Cognitive Mimetics for AI Ethics: Tacit Knowledge, Action Ontologies and Problem Restructuring

摘要

Ethics and ethical information processing are an important problem for AI development. It is important for self-evident reasons, but also challenging in its' implications and should be welcomed by designers and developers as an interesting technical challenge. This article explores AI ethics as a design problem and lays out how cognitive mimetics could be used a method for its design. AI ethics is conceptualized as a problem of implementation on the one hand, and as a problem of ethical contents on the other. From the viewpoint of human information processing, ethics becomes a special case of ethical information processing - one that has deep implications in terms of AI abilities and information contents. Here we focus on ethical information processing as a property of the system (rather as a general constraint on it). We explore three specific concepts relevant for cognitive mimetics from the perspective of ethics: tacit knowledge, ontologies, and problem restructuring. We close with a general discussion on the difference between abilities and mental contents noted as relevant in previous articles on cognitive mimetics and reiterate its importance in this context as well.