em\u003e are assembled to produce evidential knowledge.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e6.3. The dry lab: toward the algorithmic gaze\u003c/h2\u003e\u003cp\u003eProcessed samples are then sent to the sequencing center, where Next-Generation Sequencing (NGS) technicians \u0026ldquo;take care of getting out from [biological samples] all the computer data that bioinformaticians need\u0026rdquo; (Martina, NGS technician). After that, the dry lab handles the data to identify biomarkers. This \u0026ldquo;industrialization\u0026rdquo; of biology, this \u0026ldquo;politics of productivity\u0026rdquo; based on automation and informatics (Stevens \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2011\u003c/span\u003e, 240), provides bioinformaticians with an alleged positional advantage, advancing claims of epistemic authority over other professional visions:\u003c/p\u003e\u003cp\u003eMany biologists\u0026hellip; have no background in quantitative fields like computer science or statistics. However, none of them would deny the importance of bioinformatic analysis in research. \u0026hellip; That is, there would be no bioinformatics if there were no data and no biological samples from which to collect that data. Likewise, there would be no wet lab as we know it today if massively parallel DNA sequencing did not exist. \u003cem\u003eWithout bioinformatics\u003c/em\u003e, and all that concerns the analysis of biological data, \u003cem\u003ebiology as we know it today would not exist\u003c/em\u003e. (Julien, SBI, emphasis added)\u003c/p\u003e\u003cp\u003eHowever, in accordance with the relational approach to data (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) that we adopt, what counts as data, and its related evidential value, is always determined within specific research situations. This means that bioinformatics does not exert hegemonic dominance over other styles of reasoning. Instead, within the SMAP project, other gazes must be incorporated into the algorithmic gaze of computational procedures in order to gain a clearer view of the biology of cancer under investigation. Salvatore, an SBI working on the SMAP project, explains how working with oncologists allows bioinformaticians to understand \u0026ldquo;what we\u0026rsquo;re looking at, what questions we need to ask or answer.\u0026rdquo; Silvio, a clinical oncologist, argues that \u0026ldquo;clinical parameters are just numbers\u0026rdquo; to bioinformaticians, who \u0026ldquo;don\u0026rsquo;t know what they mean\u0026rdquo; and therefore \u0026ldquo;do things that don\u0026rsquo;t make sense from a clinical point of view.\u0026rdquo;\u003c/p\u003e\u003cp\u003eSimilarly, the molecular gaze must also be brought into the \u003cem\u003eperceptual field\u003c/em\u003e (Goodwin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1994\u003c/span\u003e) when the object of relevance is fashioned: the biomarker identified through the algorithm. Luciano, a JBI working on the SMAP project, states that in \u0026ldquo;processing data\u0026hellip; analyzing them\u0026hellip; and in explaining the data you obtain,\u0026rdquo; you must \u0026ldquo;think at every moment about the biology behind\u0026rdquo; the data, always proceeding \u0026ldquo;in light of their biological relevance.\u0026rdquo; Otherwise, as Martina explains:\u003c/p\u003e\u003cp\u003e\u0026hellip; the bioinformatician runs their analyses based on their preferred algorithms\u0026hellip; but sometimes it happens that the bioinformatician, not knowing how the whole upstream part of the experiment was done on the wet lab side, comes here and asks for a meeting to try to make sense of things that don\u0026rsquo;t quite add up. (Martina, NGS technician)\u003c/p\u003e\u003cp\u003eThis once again brings the theme of multifocality into focus, along with its role in warranting evidence. However, we now need to explore more closely how the algorithmic gaze is constituted.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e6.4. The algorithmic gaze\u003c/h2\u003e\u003cp\u003eDavide is a senior researcher, who defines himself a computational biologists and declares to have \u0026ldquo;an algorithmic background\u0026rdquo; and therefore he tends to see \u0026ldquo;every\u0026hellip; biological problem\u0026rdquo; from what he calls an \u0026ldquo;algorithmic lens\u003csup\u003e[2]\u003c/sup\u003e.\u0026rdquo; Asked to specify the point he explains:\u003c/p\u003e\u003cp\u003e\"Algorithmic thinking,\u0026rdquo; in my opinion... uh... it\u0026rsquo;s like, you reason in terms of properties. You have an input, and you want a certain result at the end, but you need to figure out what happens in the middle, like, what kind of properties this \u0026ldquo;middle\u0026rdquo; needs to have. And usually, in the middle, there\u0026rsquo;s some kind of representation, I use graphs, basically, like, there are people who use strings... If you don\u0026rsquo;t have algorithmic experience, you only think in terms of start and end, but you can\u0026rsquo;t really figure out what should be in between. If you \u003cem\u003edo\u003c/em\u003e have experience, then you can say something like: \u0026ldquo;Okay, this problem can be tackled as a graph theory problem\u0026rdquo; \u0026hellip; I map everything into some n-dimensional space, then I look where the little point clouds end up, I do stuff. But that \u003cem\u003efeeling\u003c/em\u003e, of what\u0026rsquo;s supposed to be in the middle, only comes from algorithmic experience. If you\u0026rsquo;re a biologist, you just don\u0026rsquo;t think that a biological problem might actually be a graph problem \u0026hellip;So \u0026ldquo;algorithmic thinking\u0026rdquo; means understanding what could be in between the problem and the solution, and who can actually work on that. Like, if it\u0026rsquo;s a graph, we know how to work on graphs; if it\u0026rsquo;s strings, we have algorithms for strings; if it\u0026rsquo;s geometry, we\u0026rsquo;ve got geometric algorithms, etc. And once you see that, it kind of opens up a whole world, which might be a world you\u0026rsquo;re familiar with, more or less, but that the biologist maybe has no idea even exists (Davide, SCB)\u003c/p\u003e\u003cp\u003eIn this long excerpt, Davide outlines some of the main features of what the algorithmic gaze is. First, how this gaze translates biological mechanisms into informational ones. Chiara (SBI) says that what a bioinformatician does is to \u0026ldquo;translate\u0026rdquo; biology \u0026ldquo;into an interpretation closer to how an information engineer would think.