Kangyu Wang

This thesis investigates hard choices—decision-making cases in which an agent’s options are incommensurable—and what their rise and resolution reveal about practical reason, the self, and human and non-human agency. I develop my arguments on three distinct but connected levels: the formal structure of hard choices, the mechanisms underlying them, and the practical reasons involved. In Chapter 0, I lay out these three levels and introduce my account of incommensurability. In Chapter 1, I consider which theory of practical reason can make sense of the resolution of hard choices. I critically examine reason externalism and Ruth Chang’s hybridism of reason and conclude that reason internalism is the best candidate. In Chapter 2, the core chapter of this thesis, I argue that hard choices arise from multiple permissible perspectives and that resolving a hard choice, for a human agent, means remaking oneself by moderating one’s desires. I argue that this process is rational and agentic in one sense, pre-rational and sub-agentic in another. In Chapter 3, I turn to the puzzling phenomenon that arbitrary picking is unfitting in some hard choices but not others. I reject five existing explanations and defend my own account involving Williamsian integrity and the deep-self. In Chapter 4, I pivot to the intersection between decision theory concerning incommensurability and dynamic decision theory and argue that neither Richard Pettigrew’s Aggregate Utility Solution nor similar aggregate solutions can satisfactorily accommodate incommensurability. In Chapter 5, I consider AI agents, demonstrating that under existing technological frameworks they are unable to identify or resolve hard choices, and reflect on the implications for alignment and the distinctiveness of human agency. Chapter 6 summarises the thesis and considers directions for future research.