On May 4 and 5, 2023, an interdisciplinary conference on the topic of took place at the Collaborative Research Center 1385 Law and Literature (Münster, Germany). In the following, the conference’s organizers report on the issues discussed at the conference and outline the project of an associated book publication (this project has been meanwhile realized in a slightly different form, Harst/Celik/Jengdes 2024).

1. Questions

When Ginzburg (1995) formulated the thesis that the humanities, like crime fiction, were based on the so-called evidential paradigm, he had Sherlock Holmes in mind, a detective who would visit crime scenes himself. He would then collect clues there, combine them and come to the solution of his case through often ingenious, but also highly speculative conclusions. However, as clues can only be identified as such within the framework of an overall narrative (course of events), Ginzburg emphasized the narrative aspect of interpreting clues and argued that the humanities search for and interpret clues in a similar fashion. Their acquisition of knowledge strongly resembles the detective’s. In addition, the evidential paradigm has been used in various ways to analyze literary and pictorial works of art: on the one hand, works of art can be examined for traces of past reality or, conversely, the evidential paradigm may be utilized as an artistic device to construct reality or guide our reception.

Against the backdrop of current developments in research and investigation, however, the evidential paradigm requires revision. Ever since the private detective is facing competition from Kommissar Computer (Hartung 2010), investigative practices have changed fundamentally. Computer-aided search and investigation methods can now supplement or even replace a crime scene inspection. In the popular narrative of modern investigation, the individual investigator is thus replaced by a team of forensic experts who solve cases using digital data processing. We can therefore observe a fundamental change in the patterns of interpretation of the world of crime: While Columbo’s investigations, for instance, mostly started on a social and psychological level, murderers and victims are now only of interest as carriers of traces that can generate evidence (Gugerli 2007: 12, our translation). With reference to the successful series (USA 2000–2015), the question of what repercussions this has on criminal and legal practice is discussed under the heading of the CSI effect (ibid.). One of the key problems caused by it is the question as to what extent a judge can understand how algorithmic trace analysis works. Additionally, there is a potential risk that a judge’s assessment of evidence will primarily be based on a belief in the proper functioning of algorithms.

While such investigative practices are associated with positivist claims regarding knowledge in popular narratives, their constructivist dimension needs to be examined from an academic perspective: Reconstructions, visualizations, and simulations produce evidence. Visualizations often achieve an evidentiality that goes beyond the data they are based on. They are preceded by the digital amplification and aggregation of data into traces. Similar to the scientific experimental system (Rheinberger 2007), digital traces can be considered technically produced mediators that indicate the invisible.

Today’s open source investigations, such as those conducted by Forensic Architecture and Bellingcat, lend themselves to such reflections. Firstly, the investigations of these “digital sherlocks” are based on the digitization of crime scenes and computer-aided data processing, often drawing on freely accessible documentary material. Their task lies in making the often randomly generated material legible so that traces emerge from it. Secondly, and in addition to research reports, Forensic Architecture produces artfully composed documentaries that aesthetically and rhetorically underpin their material-based and activist standpoint. Thirdly, Forensic Architecture and comparable investigative artists (e.g. the photographers Trevor Paglen and Edmund Clark) provide their media artifacts with a meta-reflexive discourse that reflects virtual investigations and their claim to knowledge – also with reference to Ginzburg and the evidential paradigm (cf. Weizman/Fuller 2021: ch. 10).

2. Virtuality

The subject of virtual investigations does not only refer to contemporary open source intelligence, but also addresses the epistemological configurations of virtuality and traces in recent cultural history. For that reason, increasing attention has been paid to literary crime and detective fiction, which in turn has provided decisive impulses for reflecting virtuality. In his reading of Poe’s , Lacan (1966) described the purloined letter as a virtual object, which differs from the real object in that it can be missing from its place: While the police investigate the suspect’s property in search of the stolen letter as meticulously as unsuccessfully, the private detective Dupin recognizes the letter precisely where one would never expect to find it, namely on the letter holder – it has simply been made unrecognizable by being turned over. Deleuze also refers to this reading when he negotiates difference and repetition in the sign of virtuality (Deleuze 1968: 135f.). Like Lacan, Deleuze links the virtual with the symbolic in order to ultimately connect it with structure and difference: Language, for example, is virtual as a differential structure (langue) that is actualized in individual speech acts (parole). Accordingly, reality and virtuality must not be understood as opposites (structures are very much real), just like, conversely, virtuality and possibility/potentiality are not synonyms (the actualization of a structure is not a reduction of a spectrum of possibilities).

The significance of this distinction becomes even clearer once we look at another domain of the virtual: the use of statistics and probability to observe, manage, and govern modern societies, which began in the 18th and 19th centuries (cf. Schäffner 1999, Nassehi 2019). It allows a calculation with the unknown, such as one of the current population on the basis of birth registers and death probability (cf. Schäffner 1999: 124). The accompanying shift from truth to probability leads to a virtualization of knowledge, resulting in the distinction between real and non-real being transformed into a continuity of more or less probable events. In this regard, too, virtuality is not opposed to reality but encompasses the continuum of both occurring and (simply) probable events (cf. Vogl 1998). The use of statistical surveys for government purposes and their elimination of unknown individual factors is also reflected in contemporary detective stories. Holmes for instance summarizes a treatise by Winwood Reade (, 1872) according to which man, as an individual an unsolvable riddle, becomes a “mathematical certainty” in the aggregate (Doyle 2005: vol. 3, 330; cf. Boltanski 2012: 137, note 76; Campe 2012). Later investigative narratives such as Stanislaw Lem’s (, 1976) also draw on statistical correlations, but tend to elaborate the absurdity of detective trace-reading in the face of the “law of large numbers”: here, the solution to the case is reflected as the statistically necessary product of an infinite series of attempts. J. L. Borges also implies a multiplicity of virtual worlds in stories with an investigative plot: (, 1941) designs a book in which fictional worlds multiply with each turning point of its plot, while in (, 1942), the initial death turns out to be a coincidence that the criminal uses as an opportunity to set a virtual trap for his antagonist: upon arriving at the scene of the anticipated fourth murder, which the detective had calculated based on a series of crimes, he only discovers his own killer.

