My central aim in the dissertation is to defend an account of the epistemic condition of moral responsibility that distinguishes culpable ignorance from non-culpable ignorance. The view that I defend is that ignorance is culpable just when an agent flouts or ignores moral reasons that underlie her epistemic norms or obligations. This view is a quality-of-will theory of moral responsibility that emphasizes the agent’s reasons-responsiveness. It holds that only relevant epistemic obligations are those that require acts of investigation or reflection. In the dissertation, I examine extant theories of culpable ignorance and suggest that they all fall short in some important respect. Then, I propose and defend an account in which epistemic norms play a leading role. I analyze the nature of epistemic norms and their normativity, and I argue that agents who ignore or flout actional investigative norms and then act on subsequent false beliefs are connected to the wrongness of their action in a way that establishes their blameworthiness. I also argue that epistemic norms that require agents to hold certain beliefs or make certain inferences are not relevant to culpable ignorance. Finally, I explore the implications of my view for certain interesting cases of moral ignorance. I discuss ignorance that results from an agent’s social or historical circumstances, ignorance that stems from pure moral deference, and ignorance that is explained by epistemic difficulty of getting certain moral facts right. There are two striking outcomes of my research. The first is that reflection on the epistemic condition shows that one cannot think deeply about moral responsibility without also engaging issues in epistemology relating to the nature and normativity of belief, and issues in normative ethics relating to what our moral obligations actually are. The second striking outcome is that bringing these rather disparate topics together, as I attempted to do, reveals that much of our ignorance is actually non-culpable, and that many of our beliefs about the blameworthiness of ignorant agents are unwarranted.

Für die Konzeption sozio-technischer Systeme zur wissenschaftlichen Erkenntnisgenerierung ist das Wissen über die spezifischen Inhalte und Bedingungen der Arbeit einer Fachgemeinschaft essenziell. Im Kontext der Automatisierung von Wissensproduktion ist unklar, welche fachspezifischen Faktoren die Möglichkeiten einer Arbeitsteilung zwischen Mensch und Maschine beeinflussen. In der vorliegenden Arbeit wird eine grundlegende Voraussetzung für die Automatisierung von Forschungsprozessen, nämlich die Formalisierbarkeit typischer Handlungen und Wissensbestände, hinsichtlich ihrer fachspezifischen Bedingungen untersucht. Dafür wurde ein qualitativer Vergleich der Evidenzkonstruktion zweier Fachgebiete, der Editionsphilologie und der Klimaforschung, durchgeführt. Um deren Forschungsprozesse systematisch vergleichen und Zusammenhänge zwischen den Eigenschaften eines Forschungsprozesses und den Möglichkeiten seiner Formalisierbarkeit empirisch untersuchen zu können, wurde ein Vergleichsrahmen entwickelt, der auf dem wissenschaftssoziologischen Konzept der epistemischen Bedingungen beruht. Die fachspezifischen Bedingungen des Forschungshandelns stellen einen Erklärungsansatz für Varianten der Wissensproduktion und damit auch für unterschiedlich formalisierbare Forschungsprozesse dar. Es konnte gezeigt werden, dass insbesondere der Grad an Kodifizierung des Wissens einen wesentlichen Einfluss auf das Auftreten bzw. die Abfolge unterschiedlich formalisierbarer Handlungstypen und Wissensbestände hat. Der Anteil persönlicher Perspektive in der Evidenzkonstruktion und der Grad der Zerlegbarkeit eines Forschungsprozesses sind ebenfalls wichtige Faktoren für die Möglichkeiten der Delegation von Handlungen an Maschinen. Desweiteren konnte gezeigt werden, dass selbst bei einem hohen Formalisierungsgrad das informelle menschliche Handeln das wesentliche Komplement automatisierter Abläufe darstellt und dass die Formalisierbarkeit einer zeitlichen Dynamik unterliegt. / Knowing about the field-specific content and conditions of work in a scientific discipline is essential for the design of socio-technical systems used for the production of scientific knowledge. In the context of automated knowledge production, it remains unclear which field-specific factors influence the possibilities to distribute labour between humans and machines. This study analyses a fundamental prerequisite for the automation of research processes, namely the possibility to formalise typical actions and knowledge, with regard to its field-specific conditions. A qualitative approach is used to compare the construction of evidence in two scientific fields, textual studies and climate research. In order to systematically compare research processes and to empirically investigate correlations between the properties of a research process and the possibilities of its formalisation, a comparative framework based on the sociological concept of epistemic conditions was developed. Field-specific conditions of doing research represent an explanatory approach for variants of knowledge production and thus for variant degrees of formalised processes. Results show that the degree of codification of knowledge has a significant influence on the occurrence and on the sequence of types of action and of knowledge resources with variant degrees of formalisation. In addition, the role of personal interpretation in problem formulation and construction of empirical evidence and the degree of decomposability of a research process are decisive factors for being able to delegate actions to machines. Furthermore, the study shows that a high degree of formalisation requires informal human action to complement automated processes and that formalisability is subject to temporal dynamics during research processes.