Ethics as a service: a pragmatic operationalisation of AI Ethics

Abstract

As the range of potential uses for Artificial Intelligence (AI), in particular machine learning (ML), has increased, so has awareness of the associated ethical issues. This increased awareness has led to the realisation that existing legislation and regulation provides insufficient protection to individuals, groups, society, and the environment from AI harms. In response to this realisation, there has been a proliferation of principle-based ethics codes, guidelines and frameworks. However, it has become increasingly clear that a significant gap exists between the theory of AI ethics principles and the practical design of AI systems. In previous work, we analysed whether it is possible to close this gap between the 'what' and the 'how' of AI ethics through the use of tools and methods designed to help AI developers, engineers, and designers translate principles into practice. We concluded that this method of closure is currently ineffective as almost all existing translational tools and methods are either too flexible (and thus vulnerable to ethics washing) or too strict (unresponsive to context). This raised the question: if, even with technical guidance, AI ethics is challenging to embed in the process of algorithmic design, is the entire pro-ethical design endeavour rendered futile? And, if no, then how can AI ethics be made useful for AI practitioners? This is the question we seek to address here by exploring why principles and technical translational tools are still needed even if they are limited, and how these limitations can be potentially overcome by providing theoretical grounding of a concept that has been termed 'Ethics as a Service.'

Guide for Artificial Intelligence Ethical Requirements Elicitation - RE4AI Ethical Guide

“…The objective of the evaluation of the Guide through a survey is to verify the viability of the guide, as well as the perceptions of users about the content provided. According to Morley et al [1], there is little evidence that the use of tools that operationalize AI ethics impacts the governability of a system. Thus, the overall aim of the evaluation of the Guide is to provide evidence that its use may have an impact on the governability of AI-based systems.…”

“…Ethical evaluation must become an integral part of the operation of a system, or there is no guarantee that tools -such as this guide -will have any positive impact on the ethical implications of AI systems [1]. There is a caution to insert on some cards throughout the deck, under "Issues to be addressed", whether the system allows for evaluation (e.g., internal, or external, and to what extent), however, it is observed the urgency of inserting a card exclusively for this purpose, in that oversight, at the evaluation stage, is "concerned with whether the algorithmic system is continuing to operate in the right way once deployed, needs to be revised, or can be improved" [1]. In other words, even after the system is deployed, the development team should define a time interval between one evaluation and another.…”

“…In other words, even after the system is deployed, the development team should define a time interval between one evaluation and another. The need for the inclusion of this card also emerges from the distribution of responsibility between the components defined by Morley et al [1] on AI ethics governance: an independent multi-disciplinary ethics board; and the AI professionals themselves. According to Morley et al [1], positive ethical qualities are susceptible to progressive increase, that is, "an algorithm can be increasingly fair, and fairer than another algorithm or a previous version, but makes no sense to say that it is fair or unfair in absolute terms".…”

“…The evolution of the emergence of software that makes use of Artificial Intelligence (AI) techniques, mostly Machine Learning (ML), amplifies the manifestations of accidents and the awareness of the associated ethical issues [1]. As an example, a ML algorithm with extensive use in hospitals in the United States had a racial bias when referring black people to better health programmes [2].…”

Some critical and ethical perspectives on the empirical turn of AI interpretability

“…In a recent paper, (Morley et al, 2021) claim that AI Ethics "tools and methods are either too flexible (and thus vulnerable to ethics washing) or too strict (unresponsive to context)". In the same lines, how to reconcile context-independent ethical principles and context-dependent empirical methods for the future development of AI?…”