fprs.19

2019
videoinstallation - transmedia
Video: HD, 52:26 min (loop)
148,5 x 148,5 x 240 cm

@Format (*.strk) - 2020, Vienna

in cooperation with:
Format (*.strk)

supported by:
Stadt Wien Kultur
Bundeskanzleramt Österreich
Kunstverein Format (*.strk)

The video installation fpr.19 examines misinterpretations of data-based systems as artifacts that mirror traits of technological identity and reveal exploitable vulnerabilities that yield potential for agency.

Research on countersurveillance strategies and communication networks of Hong Kong's protest movement in 2019 led to a monitoring of several telegram groups and channels. The impressive size and decentralized organization of doxing strategies on this encrypted instant messenger indicated a civilian use of face recognition systems. In a decentralized manner, numerically indexed, detailed identity profiles, of violent, unmarked police officers and alleged political opponents were compiled, published, filtered and continuously updated. The dissemination of these mugshot profiles drew international attention in mid-2019 and led to the arrest of alleged administrators, an unusually large-scale DDoS attack on the platform's servers, and later to a brief injunction and offlining of one of the channels.

In the period before and after these events, image, video, audio and text material from various of these groups and channels was crawled. The data collected in this way was then analyzed by a face recognition system and filtered for its misinterpretations. These false positive identity markers, erroneously inscribed by the face recognition system, subsequently became the database for an artificial neural network to mimic facial expression and motion. To this end, an autoencoder was trained with the visual information related to these mismapped and erupted facial landmarks to interpret a representative model. The facial patterns resulting from this process mutate continuously between the alienated, the uncanny and the familiar, as hybrids of diverse contexts.

This process of personification of structural conflicts, criticized by parts of the movement as nativistically motivated and essentially misleading, embodies how regressive patterns of interpretation can reproduce within contemporary agency. Obscured in a narrative of innovation, it is not not only the retrospective functionality of data-based models, like face recognition systems, that reactualizes pasts in the anticipation of futures; but their conceptual antecedents also reincarnate within the the logics induced by our technological capabilities and limits. Haunted by biometrics' eugenic past, their shadows can be traced across notions of ideal proportions of artistic canons back to the earliest artistic personifications that aimed to capture and comprehend external complexities in anthropomorphic representations. An imprint of technological conditions that, apart from its implication of turning every smartphone into a mobile surveillance camera, represents no singularity. In fact, recent scientific and commercial approaches that promise classifications of sexual, political, or criminal characteristics based on facial proportions echo and normalize this technologically manifested notion of dredging hybrid complexities into genetic determinants.

The artistic method aims to examine spaces of imagination and action within the fluid deviations normally evaluated as Error Type I and Error Type II. These false-positive and false-negative interpretations are not coincidentally reoccurring singularities, they are structurally inevitable implications of data-based systems and emphasize that there is no capture that is free of any interpretation and therefore free of any bias. As scattered fragments along the lines of automated decision making, they throw light on these systems' codified structures of power and definition. The affirmative aspect of false positives renders them undetectable to their evoking capture systems, making them an essential source of spurious correlations that silently obscure or oversimplify causalities, but also allow for the intentional dredging of data into correlations and meaning. As both confirmation and discrepancy between a reality and its model, between the complexity of our universe and its reduction to an imaginable, thus manageable world, they can be a source of adversarial examples; data objects that can confuse, sabotage or even harm statistical systems by disruptively altering their norm.