\u0026rdquo; For her, cellular mechanisms are signals and pathways (see Landecker \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), which can be reconstructed and understood by \u0026ldquo;reasoning in terms of data.\u0026rdquo;\u003c/p\u003e\u003cp\u003eSecond, it highlights the theme of complexity. Bioinformaticians and computational biologists conceive their work as essential to dealing with the vast amount of data \u0026ndash; not simply in terms of managing a \u0026lsquo;data deluge\u0026rsquo;, but in enabling an understanding of biological phenomena that would be impossible without an algorithmic way of reasoning. They model \u0026ldquo;a systemic view of complex biological systems by involving integrative computational methods,\u0026rdquo; which provides \u0026ldquo;a tool to test the hypothesis, to understand whether you\u0026rsquo;re really grasping how the process works\u0026rdquo; (Flavio, SCB, emphasis added). They are developing computational methods \u0026ldquo;to make sense\u0026rdquo; (Emanuele, SCB) of the biological complexity that can only be addressed by \u0026ldquo;someone who works in information theory,\u0026rdquo; because \u0026ldquo;if you apply information theory to a DNA string, and extract some notion of how chaotic or not chaotic that language is, then you can get a measure of this complexity\u0026rdquo; (Riccardo, SBI). Algorithms step in when biological complexity exceeds the limits of human perception. This is the case with n-dimensionality, as mentioned by Davide, which, according to Chiara:\u003c/p\u003e\u003cp\u003eThen there\u0026rsquo;s the reasoning part of the algorithm, precisely because I\u0026rsquo;m dealing with complex data, complex patterns, complex relationships... I need to abstract a kind of reasoning that I\u0026rsquo;d normally use with just a few variables, in low-dimensional spaces\u0026hellip; But when I\u0026rsquo;m talking about machine learning, identifying patterns, I\u0026rsquo;m already on another level\u0026hellip; I can extend to higher-dimensional spaces, and I do that through an algorithm. (Chiara, SBI)\u003c/p\u003e\u003cp\u003eComplexity is closely tied to the third theme: the difference between experimental biologists and bioinformaticians in the tools they use to explore biological phenomena. Bioinformaticians draw a boundary between the experimental and the algorithmic that aligns with the point made by Stevens (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, 45), who argues that bioinformatics poses and answers \u0026ldquo;general questions\u0026rdquo; such as \u0026ldquo;what are the overall rules or patterns for how genomes or diseases behave?\u0026rdquo; (Stevens \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2011\u003c/span\u003e, 236). Alberto (JBI), for instance, points out that \u0026ldquo;at the wet-lab level you can\u0026rsquo;t study every single element of [the DNA] sequence in detail,\u0026rdquo; whereas \u0026ldquo;computational analysis and bioinformatics allow you to explore this world in depth.\u0026rdquo; Similarly, Riccardo (SBI) notes that geneticists using \u0026ldquo;classical methods\u0026rdquo; face serious difficulties when dealing with complex diseases, while algorithms can detect \u0026ldquo;what happens with 10, 20, or even 100 genes at once.\u0026rdquo; By stressing that bioinformaticians have tools that enable them \u0026ldquo;to ask questions that biologists and clinicians cannot ask themselves\u0026rdquo; (Riccardo, SBI), they claim professional jurisdiction (Abbott \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1988\u003c/span\u003e) and reject the idea that their work is merely a service (Lewis and Bartlett \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIs the algorithmic gaze therefore establishing its sovereignty over other biomedical gazes, as \u0026ldquo;the [algorithmic] eye that knows and decides, the [algorithmic] eye that governs\u0026rdquo; (Foucault \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2003\u003c/span\u003e, 89)? Is it supplanting the clinical and molecular gazes? Is data colonialism seeking \u0026ldquo;new territories to set its algorithmic eyes on\u0026rdquo; (Kotliar \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 922) and denying \u0026ldquo;the existence of alternative worlds and epistemologies\u0026rdquo; (Ricaurte \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2019\u003c/span\u003e, 381)?\u003c/p\u003e\u003cp\u003eRather, our definition of the algorithmic gaze suggests that, even if it does not rely on stable and bounded categorizations, it does not displace existing ones either (Kotliar \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 928). In this sense, Chiara (SBI) explains that \u0026ldquo;where there is prior knowledge \u0026hellip; of a physiological process,\u0026rdquo; this must be integrated into the algorithm. Moreover, data-driven algorithms are tested on data collected through wet-lab experiments, designed according to parameters defined by existing biological knowledge \u0026ndash; which, in turn, \u0026ldquo;has allowed us to forbid certain relationships between variables or make them mandatory.\u0026rdquo;\u003c/p\u003e\u003cp\u003eSo, while the algorithmic gaze detects \u0026ldquo;nonlinearities in the data that might escape the biotechnologist or biologist, but do not escape these algorithms\u0026rdquo; (Tommaso, JBI), the clinical and molecular gazes are not supplanted, because any \u0026ldquo;candidate pathway \u0026hellip; has to be investigated further in the wet lab to validate the result\u0026rdquo; (Marta, JBI). This leads to the question of validation and, consequently, to what counts as evidence: what is considered authoritative, objective knowledge \u0026ndash; even when it is produced through an algorithmic gaze.