Issues surrounding Generative Artificial Intelligence (GAI) have become central in political debates. However, most recent literature overlooks the interactions between political agents and GAI systems in concrete situations whereby processes of (non-)democratic decision-making occur. The proposed paper moves the first steps in addressing this research gap by answering how and under which conditions partisan actors retain agency over GAI systems. Consequently, GAI may become an ally or a barrier to processes of democratic innovation.
The paper draws on Stuart Hall’s communication model, proposing that political agents ‘decode’ GAI’s contents in one of three modes: hegemonic, negotiated, and oppositional. I have applied this framework to a pilot empirical study of tabletop simulations of intra-party decision-making with 15 regional party officers in Spain and Italy. The pilot study involves four parties per country, ranging from the radical right to the radical left. Participants work in teams: I use a cloud-based communication platform on which I embed an AI chatbot that analyses group social interactions. The teams also use AI as a tool. This allowed me to gather vast data on human-GAi interactions. The simulations regard sorting out intra-party conflicts between local branches and regional executives, advancing consensus over divisive issues, and interrogating GAI systems to design party recruitment campaigns on social media. Further, I conducted pre- and post-event surveys to measure whether participants changed their views on using GAI for intra-party decision-making. The paper will present a qualitative analysis of these interactions.
There are four preliminary findings: (1) participants in the sessions follow hegemonically the results of more mechanical tasks performed by GAI, such as summarising discussions and plans of action; (2) on average, political officers with higher levels of responsibility become more oppositional when they perceive GAI’s responses as trying to lead intra-party discussions; (3) the ability to negotiate with GAI platforms correlates with the use of the platforms outside party engagement more than levels of seniority or in different parties; (4) there are no substantial differences on how party officers of different ideologies interact with GAI systems.
Ultimately, the paper proposes a research agenda to track large-N samples of how party actors negotiate political agency with GAI systems to track under which conditions GAI can be an ally or a barrier to democratic innovations in political parties.
