The Use of Generative AI to Create Artificial Political Personas

The next speaker in this IAMCR 2024 session is Nicole Stewart, whose interest is in the impact of generative AI on the propaganda of tomorrow. How might we democratise AI, and what does it mean for political systems?

In particular, what about AI systems that have been trained to appear human, and perhaps even to mimic real, historical humans: there is a long history of the preservation of real people’s traces, and these are increasingly also available in digital form; such traces can be used to train generative AI to produce content which resembles the traces of real persons.

Indeed, there are already information agents that serve as an archive or corpus of historical personas; some of these are also being used as grief bots, to enable the family and friends of deceased persons to grieve for their friends and relatives. But often the content that AI is trained on contains a considerable volume of toxic or otherwise problematic content.

It is just as possible to create AI bots that are trained to resemble politicians and promote political propaganda; this has started to happen in some contexts already – in India, for instance, a candidate campaigned alongside an AI-generated video of his politician father. A universal code of ethics on the creation and use of such AI bots is urgently needed; what does a new AI informational order look like, and how can we avoid power concentrations (due to the availability and ownership of AI), the use of ‘dark archives’ (that are intransparent and incomplete), and the challenges of epistemological and informational inequality?