MediaFutures Project Update | Synthetic Identity Speculations

With Synthetic Identity Speculations, media artist Michael Wallinger explores possibilities of collectively monitoring algorithmic and corporate agency in the curation of social media feeds. In collaboration with artists Kevin Daryl Ferdinandus, Alina Smith, Claudia Strate and Erico Zago they currently produce a multi channel media installation that acts as a research infrastructure as well as a hybrid interface, linking performative online investigation with contextual and methodological reflection. This is launching a long term participatory artistic research project that aims to investigate how body images affect algorithmic decision-making on content dissemination.

Integrated into large software architectures, these systems we are interested in are an essential part of the recommendation and dissemination processes on social media platforms. They scan content for violations of platform policies, copyrights, intellectual property or privacy rights, but also to identify content that is assumed to trigger user engagement. Increasingly, the agency of such filters is being extended not only to capture sensitive content, but also fact-check and flag or block misinformation and disinformation. Even though it is hard to determine from the outside how deeply they are integrated and how capable they are to fulfil these tasks, research and development towards automated holistic scene understanding continuously makes quantum leaps and can be expected to become an even bigger influence on future media landscapes.

What we can regularly observe, however, are the manifold effects of automated content curation. Repeatedly portrayed as singularities by platforms, they manifest themselves in misclassifications that silence activist, artistic, and individual content, or in the overrepresentation of motifs that promote illusory ideals with potentially extensive implications for individuals’; mental autonomy and health. While accountability is effectively shifted from platforms and algorithms to users, these technologies craft their own kind of media realities straddling the maintenance of affective economies, corporate integrity, and technological capabilities.

In order to investigate how these systems might influence contemporary and future aesthetics, choreography, iconography and authenticity in relation to body images, we are currently developing a software implementation that helps us generate and evaluate visual body representations in an automated pipeline. We built and trained several artificial neural networks that separately generate 3D-sequences of different body properties like posture, gesture, expression and body shapes while directly evaluating on consistencies between them and their dissemination processing online. This processing of synthetic 3D data allows us to directly compare the individual body properties. Not only makes this our approach highly efficient and precise, but also independent of the usage of personal data from individuals, thus making the data reusable and, in the long run, more sustainable.

Within the next weeks we will launch our first research phase via an Instagram account of a virtual Identity using the artist’s body as an avatar. In this phase we want to investigate and better understand the fundamental conditions of researching social media environments, test our implementation under different configurations and improve it in order to start investigations within the accounts of our first participants by the end of this year. We strongly believe that our approach can be of interest to a wide variety of research disciplines, such as fields from the social sciences, political science, gender studies, and youth studies, to name a few. To that end, we are simultaneously working to establish interdisciplinary collaborations with researchers and institutions.

Artistically we integrate our software architecture into a hybrid multi channel media installation, which is both the the technical infrastructure for our research as well as the interface and backend to the virtual identity on Instagram. The core of the installation and the main focus of our first research phase is the iterative body synthesis of this virtual identity. It is a long term, continuously accumulating virtual performance generated in an ongoing feedback-loop communication between the installation and Instagram. We derived this approach from state of the art methodologies in neural reverse engineering of black box functions. By gradually building up more and more informative queries to the algorithmically unique maintained user environments, we aim to identify consistent dissemination behaviour. This also means that with each new, daily shared sequence, Instagram’s affective economies, integrity interests and hyperrealistic ideals are inscribed more and more deeply into the virtual body’s appearance and its agency.

This visual element is contextualised within the installation with data visualisations giving, insights into our evaluation and selection processes as well as into intermediate results. We integrate another contextualising dimension in the form of an essayistic video documentation that elaborates on the research motivations and methodologies as well as on the biases and limitations inscribed into our architecture and more generally into statistical body representations. We hereby draw lines to idealised and invisible bodies and the potential misuse of such data yielding information on iconographic user affection.

AUTOR: Michael Wallinger – Synthetic Identity Speculations