Rodrigo Sanz

We examine epistemic trespassing in a particular context—interdisciplinary research—with the goal of exploring the role of LLMs in addressing the epistemic challenges that arise from such practices. We first argue that trespassing is not incidental but rather a necessary feature of interdisciplinarity due to three conditions: the flexible boundaries of research questions, the hyperspecialization of expertise, and the socio-economic constraints of academic institutions. These conditions entail that epistemic trespassing may be an unavoidable phenomenon in contemporary interdisciplinary research. While epistemic trespassing is often regarded as problematic, we develop an account of how interdisciplinary research can be conducted under enhanced epistemic credentials. We propose that LLMs function as epistemic scaffolds: LLMs are constructive—they participate in the process of knowledge acquisition—, dynamically embedded—they interact continuously throughout the research process—and non-constitutive tools—can be removed once the process is completed. We further discuss the possibility of LLMs functioning as academic experts and defend an intermediate position between novice and expert status, which we call ghost expert. LLMs possess a set of epistemic credentials grounded in the transitive character of their inferential processes, the structured organization of information, and their finely calibrated capacities. However, they lack certain features required for full expert status, giving rise to persistent and highly dynamic challenges in the contemporary AI landscape. At the intersection of these analyses, we advance our central thesis: the epistemic scaffolding role of LLMs functions as a ghost expert in interdisciplinary research, enhancing researchers’ epistemic credentials and thereby making unavoidable epistemic trespassing more epistemically reliable.

Since the automation revolution of our technological era, diverse machines or robots have gradually begun to reconfigure our lives. With this expansion, it seems that those machines are now faced with a new challenge: more autonomous decision-making involving life or death consequences. This paper explores the philosophical possibility of artificial moral agency through the following question: could a machine obtain the cognitive capacities needed to be a moral agent? In this regard, I propose to expose, under a normative-cognitive perspective, the minimum criteria through which we could recognize an artificial entity as a genuine moral entity. Although my proposal should be considered from a reasonable level of abstraction, I will critically analyze and identify how an artificial agent could integrate those cognitive features. Finally, I intend to discuss their limitations or possibilities.