Search

In the field of algorithmic fairness, many fairness criteria have been proposed. Oftentimes, their proposal is only accompanied by a loose link to ideas from moral philosophy -- which makes it difficult to understand when the proposed criteria should be used to evaluate the fairness of a decision-making system. More recently, researchers have thus retroactively tried to tie existing fairness criteria to philosophical concepts. Group fairness criteria have typically been linked to egalitarianism, a theory of distributive justice. This makes it tempting to believe that fairness criteria mathematically represent ideals of distributive justice and this is indeed how they are typically portrayed. In this paper, we will discuss why the current approach of linking algorithmic fairness and distributive justice is too simplistic and, hence, insufficient. We argue that in the context of imperfect decision-making systems -- which is what we deal with in algorithmic fairness -- we should not only care about what the ideal distribution of benefits/harms among individuals would look like but also about how deviations from said ideal are distributed. Our claim is that algorithmic fairness is concerned with unfairness in these deviations. This requires us to rethink the way in which we, as algorithmic fairness researchers, view distributive justice and use fairness criteria.

By combining the philosophical literature on statistical evidence and the interdisciplinary literature on algorithmic fairness, we revisit recent objections against classification parity in light of causal analyses of algorithmic fairness and the distinction between predictive and diagnostic evidence. We focus on trial proceedings as a black-box classification algorithm in which defendants are sorted into two groups by convicting or acquitting them. We defend a novel principle, causal equal protection, that combines classification parity with the causal approach. In the do-calculus, causal equal protection requires that individuals should not be subject to uneven risks of classification error because of their protected or socially salient characteristics. The explicit use of protected characteristics, however, may be required if it equalizes these risks., Comment: 18 pages, 7 figures

Group fairness metrics are an established way of assessing the fairness of prediction-based decision-making systems. However, these metrics are still insufficiently linked to philosophical theories, and their moral meaning is often unclear. In this paper, we propose a comprehensive framework for group fairness metrics, which links them to more theories of distributive justice. The different group fairness metrics differ in their choices about how to measure the benefit or harm of a decision for the affected individuals, and what moral claims to benefits are assumed. Our unifying framework reveals the normative choices associated with standard group fairness metrics and allows an interpretation of their moral substance. In addition, this broader view provides a structure for the expansion of standard fairness metrics that we find in the literature. This expansion allows addressing several criticisms of standard group fairness metrics, specifically: (1) they are parity-based, i.e., they demand some form of equality between groups, which may sometimes be detrimental to marginalized groups; (2) they only compare decisions across groups but not the resulting consequences for these groups; and (3) the full breadth of the distributive justice literature is not sufficiently represented.

In prediction-based decision-making systems, different perspectives can be at odds: The short-term business goals of the decision makers are often in conflict with the decision subjects' wish to be treated fairly. Balancing these two perspectives is a question of values. However, these values are often hidden in the technicalities of the implementation of the decision-making system. In this paper, we propose a framework to make these value-laden choices clearly visible. We focus on a setting in which we want to find decision rules that balance the perspective of the decision maker and of the decision subjects. We provide an approach to formalize both perspectives, i.e., to assess the utility of the decision maker and the fairness towards the decision subjects. In both cases, the idea is to elicit values from decision makers and decision subjects that are then turned into something measurable. For the fairness evaluation, we build on well-known theories of distributive justice and on the algorithmic literature to ask what a fair distribution of utility (or welfare) looks like. This allows us to derive a fairness score that we then compare to the decision maker's utility. As we focus on a setting in which we are given a trained model and have to choose a decision rule, we use the concept of Pareto efficiency to compare decision rules. Our proposed framework can both guide the implementation of a decision-making system and help with audits, as it allows us to resurface the values implemented in a decision-making system.

