The Case of AI-Based Political Theorising

Methods in contemporary political philosophy are diverse. Approaches employed by theorists include empirical methods and data, conceptual analysis, thought experiments, and intuitions. Of these methods, intuitions and mental states are particularly prominent. Floyd goes as far as to characterise the use of such methods in contemporary political philosophy as the ‘mentalist paradigm’. Despite the extent of the ‘mentalist paradigm’, intuitions are increasingly seen as being flawed. Studies inspired by cognitive psychology and the history of ideas, for example, interrogate the veracity of intuitions as the basis for political theorising. These studies suggest that intuitions are either inconsistent across individuals or unreliable. These critiques beg the question of an alternative. If we accept that intuitions are too inconsistent or unreliable, what alternative method should we use to generate political principles? Floyd’s normative behaviourism, for example, seeks to derive normative principles from the preferences expressed through behaviour. Looking at statistics on levels of insurrection and crime, for example, Floyd suggests that we can generate normative principles regarding preferred institutional arrangements. Despite the prominence of this approach normative behaviourism has been subject to a great deal of recent critical attention. The purpose of this paper is to assess an alternative and distinct method for generating organising political principles. Specifically, I follow in Floyd’s footsteps by examining an alternative to the mentalist paradigm. Rather than focus on behaviour as a source of political principles, however, I examine the potential for Artificial Intelligence (‘AI’) to form the basis of political theorising. AI has come to occupy an increasingly prominent role in contemporary society. This extends to political theory and philosophy, encompassing efforts to develop rules and principles that govern and regulate the use of AI, the effect that AI has on political institutions and the nexus between AI and theories concerning race, gender and politics. In this paper, I examine an alternative role for AI in political theory: generating political principles for organising and structuring political institutions. Just as we might ask AI models for answers to the Trolley Problem, or ask AI to write laws and policies, this approach involves, for example, asking AI models for answers to organising questions such as ‘how should we live?’ I explore three lines of reasoning concerning AI as an alternative to intuition-based political theorising. First, AI represents a more coherent way of generating political principles. In effect, AI strengthens the case for using intuitions if it has enough unbiased training data, in that it condenses and summarises our intuitions, without the cognitive biases and heuristics that we (as humans) are subject to. The second line of reasoning suggests that the extent of the biases in training data has the potential to lead to as much inconsistency or unreliability as cognitive biases in intuitions. Third, when posed questions such as ‘how should we live?’ or ‘what is the ideal political regime type?’ AI models resort to either cultural or philosophical relativism or subjectivism. As such, these models do not generate clear and determinative answers to these questions.