The omnipresent play with symmetries, doublings, and reflections in the fictional examples mentioned above also forms a link to a third aspect of virtuality, which relates to the optical register. Mirror images open up a virtual space by pretending that mirrored objects are located behind the mirror surface (Krämer 1998: 32). In a similar sense, today’s computer screens can be understood as interactive mirrors (ibid.). Because large amounts of data cannot be processed on a purely linguistic level, data-based investigations require virtual interfaces. These can include complex query tools as well as applications for visualizing, modeling, and navigating data. Like the mirror, computers and screens can open up a new perspective on what is mirrored, which in this case lies in the interactive, possibly also synesthetic handling of data structures (Krämer 1998: 32f., our translation). If every investigator has to construct a mental model in order to examine possible connections between clues (Rothöhler 2021: 17), then computer-aided models and simulations can visualize this virtual intermediate product of criminal reverse engineering (ibid., our translation) and make it navigable. This is what Weizman and Fuller mean when they say that digital or digitized shards of evidence only become fully valid evidence in the virtual model (Weizman/Fuller 2021: 5f.). On the other hand, it must be recognized that virtual models and simulations have both a representative and an interpretative function: Visual models suggest interpretations, they emphasize and conceal differences and exclude other interpretations (Reichle/Siegel/Spelten 2008: 12).

3. Data, traces, patterns

The consideration of investigations in the light of virtuality also leads to the need for a reconfiguration of the concept of the trace. As Ginzburg pointed out, in the 19th century the fingerprint was considered the archetype of the trace, which can be understood as the material remnant of a past presence. As a result, the trace gains an indexical relationship to its cause. In the case of digitized traces, however, indexicality is lost through the conventionality and universality of binary code: binary-coded data can be read as the effect of writing processes, but no longer as an indexical trace (Grube 2007: 238f.). This applies even more to digital data, which – as can be proven by various investigations of Forensic Architecture – only become meaningful traces through complex, virtually mediated design processes. Using the fingerprint as the archetype of the trace, a comparison with digital “fingerprints” is instructive: for example, various aspects of an individual hardware and software configuration can be combined to form a “browser fingerprint”, which enables precise identification of devices and users on the Internet. When discussing “data traces” of internet usage, it is hence important to note that this term refers to a digital artifact created with huge computational effort rather than to an involuntary remnant (Reigeluth 2015).

However, the absence of indexicality in digital traces does not diminish the compelling nature of the information they provide. Rather, it is simply a different type of evidence. Whereas the indexical relation, in addition to the certainty of a cause, opens up a qualitative need for interpretation with regard to an individual case, “data trails” are often processed into statistical profiles that point to possibilities and futures. The “mathematical certainty” – to use Holmes’ term – with which such virtual traces provide information about future or past behavior remains trapped in the continuum of probabilities and cannot be related back to the individual and empirical level beyond doubt.

The claim to knowledge that is free of hypotheses and interpretation is of course no longer put forward with the verve of 15 years ago. But even if nobody wants to let the data speak for itself anymore, algorithms are supposed to recognize potentially relevant patterns, abstract hypotheses from them and in turn statistically verify them based on the material (Kitchin 2014). A similar process takes place at the level of internet search engines and generative language models, which use the internet as a “database”. Search engines index the internet and suggest websites depending on search queries and user profiles. From this point of view, the data on the Internet is ‘virtual information’ that only becomes real if you search for it, produce it and allow yourself to be surprised by it (Esposito 1998: 292, our translation). This is all the more true for so-called artificial intelligence, which no longer outputs the data it finds, but formulates its own answers – virtual knowledge per se – on the basis of the language model calculated from them.

However, virtual investigations do not cement the categorical difference between evidential and data-based knowledge. Instead, they focus on research objects that interweave the latter with the former: Whether it is the combination of literary searches for clues with “large number” problems, or the combination of data-driven research with detective narrative forms. In this way, virtual investigations reveal a fundamental transformation of the evidential paradigm in the age of digitalization. By linking the investigation back to concepts that the methodological self-reflection of the humanities has engaged with since the 1970s, it is possible to highlight both the epistemic ruptures which make virtual investigations possible and the methodological and narrative continuities with conventional investigative work.

4. Structure of the publication

Based on this concept, an interdisciplinary conference was held on May 4 and 5, 2023 as part of the CRC 1385 at University of Münster, Germany, the contributions of which are currently being revised for an anthology. In keeping with the thematic framework of the CRC, the conference focused not only on literary and artistic investigations, but also on the legal dimension of the evidential paradigm. Participants examined the semiotic status of circumstantial evidence in historical and contemporary legal practice and explored the virtualization of police work in the context of dragnet investigations and predictive policing. A focal point of the conference was an evening lecture by the director of the Frankfurter Kunstverein, Professor Franziska Nori, which took place at the Picasso Museum in Münster and included a panorama of contemporary investigative art both from the national and international art landscape. Drawing on her curatorial collaboration with Forensic Architecture, she discussed the relationship between investigative art and its socially relevant potential for exposure.

The following three thematic sections are planned for the anthology: