The field of AI ethics during the current and previous decade is receiving an increasing amount of attention from all involved stakeholders: the public, science, philosophy, religious organizations, enterprises, governments, and various organizations. However, this field currently lacks consensus on scope, ethico-philosophical foundations, or common methodology. This thesis aims to contribute towards filling this gap by providing an answer to the two main research questions: first, what theory can explain moral scenarios in which AI entities are participants?; and second, what (...) theory can explain the process of moral reasoning, decision and action, for AI entities in virtual, simulated and real-life moral scenarios? This thesis answers these two research questions with its two main contributions to the field of AI ethics, a substantial and a methodological contribution. The substantial contribution is a coherent and novel theory named Ethics of Systems Framework, as well as a possible inception of a new field of study: ethics of systems. The methodological contribution is the creation of its main methodological tool, the Ethics of Systems Interface. The second part of the research effort was focused on testing and demonstrating the capacities of the Ethics of Systems Framework and Interface in modeling and managing moral scenarios in which AI and other entities participate. Further work can focus on building on top of the foundations of the Framework provided here, increasing the scope of moral theories and simulated scenarios, improving the level of detail and parameters to reflect real-life situations, and field-testing the Framework on actual AI systems. (shrink)

Organisations that design and deploy artificial intelligence (AI) systems increasingly commit themselves to high-level, ethical principles. However, there still exists a gap between principles and practices in AI ethics. One major obstacle organisations face when attempting to operationalise AI Ethics is the lack of a well-defined material scope. Put differently, the question to which systems and processes AI ethics principles ought to apply remains unanswered. Of course, there exists no universally accepted definition of AI, and different systems pose different ethical (...) challenges. Nevertheless, pragmatic problem-solving demands that things should be sorted so that their grouping will promote successful actions for some specific end. In this article, we review and compare previous attempts to classify AI systems for the purpose of implementing AI governance in practice. We find that attempts to classify AI systems proposed in previous literature use one of three mental models: _the Switch_, i.e., a binary approach according to which systems either are or are not considered AI systems depending on their characteristics; _the Ladder_, i.e., a risk-based approach that classifies systems according to the ethical risks they pose; and _the Matrix_, i.e., a multi-dimensional classification of systems that take various aspects into account, such as context, input data, and decision-model. Each of these models for classifying AI systems comes with its own set of strengths and weaknesses. By conceptualising different ways of classifying AI systems into simple mental models, we hope to provide organisations that design, deploy, or regulate AI systems with the vocabulary needed to demarcate the material scope of their AI governance frameworks. (shrink)

What is nontrivial digital computation? It is the processing of discrete data through discrete state transitions in accordance with finite instructional information. The motivation for our account is that many previous attempts to answer this question are inadequate, and also that this account accords with the common intuition that digital computation is a type of information processing. We use the notion of reachability in a graph to defend this characterization in memory-based systems and underscore the importance of instructional information for (...) digital computation. We argue that our account evaluates positively against adequacy criteria for accounts of computation. (shrink)