Symposium on Forensic and Counter-Forensic Approaches to International Law: Visual Truths and Factual Realities – The Impact of Computational Photography on Legal Evidence

[Dr. Akanksha Bisoyi is a post-doctoral researcher at the Professorship of Law, Innovation and Legal Design, Technical University of Munich in Germany] 

Introduction

Photographs as legal tools for truth-telling reflect the aphorism ‘seeing is believing’. These images range from visual depictions of war crimes to human rights violations, affirming their role in the objective portrayal of historical events. Photographs are forensic evidentiary mediums that are, paradoxically, valued for their ‘true to life’ factual reality and machine-like objectivity while concomitantly scrutinized for their aptitude to mislead. However, with the upsurge of computational photography, which produces machine-readable renderings of reality, the photographs can be both revealing as well as delusive challenging the legal system’s dexterity to assess the authenticity and reliability of photographic evidence.

Therefore, while examining the implications of computational photography on established standards of evidence, this post specifically explores the role of photographs as evidence in international criminal law. The algorithmically reconstructed visualizations of atrocities challenge the traditional evidentiary standards by introducing computationally derived truths that are both technically rigorous and narratively compelling. For instance, the investigative platforms such as Cartography of a Genocide by Forensic Architecture employ satellite imagery, 3D modeling, and OSINT techniques, which underscores the epistemological shift in how such visual evidences are produced, interpreted, and legitimized. Since we live in a time when code influences our perceptivity towards reality, the post also touches upon the possibility of preserving legitimacy and accountability by design into the legal system architecture.

Photographs: Medium for Truth-telling

In 1882, the Supreme Court of Georgia viewed a photograph to be ‘a more impartial and truthful witness’ than the human memory. This view was also orchestrated in the Nuremberg trial, which was the first to admit photographic and film artefacts, in international crimes, as direct evidence of Nazi atrocities, constituting these records to ‘speak for themselves’ in ways that mere witness testimony could not. Subsequent tribunals upheld this practice, as seen in the ICTR and ICTY, where visual evidence of a massacre contributed to the conviction of Jean-Paul Akayesu for genocide, and aerial photographs depicting unlawful attacks on cultural and civilian sites were admitted as evidence, aiding in the conviction of Ðorðević for crimes against humanity. The ICC has likewise relied on visual artefacts, such as in the Lubanga case, which showed the conscription of child soldiers, and in the Al Mahdi case, which recorded the destruction of the cultural heritage of Timbuktu.

It can be said that photographs do more than depict; they have the power to establish temporal and spatial facts, corroborate testimony, and shape the picture of accountability.

Computational Photography: A New Epistemology

The shift from traditional to computational photography alters the material substrate of images and fundamentally reshapes the epistemological assumptions underpinning their production and interpretation. While traditional photography is often imagined as capturing a decisive moment in time, critical theorists have long challenged this ideal. Barthes’ claim that a photograph ‘attests to what has been’, emphasizing its evidentiary force as a certificate of presence or temporal occurrence of something or someone, and Sontag’s reflection on photography’s role in shaping our perception of reality rather than simply recording it, underscore the medium’s inherent mediation. Barad further complicates this view by emphasizing the performativity of the apparatus: images do not merely represent the world but actively participate in its construction.

Computational photography intensifies these dynamics by enabling construction of images across multiple frames, applying algorithms to enhance, stitch together, or even produce entirely synthetic scenes. This automated behaviour democratizes access to high-quality imaging but concurrently creates a gap between users and the act of image-making, shifting knowledge and agency into machine systems. The legal consequences of this shift are significant since computational artefacts push the mediation inherent in photographs to new extremes. Images can now be synthetically generated to appear photographic without any direct referent in the physical world, undermining their placeholder as ‘indexical signs’. This results in double erosion of trust, where photographic indexicality risks devolving from referentiality to sophistry, a shift from depicting ‘what is’ to persuading viewers of ‘what ought to be believed’.

Moreover, the evidentiary value of photographs is derived not solely from the image but also through narrative framing, intentional deployment, and interpretive reception. Morris emphasizes that the images are never self-evident. The infamous Abu Ghraib image, for example, implicated a soldier smiling over the corpse while obscuring the larger chain of accountability. This exemplifies that images both reveal and conceal. Computational artefacts intensify this dilemma: their ‘truth effects’ rely on algorithmic decisions and foundational data, making interpretation even more critical.  

Normative and Doctrinal Assessment

The international judicial infrastructure has been conventionally dependent on procedural safeguards to regulate visual artefactual evidence. Such safeguards assess the visual artefact’s authenticity, probative value, and chain of custody, which includes cross-examination, contemporaneity, and corroboration. However, the emergence of computational photography exposes the inadequacy of these safeguards and highlights weaknesses in existing evidentiary frameworks.