\u003c/p\u003e\u003c/div\u003e"},{"header":"7. What counts as evidence: the problem of validation","content":"\u003cp\u003eWhile we agree with Stevens (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, 67) that the algorithmic gaze involves \u0026ldquo;specific epistemologies, practices, and modalities of knowing,\u0026rdquo; this does not suffice to constitute a colonizing style of scientific reasoning that eliminates experimental practices. According to Hacking, a style of reasoning is \u0026ldquo;a standard or model of what it is to be reasonable about this or that type of subject matter\u0026rdquo; (Hacking, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e1992\u003c/span\u003e, p. 10). The question, then, becomes where the boundaries of this \u0026ldquo;subject matter\u0026rdquo; lie. Our interviewees, while defending their work as epistemically robust, do not seek to \u0026lsquo;extend the boundaries\u0026rsquo; (Gieryn \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1999\u003c/span\u003e) of their subject matter to encompass all biological and biomedical knowledge.\u003c/p\u003e\u003cp\u003eWhen we asked our interviewees about validation processes, their answers stand in contrast to the tenet made by Stevens that \u0026ldquo;computers have created new ways of making \u003cem\u003eauthorized\u003c/em\u003e and valuable knowledge\u0026rdquo; (Stevens \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, 102 emphasis added). They explained how they statistically validate their models and algorithms, how they split samples into training and validation sets, or how they use datasets retrieved from public databases to check whether the algorithm can generalize correctly. However, they consider these \u0026ldquo;sanity checks\u0026rdquo; (Stevens \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e, 64) insufficient to produce authorized evidence.\u003c/p\u003e\u003cp\u003eThe first point to consider is that evidential value is ascribed to the specific moment of inquiry (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e, 70), and it is shaped by the epistemic cultures involved\u0026mdash;since they \u0026ldquo;create and warrant knowledge\u0026rdquo; (Knorr-Cetina 1999). For example, in Genome-Wide Association Studies (GWAS), \u0026ldquo;validation is purely statistical\u0026hellip; there, wet-lab validation doesn\u0026rsquo;t even exist\u0026rdquo; (Elisa, SBI). Similarly, in metagenomics, the standard validation \u0026ldquo;is to replicate the study in another population\u0026rdquo; (Emanuele, SCB). In contrast, when we examine computational biology and bioinformatics applied to cancer genomics or other clinical investigations, the evidential value ascribed to statistical validation changes dramatically.\u003c/p\u003e\u003cp\u003eAccording to Chiara (SBI), in biomarker identification, algorithm-based models are scarcely reproducible, making them only \u0026ldquo;great exploratory tools.\u0026rdquo; However, in order \u0026ldquo;to verify that the biomarker is actually valid and that I can really use it as such,\u0026rdquo; the wet lab, with its \u0026ldquo;more standardized and standardizable techniques,\u0026rdquo; becomes \u0026ldquo;essential.\u0026rdquo; Similarly, Elisa (SBI) states that when \u0026ldquo;it comes to gene function or biological mechanisms, it\u0026rsquo;s more important, or anyway fundamental, to have the wet lab component.\u0026rdquo;\u003c/p\u003e\u003cp\u003eAmong our interviewees, there is a shared belief that experimentation (whether clinical or laboratory-based) holds an epistemic power that \u003cem\u003ein silico\u003c/em\u003e validation does not possess. This belief is sometimes attributed to epistemic cultures; for example, the biomedical field \u0026ldquo;still expects the lab to be the final step\u0026rdquo; (Matteo, SBI), and consequently, biomedical journals are \u0026ldquo;more oriented toward experimental validation\u0026rdquo; (Davide, SCB). Other times, it is simply seen as part and parcel of the research endeavor: since \u0026ldquo;biology is very complicated\u0026rdquo; and one \u0026ldquo;will never exactly reproduce biology,\u0026rdquo; the outcomes of any computational model \u0026ldquo;need to be taken with a pinch of salt\u0026rdquo; (Antonia, JBI). In fact, \u0026ldquo;going from saying that something is statistically significant\u0026hellip; to saying that it\u0026rsquo;s valid, there\u0026rsquo;s a whole world in between\u0026rdquo; (Tommaso, JBI).\u003c/p\u003e\u003cp\u003eThis does not mean that our interviewees devalue their algorithmic models; rather, they are aware that the algorithmic style of reasoning must be situated within a complex ecology of what counts as evidence in the biosciences. In this ecology, the algorithmic gaze plays an important role, but it is interwoven with other gazes and styles of reasoning in order to gain a more layered understanding of biological phenomena. The algorithmic gaze is linked to an emergent style of reasoning that adopts specific statistical procedures for pattern detection (Keating and Cambrosio \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), merging them with other, more traditional styles \u0026ndash; such as model construction, statistical analysis, and the ordering of variety (see Hacking \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e1992\u003c/span\u003e). However, \u0026ldquo;objectivity comes into being\u0026rdquo; (Hacking \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e1992\u003c/span\u003e, 10) not solely in silico, but rather through the interplay of a plurality of styles and a network of theories, practices, tools, and technologies that together constitute epistemic cultures (Knorr-Cetina 1999).\u003c/p\u003e"},{"header":"8. Nested multifocality or the meeting of multiple gazes","content":"\u003cp\u003eHow is the algorithmic gaze multifocal? A gaze is multifocal when it \u0026ldquo;stems from a complex combination of very different types of lenses\u0026rdquo; (Kotliar \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, 934). The algorithmic gaze is already multifocal when an algorithm explores data to find hidden patterns. Not only because it embeds a system of classification rooted in the construction of databases \u0026ndash; the classificatory theorization discussed by Leonelli (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) \u0026ndash; or because data interpretation is influenced by metadata that, according to Leonelli (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), express the embodied knowledge of the experimental craftwork that produced the data. Rather, it is the very construction of algorithms that is carried out from a multifocal perspective.\u003c/p\u003e\u003cp\u003eFlavio (SCB), for example, explains that when his team has to develop a novel pipeline, they have \u0026ldquo;periodical meetings with [biomedical] experts\u0026rdquo; in order to \u0026ldquo;understand the details of the data that they are going to share, [and] of the technique that is used to collect the data.\u0026rdquo; At the same time, computational people also need \u0026ldquo;feedback from them on the quality of the output that we have to produce or the quality of the results, etc.\u0026rdquo; Indeed, Davide (SCB), who has solid experience in graph theory and graph analysis, told us that applying algorithmic \u0026ldquo;tricks\u0026rdquo; used on other kinds of graphs doesn\u0026rsquo;t \u0026ldquo;work at all on biological ones,\u0026rdquo; requiring \u0026ldquo;new algorithmic challenges and solutions that [are] more specific to biological graphs.\u0026rdquo;\u003c/p\u003e\u003cp\u003eUsing the framework developed by Charles Goodwin (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1994\u003c/span\u003e), multifocality does not mean warping the practices of \u003cem\u003ecoding\u003c/em\u003e and \u003cem\u003ehighlighting\u003c/em\u003e toward a single, unified biomedical vision, but rather constructing algorithms and tools in such a way as to produce objects of knowledge understandable to, and consistent with, the professional visions of biologists and clinicians. This is well explained by Elisa:\u003c/p\u003e\u003cp\u003eIn my personal experience, this approach works when the other side [the clinical and the experimental ones] has contributed not to the concrete, implementational development of the code, but to the design phase of the method itself. ... This helps me understand whether the expected outputs align with the experience and knowledge of the clinician, geneticist, or biologist \u0026hellip; a geneticist, in the case of genetics, or a molecular biologist if lab experiments are involved, have, in a way, some a priori knowledge of a specific disease that I don\u0026rsquo;t have. They may \u003cem\u003ealso have seen the same problem from many different angles\u003c/em\u003e, so it becomes important to understand whether my \u003cem\u003eperspective\u003c/em\u003e is consistent or at odds with theirs. \u0026hellip; through this exchange and comparison, it becomes possible to refine the method, the algorithm, and consequently the tool itself, making it into something that can actually be used by clinicians. (Elisa, SBI, emphasis added)\u003c/p\u003e\u003cp\u003eAlgorithms must be free to explore data and detect patterns, but in ways that account for different focal layers. The algorithmic gaze must be multifocal not only to be validated, but also because biomedicine operates within multifocal platforms, where different epistemic cultures and professional visions meet and seek to establish alignment on what counts as evidence. Without multifocal lenses, the algorithmic gaze \u0026ldquo;may not be able to capture a difference that, instead, the wet lab identified because they were working at a different, downstream level\u0026rdquo; (Alberto, JBI).\u003c/p\u003e\u003cp\u003eNested multifocality seeks to capture how a multifocal gaze operates across different optical layers, where evidential value emerges from the convergence of multiple gazes. This meeting, however, is challenging, always open to potential contestation and friction, because different gazes are not always easily accommodated. Each gaze is entrenched in a professional vision, with its own consolidated styles of reasoning through which \u0026ldquo;they have settled what it is to be objective\u0026rdquo; (Hacking \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e1992\u003c/span\u003e, 4). The algorithmic gaze must be multifocal in order to prevent different professional visions from clashing in jurisdictional disputes over biological phenomena, rather than constituting a shared \u0026ndash; though multilayered \u0026ndash; domain of scrutiny. Or, to use Abbott\u0026rsquo;s terms (1988, 41), to prevent the process of \u003cem\u003ecolligation\u003c/em\u003e \u0026ndash; that is, the act of establishing \u0026ldquo;rules declaring what kinds of evidence are relevant and irrelevant, valid and invalid\u0026rdquo;\u0026mdash;from turning into an irreconcilable wrangle.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this paper, we have explored how \u003cem\u003ewhat counts as evidence\u003c/em\u003e in data-intensive biosciences can be rethought through the conceptual lenses of the algorithmic gaze and nested multifocality. Drawing on the idea of \u003cem\u003edata as relational categories\u003c/em\u003e (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), we argued that evidence does not simply emerge from the technical output of computational methods. Rather, the production of evidential value is a situated epistemic achievement, conditioned by the alignment of an emergent computational \u003cem\u003estyle of reasoning\u003c/em\u003e (Hacking \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e1992\u003c/span\u003e) with other, coexisting professional visions (Goodwin \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1994\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn this sense, the algorithmic gaze does not assert epistemic sovereignty over clinical or molecular gazes. Instead, what counts as objective knowledge emerges through the negotiation of multiple styles of reasoning within historically specific \u003cem\u003eepistemic cultures\u003c/em\u003e (Knorr-Cetina 1999) and contingent \u003cem\u003eresearch situations\u003c/em\u003e (Leonelli \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The complexity of contemporary biosciences, marked by large heterogeneous collaboration networks (Vermeulen 2009; Leonelli \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), requires the careful articulation of multiple epistemic perspectives.\u003c/p\u003e\u003cp\u003eIt is in this space of encounter that we situate the notion of \u003cem\u003emultifocality\u003c/em\u003e: a metaphor for the epistemic condition in which evidence is no longer anchored to a singular mode of seeing, but emerges from the dynamic interplay of heterogeneous gazes. The \u003cem\u003eknowing eye\u003c/em\u003e, must learn to shift across layered focal planes in order to achieve a comprehensive sight.\u003c/p\u003e\u003cp\u003eBy exploring a translational medicine project involving the clinic, the wet lab, and the dry lab \u0026ndash; and by interviewing bioinformaticians and computational biologists \u0026ndash; we have defined a particular epistemology, which we term \u003cem\u003enested multifocality\u003c/em\u003e.