Ethical Design Patterns for Automated Decision Platforms

This article is relevant to the ethical risks posed by automated decision platforms which are often due to architectural loopholes and not only to isolated algorithmic biases. To address the risks, a new set of ethical design patterns is proposed in the paper that is specifically aimed at inserting responsibility, fairness, and human control into the automated systems at the architectural level. The above-described patterns are contestable decisions, whereby one can challenge and review automated results; proportional automation, where automated results can be applied based on context; human override pathways, whereby, there is a way of determining any negative effects; transparency triggers whereby, users and other stakeholders can know about automated decisions; and continuous harm monitoring whereby, unsatisfactory results can be evaluated on a recurring basis. This can enable engineers to always integrate ethical considerations in the development process as they translate abstract ethical principles into concrete, measurable behaviors within the system. This article restructures the concept of ethics within the context of system design as opposed to an immediate post-deployment analysis but presents a pragmatic and feasible plan of designing automated systems that value human agency and enforce organizational responsibility. The suggested ethical design patterns comprise a detailed scheme of enhancing the responsible automation in the current technological-driven decision-making procedures.