3nd Nagoya Meta-Philosophy Workshop 2.0
Date: 15:00 to 17:55pm, Oct 18, 2024; 2024年10月18日(金) 15:00~17:55pm
Venue: Liberal Arts and Sciences Main Building SIS4, 4th Floor; 名古屋大学 東山キャンパス 全学教育棟中棟4階SIS4教室
Program:
15:00 to 15:55pm (including Q & A)
Rethinking the Person Affecting Approaches in Population Ethics
Makoto Suzuki (Graduate School of Letters, Nagoya Unversity)
Abstract: In population ethics, a new version of the Person Affecting Approaches (PAA) has gained prominence among influential writers. This new PAA holds that having a life of a certain quality can be better or worse for a person than nonexistence, and embraces the Wide Person Affecting Restriction (WPAR): if an outcome A is ethically better than B, A would be better than B for somebody that would exist if either A or B were to obtain. However, this approach faces several critical challenges. Firstly, it fails to uphold the Neutrality intuition central to the original PAA, which regards creating happy people as ethically neutral. Secondly, coupled with the Subjunctive Weak Pareto principle, the new PAA implies the Repugnant Conclusion: a large population with lives barely worth living could be ethically preferable to a smaller population with high-quality lives. Thirdly, while the new PAA excludes certain moral theories, this exclusion can be achieved through the simpler Mere Addition principle. Lastly, accepting that existence can be better or worse for a person than nonexistence leads to either metaphysical absurdities or semantic difficulties, weakening the intuitive appeal of WPAR. Given these issues, the new PAA lacks strong motivation, and ethicists may well need to reconsider basing their views on this framework.
16:00 to 16:55pm (including Q & A)
The Positive Role of Users in Knowledge Acquisition from Conversational AI
Masashi Kasaki (Graduate School of Informatics, Nagoya University) & Shun Tsugita (Graduate School of Humanities, Arts, and Social Sciences, Toyama University)
Abstract: We explore the epistemology of conversational AI, focusing on how users can acquire knowledge from AI's output. We propose a "prompt-responsive view" of quasi-testimonial warrant, inspired by the trust-responsive view in the epistemology of testimonial warrant: one can acquire warrant for believing testimony when one's trust motivates the testifier to tell the truth, due to the testifier valuing the good opinions of others or their commitment to social norms. The prompt-responsive view claims that a user can acquire warrant for believing AI's quasi-testimony when their prompt motivates the AI to provide a true or accurate response, due to its valuing of the good opinions of users. We support this view with empirical evidence concerning prompt engineering techniques that instructive prompts can improve the reliability, reasonableness, and comprehensibility of AI's outputs. We then compare the prompt-responsive view with the reductionist and the non-reductionist view of quasi-testimony in the epistemology of conversational AI, arguing that the prompt-responsive view better accounts for AI's variable reliability and potential lack of adherence to epistemic norms. The reliability of conversational AI varies significantly with factors such as datasets, alignments, topics of conversation, and user prompts. In addition, AI does not consistently comply with any epistemic norm. AI, however, can be motivated to provide truth or accurate responses if a user prompts it effectively. Thus, the effectiveness of user prompts is the generative source of quasi-testimonial warrant.
17:00 to 17:55pm (including Q & A)
Demarcating Conceptual Engineering from Mere Change of Belief
Jennifer Nado (Department of Philosophy, University of Hong Kong)
Abstract: Certain commonly-cited examples of conceptual engineering seem, on closer inspection, to potentially be merely cases of belief-change. One example is Carnap's case of explicating the term 'fish' in order to exclude aquatic mammals like whales - on certain externalist metasemantic views, it's plausible that 'fish' excluded whales all along. Similarly, some engineers appeal to 'marriage' as a prototype case of engineering, the thought being that we ameliorated the term 'marriage' by advocating its extension to same-sex partnerships. But did we really change the meaning of 'marriage'? Or did we simply discover that, in fact, same-sex partnerships count as marriages? Whether these cases count as conceptual engineering or mere belief-change seems to depend on one's theoretical commitments - on one's views on the targets of conceptual engineering, for instance, and potentially on one's metasemantic views or one's views on the nature of concepts. In this talk I'll explore how different accounts of conceptual engineering might deal with this 'demarcation problem'. I'll argue that many views face potentially serious difficulties drawing the distinction between engineering and discovery; certain views seem to imply unintuitive boundaries, while others seem to put the boundaries beyond our epistemic reach.