\u003c/p\u003e\u003cp\u003eThis notion captures both the multifocal nature of algorithms, designed to traverse and include multiple socio-epistemically organized perceptual frameworks, and the normative requirement that validated knowledge is what is visible through the various professional visions involved. The algorithmic gaze, in this view, is not an autonomous epistemic force, but a situated way of seeing that gains epistemic legitimacy only when it contributes to a shared \u0026ndash; though multilayered \u0026ndash; domain of scrutiny.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eF.G. conducted the ethnographic fieldwork and 12 interviews under the supervision of L.B. E.H. conducted 10 interviews under the supervision of L.B. L.B. analysed the interviews transcripts and worked on the theoretical and analytical framework.L.B. wrote the main manuscript text (section 1-2-3-4-8)L.B. and F.G. wrote together the remaining sections (5-6-7)All authors reviewed the manuscript\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbbott, A. 1988. \u003cem\u003eThe System of Professions\u003c/em\u003e. Chicago: University of Chicago Press.\u003c/li\u003e\n\u003cli\u003eAllen, J.F. (2001) Bioinformatics and discovery: induction beckons again. \u003cem\u003eBioEssays\u003c/em\u003e 23(1): 104\u0026ndash;107.\u003c/li\u003e\n\u003cli\u003eAnderson, C. (2008). The End of Theory: The Data Deluge Makes the Scientific Method Obsolete. \u003cem\u003eWired \u003c/em\u003eJun 23, 2008 12:00 PM https://www.wired.com/2008/06/pb-theory/\u003c/li\u003e\n\u003cli\u003eBartlett, A., J. Lewis, and M.L. Williams (2016) Generations of interdisciplinarity in bioinformatics. \u003cem\u003eNew Genetics and Society\u003c/em\u003e 35(2): 186\u0026ndash;209.\u003c/li\u003e\n\u003cli\u003eBechtel, W. (2020) Data Journeys Beyond Databases in Systems Biology: Cytoscape and NDEx. In: S. Leonelli and N. Tempini (eds.) \u003cem\u003eData Journeys in the Sciences\u003c/em\u003e. Springer International Publishing, pp. 121-144\u003c/li\u003e\n\u003cli\u003eBourret, P., P. Castel, H. Bergeron, and A. Cambrosio (2021) Organizing precision oncology: introduction to the special issue. \u003cem\u003eNew Genetics and Society\u003c/em\u003e 40(1): 1\u0026ndash;6.\u003c/li\u003e\n\u003cli\u003eBrown, P.O., and D. Botstein (1999) Exploring the new world of the genome with DNA microarrays. \u003cem\u003eNature Genetics\u003c/em\u003e 21(Suppl 1): 33\u0026ndash;37.\u003c/li\u003e\n\u003cli\u003eBurgio, V., and R. Raffaet\u0026agrave; (2024) Organizing Microbial Diversity and Interspecies Relations through Diagrams: Trees, Maps, and the Visual Semiotics of the Living. \u003cem\u003eBiosemiotics\u003c/em\u003e 17(3): 817\u0026ndash;844.\u003c/li\u003e\n\u003cli\u003eCambrosio, A., P. Keating, T. Schlich, and G. Weisz (2006) Regulatory objectivity and the generation and management of evidence in medicine. \u003cem\u003eSocial Science and Medicine\u003c/em\u003e 63(1): 189\u0026ndash;199.\u003c/li\u003e\n\u003cli\u003eCambrosio, A., P. Keating, T. Schlich, and G. Weisz (2009) Biomedical Conventions and Regulatory Objectivity: A Few Introductory Remarks. Social Studies of Science 39(5): 651\u0026ndash;664.\u003c/li\u003e\n\u003cli\u003eCambrosio, A., P. Keating, E. Vignola-Gagn\u0026eacute;, S. Besle, and P. Bourret (2018) Extending experimentation: oncology\u0026rsquo;s fading boundary between research and care. New Genetics and Society 37(3): 207\u0026ndash;226.\u003c/li\u003e\n\u003cli\u003eChiapperino, L. (2024) Enacting biosocial complexity: Stress, epigenetic biomarkers and the tools of postgenomics. \u003cem\u003eSocial Studies of Science\u003c/em\u003e \u003cem\u003e54\u003c/em\u003e(4): 598-625.\u003c/li\u003e\n\u003cli\u003eCrabu, S. (2016) Translational Biomedicine in Action: Constructing Biomarkers across Laboratory and Benchside. \u003cem\u003eSocial Theory \u0026amp; Health\u003c/em\u003e 14(3): 312\u0026ndash;331.\u003c/li\u003e\n\u003cli\u003eCrabu, S. (2021) Organizing the precision clinic: arranging expertise, knowledge and technologies in cancer precision medicine clinical trials. New Genetics and Society 40(1): 58\u0026ndash;72.\u003c/li\u003e\n\u003cli\u003eDouglas, C.M.W. (2014) The Role of Bioinformatics in Facilitating Translational Science and Medicine. \u003cem\u003eTecnoscienza \u0026ndash; Italian Journal of Science \u0026amp; Technology Studies\u003c/em\u003e 5(1): 141\u0026ndash;163.\u003c/li\u003e\n\u003cli\u003eFasel, C., and L. Chiapperino (2023) Between the genotype and the phenotype lies the microbiome: symbiosis and the making of \u0026lsquo;postgenomic\u0026rsquo; knowledge. \u003cem\u003eHistory and Philosophy of the Life Sciences\u003c/em\u003e 45(4): 43..\u003c/li\u003e\n\u003cli\u003eFoucault, M. (1998) On the Archaeology of the Sciences: Response to the Epistemology Circle. In: J. D. Faubion (ed.) \u003cem\u003eAesthetics, Method, And Epistemology. \u003c/em\u003eNew York: The New Press, pp. 297-333.\u003c/li\u003e\n\u003cli\u003eFoucault, M. (2003) \u003cem\u003eThe Birth of the Clinic. An Archaeology of Medical Perception\u003c/em\u003e. London: Routledge\u003c/li\u003e\n\u003cli\u003eGieryn, T. F. (1983) Boundary-Work and the Demarcation of Science from Non-Science: Strains and Interests in Professional Ideologies of Scientists. \u003cem\u003eAmerican Sociological Review\u003c/em\u003e 48 (6): 781\u0026ndash; 795.\u003c/li\u003e\n\u003cli\u003eGieryn, T. F. (1999) \u003cem\u003eCultural Boundaries of Science. Credibility on the Line\u003c/em\u003e. Chicago: University of Chicago Press.\u003c/li\u003e\n\u003cli\u003eGilbert, W. (1991). Towards a Paradigm Shift in Biology. \u003cem\u003eNature\u003c/em\u003e 349 (6305): 99\u0026ndash;99.\u003c/li\u003e\n\u003cli\u003eGoodwin, C. (1994) Professional Vision. \u003cem\u003eAmerican Anthropologist\u003c/em\u003e 96(3): 606-633. \u003c/li\u003e\n\u003cli\u003eGraham, S. D. (2010) Combat Zones that See: Urban Warfare and US Military Technology. In \u003cem\u003eObservant States. Geopolitics and Visual Culture\u003c/em\u003e, ed. F. MacDonald, R. Hughes and K Dodds, 199-223. London: I.B. Tauris. \u003c/li\u003e\n\u003cli\u003eGriesemer, J. (2020)A Data Journey Through Dataset-Centric Population Genomics. In: S. Leonelli and N. Tempini (eds.) \u003cem\u003eData Journeys in the Sciences\u003c/em\u003e. Springer International Publishing, pp. 145-167.