Authenticity

While metadata, such as timestamps, geotags, device information, and image hash values, can facilitate establishing provenance, they are also prone to alteration, fabrication, or damage. Since it is possible to introduce or eliminate certain elements using computational methods, the concern is not solely about the accuracy of representation of an event, but about the situated nature of the knowledge they produce. As Harraway reminds us, all knowledge, including visual knowledge, is ‘situated’, partial, and shaped by the embodied position of the knower. The image, whether traditionally or computationally generated, does not offer a view from nowhere; it enacts a particular world, shaped by the apparatus, the algorithms, and the socio-political context in which it is produced. Unless the image creation process is transparent and full disclosure about the algorithms used is available, the image has to be subjected to rigorous examination to determine its authenticity. This may necessitate the courts to increasingly demand an expert to determine and validate the truthfulness of the visual artefact, whether it has been altered or is original.

Probative Value

The ICC Ntaganda and the ICTR Karemera cases illustrate that tribunals reject photographic evidence that lacks provenance or identifiable authorship. The computational artefacts have further exacerbated these issues, leading to questions such as: If an algorithm materially alters or reconstructs an image, can it still serve to substantiate what it purports to depict? The inability of defense counsel to interrogate either the photographer or the algorithm undermines the fairness of proceedings, leaving judges with artefacts that may be visually persuasive yet epistemically fragile.

Cross-examination

In traditional trials, a photographer or witness could be questioned about how and when an image was captured. But computational artefacts raise ambiguity about authorship. Is the responsible party the photographer who pressed the shutter, or the engineers who designed the algorithm? Defense lawyers cannot meaningfully interrogate a machine-learning system, leaving a gap in accountability and adversarial testing.

Contemporaneity

This requirement necessitates that photographs should be taken close in time to the events they depict. Computational tools often blend frames, reconstruct missing details, or enhance features in ways that threaten contemporaneity. Satellite or aerial images, for instance, may be algorithmically sharpened to reveal detail while masking natural degradation or reconstruction that occurred later. The risk is that courts may attribute violence or destruction to the wrong period, undermining the causal link between the act and image.

Corroboration

Rule 63(4) of the ICC Rules of Procedure and Evidence permits reliance on photographs without corroboration; however, computational images arguably demand corroboration. As partly or wholly machine-generated constructs, computational artefacts cannot “speak for themselves” and thus, their evidentiary reliability depends on triangulation with witness testimony, documentary records, or expert interpretation of the processes that transform data into a visual presentation.

The challenge ahead is to adapt evidentiary standards to ensure that computational photography remains a tool for truth, rather than an obstacle.

Towards Legitimacy and Accountability “By Design”

To preserve trust in computational images, courts and investigators must embed legitimacy and accountability by design, which demands that evidentiary artefacts, particularly those generated through algorithmic or computational means, are structured to meet legal standards of reliability, transparency, and due process from their inception. This means that photographs should be accompanied by (i) a transparency disclosure of the methodologies used; (ii) availability of unprocessed raw image data – assessing whether the image is composite or stitched output; (iii) corroborating evidence that links the visual artefact temporally and causally to the alleged conduct; and (iv) expert analysis verifying the absence of manipulative enhancements. Without such safeguards, computational visual artefacts may present an interpretative conduct rather than a faithful record of events.

International criminal procedure already embeds safeguards for expert evidence. Rule 94 bis ICTY Rules of Procedure and Evidence, for example, emphasises on relevance, probative value, and the opportunity for cross-examination to govern admissibility. However, such safeguards may not be sufficient for computational photography, since in this case, the changes could be more than mere enhancements by introducing new layers of abstraction and bias. As a result, courts would require more than expert qualification, a full disclosure of the computational infrastructure involved, such as the model architecture, training data, algorithmic parameters, and uncertainty margins. Such a move demonstrates accountability as it encompasses not only individual experts but also the technological systems.

Conclusion

Computational photography is transforming the evidentiary landscape in international law. Its ability to enhance and reconstruct images challenges conventional norms about truth, authenticity, and probative value of visual evidence. This tension is vividly illustrated in investigative platforms such as Cartography of a Genocide by Forensic Architecture, which deploy advanced visual tools, including satellite imagery, 3D modeling, and algorithmic reconstruction, to document and analyze mass atrocities. These projects exemplify how computational images can serve not only as records but as spatial and temporal arguments, enacting events that courts must interpret and assess. While courts, on one hand, want to leverage the evidentiary power of images to document and prosecute atrocities, on the other hand, they risk either placing too much trust in technological artefacts or excluding visual evidence entirely. The work of Forensic Architecture highlights this dilemma: their visual investigations are meticulously sourced and layered, yet they also foreground the constructed nature of the evidence, inviting viewers to critically engage with the apparatus and methodology. As such, legitimacy and accountability by design demand that the computational images ought to be substantiated and verifiable so that machine-generated visual evidence would serve as instruments of justice rather than mere tools of persuasion.