„By analyzing current machine learning-based AI technologies and the associated controversies, the project makes a fundamental media and social theoretical contribution to the praxeology of agentic media.“

Executive Summary

The project “Agentic Media” investigates the formation and differentiation of new media technologies composed of both software and hardware components that act (semi-)autonomously as entities capable of independent action. In its first funding phase (2020–2023), the project engaged in an analysis of the actor/subject characteristics and capacities of synthetic agents. The orders of cooperation emerging between human actors and agentic media were analyzed as configurations of semi-autonomy. Building on this, the next funding phase of the project will focus on sensor-mediated relationships of and with agentic media. The latter both form the basis of the semi-autonomous capacities for action evinced by synthetic agents and structure the interaction with these agents. Sensors enable the generation of heterogeneous data streams that continuously determine and update situations. On the one hand, this technical form of sense-making allows the performance of semi-autonomous operations and generates new forms of knowledge. On the other hand, however, interaction and cooperation with synthetic agents poses a challenge, as the knowledge and actions of these agents appear inaccessible and unpredictable to humans.

The project encompasses complementary case studies about figurations of cooperation between synthetic and human agents, sense-making based on foundation models, and the development of synthetic agents in and for scientific contexts that collectively explore the relational interconnectedness and modular composition of agentic media. The objective is to empirically examine and theoretically conceptualize the sensor-mediated constitution and specific otherness of synthetic agents. Taking production and management as examples, the project considers models and practices of human-machine cooperation, analyzing them as emergent forms of sociality with synthetic agents. The project pursues a comparative study of the development and appropriation of generative machine learning models, which equally figure as the subject of public controversy and as fundamental components of autonomization. This consideration aims both to reconstruct these debates and to analyze practices of sense-making about and with machine learning models. Building on the empirical findings elaborated in the case studies, the project makes a contribution to a media and social theory-based analysis of agentic media by exploring the challenges, conditions, and forms of cooperation between human actors and synthetic agents and describing their participation in the formation of overarching cooperation practices in such heterogeneous fields as industrial production, the cultural industries, and science itself.