\u003c/li\u003e\n\u003cli\u003eHacking, I. (1992) \u0026lsquo;Style\u0026rsquo; for Historians and Philosophers. \u003cem\u003eStudies in History and Philosophy of Science Part A\u003c/em\u003e 23 (1): 1\u0026ndash;20..\u003c/li\u003e\n\u003cli\u003eHarrington, J., and C. Hauskeller (2014) Translational Research: An Imperative Shaping the Spaces in Biomedicine. Tecnoscienza \u0026ndash; Italian Journal of Science \u0026amp; Technology Studies 5(1): 191\u0026ndash;201.\u003c/li\u003e\n\u003cli\u003eKay, L.E. (2000) \u003cem\u003eWho wrote the book of life? A history of the genetic code\u003c/em\u003e. Stanford, CA: Stanford University Press.\u003c/li\u003e\n\u003cli\u003eKeating, P., and A. Cambrosio (2000) Biomedical Platforms. \u003cem\u003eConfigurations\u003c/em\u003e 8(3): 337\u0026ndash;387.\u003c/li\u003e\n\u003cli\u003eKeating, P., and A. Cambrosio (2012) Too many numbers: Microarrays in clinical cancer research. \u003cem\u003eStudies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences\u003c/em\u003e 43(1): 37\u0026ndash;51.\u003c/li\u003e\n\u003cli\u003eKeller, E.F. (1995) \u003cem\u003eRefiguring life: Metaphors of twentieth-century biology\u003c/em\u003e. New York: Columbia University Press.\u003c/li\u003e\n\u003cli\u003eKeller, E. F. (2000) \u003cem\u003eThe Century of the Gene\u003c/em\u003e. Cambridge, MA: Harvard University Press.\u003c/li\u003e\n\u003cli\u003eKnorr Cetina, K. (1999) \u003cem\u003eEpistemic Cultures. How the Sciences Make Knowledge\u003c/em\u003e, Cambridge, MA: Harvard University Press.\u003c/li\u003e\n\u003cli\u003eKohli-Laven, N., P. Bourret, P. Keating, and A. Cambrosio (2011) Cancer clinical trials in the era of genomic signatures: Biomedical innovation, clinical utility, and regulatory- scientific hybrids. Social Studies of Science 41(4): 487\u0026ndash;513.\u003c/li\u003e\n\u003cli\u003eKotliar, D.M. (2020) Data orientalism: on the algorithmic construction of the non-Western other. \u003cem\u003eTheory and Society\u003c/em\u003e 49(5\u0026ndash;6): 919\u0026ndash;939.\u003c/li\u003e\n\u003cli\u003eKotliar, D.M., and R. Grosglik (2023) On the Contesting Conceptualisation of the Human Body: Between \u0026lsquo;Homo-Microbis\u0026rsquo; and \u0026lsquo;Homo-Algorithmicus.\u0026rsquo; \u003cem\u003eBody \u0026amp; Society\u003c/em\u003e 29(3): 81\u0026ndash;108.\u003c/li\u003e\n\u003cli\u003eKrohs, U. (2012) Convenience experimentation. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 43(1): 52\u0026ndash;57.\u003c/li\u003e\n\u003cli\u003eLandecker, H. (2016) The Social as Signal in the Body of Chromatin. \u003cem\u003eThe Sociological Review\u003c/em\u003e 64(1_suppl): 79\u0026ndash;99.\u003c/li\u003e\n\u003cli\u003eLatour, B. (1990) Drawing Things Together. In: M. Lynch and S. Woolgar (eds.) \u003cem\u003eRepresentation in Scientific Practice. \u003c/em\u003eCambridge, MA: The MIT Press, pp. 19-68.\u003c/li\u003e\n\u003cli\u003eLeonelli, S. (2012a) Introduction: Making sense of data-driven research in the biological and biomedical sciences. \u003cem\u003eStudies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences\u003c/em\u003e 43(1): 1\u0026ndash;3.\u003c/li\u003e\n\u003cli\u003eLeonelli, S. (2012) Classificatory Theory in Data-intensive Science: The Case of Open Biomedical Ontologies. \u003cem\u003eInternational Studies in the Philosophy of Science\u003c/em\u003e 26(1): 47\u0026ndash;65.\u003c/li\u003e\n\u003cli\u003eLeonelli, S. (2013) Global data for local science: Assessing the scale of data infrastructures in biological and biomedical research. BioSocieties 8(4): 449\u0026ndash;465.\u003c/li\u003e\n\u003cli\u003eLeonelli, S. (2015) What Counts as Scientific Data? A Relational Framework. \u003cem\u003ePhilosophy of Science \u003c/em\u003e82(5): 810\u0026ndash;821. doi:10.1080/02698595.2012.653119.\u003c/li\u003e\n\u003cli\u003eLeonelli, S. (2016) \u003cem\u003eData-centric Biology. A philosophical study. C\u003c/em\u003ehicago: University of Chicago Press\u003c/li\u003e\n\u003cli\u003eLeonelli, S. (2020) Learning from Data Journeys. In \u003cem\u003eData Journeys in the Sciences\u003c/em\u003e, ed S. Leonelli and N. Tempini, 1-24. Springer International Publishing.\u003c/li\u003e\n\u003cli\u003eLeonelli, S. and Tempini N. (eds.) (2020) \u003cem\u003eData Journeys in the Sciences\u003c/em\u003e. Springer International Publishing.\u003c/li\u003e\n\u003cli\u003eLevin, N. (2014) What\u0026rsquo;s Being Translated in Translational Research? Making and Making Sense of Data between the Laboratory and the Clinic. Tecnoscienza \u0026ndash; Italian Journal of Science \u0026amp; Technology Studies 5(1): 91\u0026ndash;113.\u003c/li\u003e\n\u003cli\u003eLewis, J. and A. Bartlett (2013) Inscribing a discipline: tensions in the field of bioinformatics. \u003cem\u003eNew Genetics and Society\u003c/em\u003e 32(3): 243\u0026ndash;263.\u003c/li\u003e\n\u003cli\u003eLynch, M. (1985) Discipline and the Material Form of Images: An Analysis of Scientific Visibility. \u003cem\u003eSocial Studies of Science\u003c/em\u003e 15(1): 37\u0026ndash;66.\u003c/li\u003e\n\u003cli\u003eLynch, M. (1988) The externalized retina: Selection and mathematization in the visual documentation of objects in the life sciences. \u003cem\u003eHuman Studies\u003c/em\u003e 11(2\u0026ndash;3): 201\u0026ndash;234.\u003c/li\u003e\n\u003cli\u003eMackenzie, A. (2003) Bringing sequences to life: how bioinformatics corporealizes sequence data. New Genetics and Society 22(3): 315\u0026ndash;332. \u003c/li\u003e\n\u003cli\u003eM\u0026uuml;ller-Wille, S. and I. Charmantier (2012) Natural history and information overload: The case of Linnaeus. \u003cem\u003eStudies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences\u003c/em\u003e 43(1): 4\u0026ndash;15.\u003c/li\u003e\n\u003cli\u003eNelson, N. C., P. Keating, A. Cambrosio, A. Aguilar-Mahecha, and M. Basik (2014) Testing devices or experimental systems? Cancer clinical trials take the genomic turn. Social Science \u0026amp; Medicine 111, 74\u0026ndash;83.\u003c/li\u003e\n\u003cli\u003eO\u0026rsquo;Malley, M.A. and O.S. Soyer (2012) The roles of integration in molecular systems biology. \u003cem\u003eStudies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences\u003c/em\u003e 43(1): 58\u0026ndash;68.\u003c/li\u003e\n\u003cli\u003eRaffaet\u0026agrave;, R. (2022) \u003cem\u003eMetagenomic Futures: How Microbiome Research is Reconfiguring Health and What it Means to Be Human\u003c/em\u003e. London: Routledge.\u003c/li\u003e\n\u003cli\u003eRheinberger, H.-J,. and S. M\u0026uuml;ller-Wille (2017) \u003cem\u003eThe Gene. From Genetics to Postgenomics\u003c/em\u003e. Chicago: University of Chicago press.\u003c/li\u003e\n\u003cli\u003eRicaurte, P. (2019). Data epistemologies, the Coloniality of power, and resistance. \u003cem\u003eTelevision and New Media \u003c/em\u003e20(4): 350\u0026ndash;365. https://doi.org/10.1177/1527476419831640.\u003c/li\u003e\n\u003cli\u003eRose, N. (2006)\u003cem\u003e The Politics of Life Itself : Biomedicine, Power, and Subjectivity in the Twenty-First Century, \u003c/em\u003ePrinceton, NJ: Princeton University Press\u003cem\u003e.\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eStevens, H. (2011) On the means of bio-production: Bioinformatics and how to make knowledge in a high-throughput genomics laboratory. \u003cem\u003eBioSocieties\u003c/em\u003e 6(2): 217\u0026ndash;242..\u003c/li\u003e\n\u003cli\u003eStevens H. (2013) \u003cem\u003eLife Out of Sequence: A Data-Driven History of Bioinformatics\u003c/em\u003e. Chicago: University of Chicago Press.\u003c/li\u003e\n\u003cli\u003eStrasser, B.J. (2012) Data-driven sciences: From wonder cabinets to electronic databases. \u003cem\u003eStudies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences\u003c/em\u003e 43(1): 85\u0026ndash;87..\u003c/li\u003e\n\u003cli\u003eTempini, N. (2021) Data curation-research: practices of data standardization and exploration in a precision medicine database. \u003cem\u003eNew Genetics and Society\u003c/em\u003e 40(1): 73\u0026ndash;94.. doi:10.1080/14636778.2020.1853513.\u003c/li\u003e\n\u003cli\u003eTempini, N. \u0026amp; Leonelli, S. (2018) Concealment and discovery: The role of information security in biomedical data re-use. \u003cem\u003eSocial Studies of Science\u003c/em\u003e 48(5): 663\u0026ndash;690.\u003c/li\u003e\n\u003cli\u003evan Baren-Nawrocka, J., Consoli, L. \u0026amp; Zwart, H. (2020) Calculable bodies: Analysing the enactment of bodies in bioinformatics. \u003cem\u003eBioSocieties\u003c/em\u003e 15(1): 90\u0026ndash;114.\u003c/li\u003e\n\u003cli\u003eVermeulen, N. (2016) Big Biology. NTM Zeitschrift f\u0026uuml;r Geschichte der Wissenschaften, Technik und Medizin 24(2): 195\u0026ndash;223.\u003c/li\u003e\n\u003cli\u003eZuboff, S. (2019) \u003cem\u003eThe age of surveillance capitalism: The fight for a human future at the new frontier of power\u003c/em\u003e. New York: Public Affairs.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e We use the term bioscience and derivatives to include biology and biomedicine.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e The Italian word used by the interviewee was \u0026ldquo;ottica,\u0026rdquo; which is polysemous: in a literal sense, it means \u0026ldquo;optical\u0026rdquo; or \u0026ldquo;lens,\u0026rdquo; but in a more abstract sense, it can also mean \u0026ldquo;perspective,\u0026rdquo; \u0026ldquo;point of view,\u0026rdquo; or \u0026ldquo;approach.\u0026rdquo; We chose to preserve the literal sense in our translation.\u003c/span\u003e \u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1: Professional profiles of ethnographic informants\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003eNo.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003e\u003cem\u003ePseudonym\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eRole in SMAP project\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eMaurizio\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eSenior oncologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eLinda\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eJunior biologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eCarla\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eJunior biologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eYuri\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eJunior biologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eMirko\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eSenior biologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eMartina\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eSenior NGS technician\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eAdriano\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eJunior bioinformatician\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eJulien\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eSenior bioinformatician\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eSilvio\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eSenior oncologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eNicola\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eJunior biologist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eLuciano\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eJunior bioinformatician\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 13.7019%;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25%;\"\u003e\n \u003cp\u003eEdoardo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 61.2981%;\"\u003e\n \u003cp\u003eSenior data manager\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 2: Professional profiles of interviewed participants\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" class=\"fr-table-selection-hover\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003eNo.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003e\u003cem\u003ePseudonym\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003e\u003cem\u003eRole\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003e\u003cem\u003eField and subfield\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eChiara\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, omics data analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eMatteo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, pangenomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eVittorio\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, meta-analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eGiulia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, cancer genomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eDavide\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, pharmacogenomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eElisa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, omics data analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eAlberto\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, omics data analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eRiccardo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, omics data analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eTommaso\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, omics data analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eAmedeo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, cancer genomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eFlavio\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, pharmacogenomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eGabriele\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, RNA regulatory networks\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eStefano\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, cancer genomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eEmanuele\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, metagenomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eMarta\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, RNA networks\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eFrancesco\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, transcriptomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eValentina\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, cancer genomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eAntonia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, pharmacokinetics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eSerena\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, cancer genomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eClaudia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJCB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eComputational biology, metagenomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eDiego\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eJBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, epigenetics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 6.42361%;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19.6181%;\"\u003e\n \u003cp\u003eRenato\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.1944%;\"\u003e\n \u003cp\u003eSBI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 60.7639%;\"\u003e\n \u003cp\u003eBioinformatics, epigenetics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNote: Abbreviations used in the table as follows: JBI = junior bioinformatician; SBI = senior bioinformatician; JCB = junior computational biologist; SCB =senior computational biologist\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"biosocieties","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [BioSocieties](https://link.springer.com/journal/41292)","snPcode":"41292","submissionUrl":"https://submission.springernature.com/new-submission/41292/3","title":"BioSocieties","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Algorithmic gaze, Nested multifocality, Validation, Bioinformatics, Computational biology","lastPublishedDoi":"10.21203/rs.3.rs-7246270/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7246270/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn this article, we introduce the notions of \u003cem\u003ealgorithmic gaze\u003c/em\u003e and \u003cem\u003enested multifocality\u003c/em\u003e as analytical categories to investigate the production of evidence in contemporary biosciences. We take a critical stance against certain rhetorics of data-driven science that suggest what counts as evidence is increasingly less the result of experimental procedures and more the outcome of computational methods and pattern recognition algorithms. Drawing on an ethnographic study of a translational medicine project involving the clinic, the biological lab, and bioinformatic work \u0026ndash; as well as interviews with bioinformaticians and computational biologists \u0026ndash; this paper shows how evidence emerges through the negotiation of different gazes and professional visions. We define the \u003cem\u003ealgorithmic gaze\u003c/em\u003e as the correlate of a computational style of reasoning, whose output is validated insofar as it incorporates both the clinical and the molecular gaze. The concept of \u003cem\u003enested multifocality\u003c/em\u003e accounts for an epistemic condition in which evidence is not only relative to epistemic cultures and specific research situations, but also emerges through a broader multifocal vision that accommodates different gazes and professional visions.\u003c/p\u003e","manuscriptTitle":"Validating what counts as knowledge: The algorithmic gaze and nested multifocality in the biosciences","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-02 10:56:13","doi":"10.21203/rs.3.rs-7246270/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-12-18T17:16:00+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-28T09:40:11+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-07T23:12:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"250780061336206178412136543625425707498","date":"2025-11-04T15:17:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"127098401129523123239568670624530470931","date":"2025-11-02T17:44:10+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-09T10:13:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"34290432722491841307039568947871009311","date":"2025-10-08T07:38:41+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"154371590457709662453199512409743377040","date":"2025-10-03T16:36:00+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-25T17:27:59+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-11T05:59:20+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-11T05:59:10+00:00","index":"","fulltext":""},{"type":"submitted","content":"BioSocieties","date":"2025-07-29T18:57:12+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"biosocieties","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [BioSocieties](https://link.springer.com/journal/41292)","snPcode":"41292","submissionUrl":"https://submission.springernature.com/new-submission/41292/3","title":"BioSocieties","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"28c12284-c413-4833-9b62-4d989a941777","owner":[],"postedDate":"September 2nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-22T18:38:18+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-02 10:56:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7246270","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7246270","identity":"rs-7246270","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}