In this paper, we provide a moral analysis of two criteria of statistical fairness debated in the machine learning literature: 1) calibration between groups and 2) equality of false positive and false negative rates between groups. In our paper, we focus on moral arguments in support of either measure. The conflict between group calibration vs. false positive and false negative rate equality is one of the core issues in the debate about group fairness definitions among practitioners. For any thorough moral analysis, the meaning of the term fairness has to be made explicit and defined properly. For our paper, we equate fairness with (non-)discrimination, which is a legitimate understanding in the discussion about group fairness. More specifically, we equate it with prima facie wrongful discrimination in the sense this is used in Prof. Lippert-Rasmussen's treatment of this definition. In this paper, we argue that a violation of group calibration may be unfair in some cases, but not unfair in others. This is in line with claims already advanced in the literature, that algorithmic fairness should be defined in a way that is sensitive to context. The most important practical implication is that arguments based on examples in which fairness requires between-group calibration, or equality in the false-positive/false-negative rates, do no generalize. For it may be that group calibration is a fairness requirement in one case, but not in another., Comment: Changes relative to the Conference version:(a) footnote 9 and 13 revealing a hidden assumption and correcting the expected utility formula for the case in which the assumption does not hold; (b) a mathematical proof of the main thesis for the case in which the assumption is relaxed

A recent paper (Hedden 2021) has argued that most of the group fairness constraints discussed in the machine learning literature are not necessary conditions for the fairness of predictions, and hence that there are no genuine fairness metrics. This is proven by discussing a special case of a fair prediction. In our paper, we show that Hedden 's argument does not hold for the most common kind of predictions used in data science, which are about people and based on data from similar people; we call these human-group-based practices. We argue that there is a morally salient distinction between human-group-based practices and those that are based on data of only one person, which we call human-individual-based practices. Thus, what may be a necessary condition for the fairness of human-group-based practices may not be a necessary condition for the fairness of human-individual-based practices, on which Hedden 's argument is based. Accordingly, the group fairness metrics discussed in the machine learning literature may still be relevant for most applications of prediction-based decision making.

We argue that an imperfect criminal law procedure cannot be group-fair, if 'group fairness' involves ensuring the same chances of acquittal or convictions to all innocent defendants independently of their morally arbitrary features. We show mathematically that only a perfect procedure (involving no mistake), a non-deterministic one, or a degenerate one (everyone or no one is convicted) can guarantee group fairness, in the general case. Following a recent proposal, we adopt a definition of group fairness, requiring that individuals who are equal in merit ought to have the same statistical chances of obtaining advantages and disadvantages, in a way that is statistically independent of any of their feature that does not count as merit. We explain by mathematical argument that the only imperfect procedures offering an a-priori guarantee of fairness in relation to all non-merit trait are lotteries or degenerate ones (i.e., everyone or no one is convicted). To provide a more intuitive point of view, we exploit an adjustment of the well-known ROC space, in order to represent all possible procedures in our model by a schematic diagram. The argument seems to be equally valid for all human procedures, provided they are imperfect. This clearly includes algorithmic decision-making, including decisions based on statistical predictions, since in practice all statistical models are error prone., Comment: 20 pages, 2 figures

We argue that the phenomena of distributed responsibility, induced acceptance, and acceptance through ignorance constitute instances of imperfect delegation when tasks are delegated to computationally-driven systems. Imperfect delegation challenges human accountability. We hold that both direct public accountability via public transparency and indirect public accountability via transparency to auditors in public organizations can be both instrumentally ethically valuable and required as a matter of deontology from the principle of democratic self-government. We analyze the regulatory content of 16 guideline documents about the use of AI in the public sector, by mapping their requirements to those of our philosophical account of accountability, and conclude that while some guidelines refer to processes that amount to auditing, it seems that the debate would benefit from more clarity about the nature of the entitlement of auditors and the goals of auditing, also in order to develop ethically meaningful standards with respect to which different forms of auditing can be evaluated and compared., Comment: 10 pages, 4 figures, 2 tables

A crucial but often neglected aspect of algorithmic fairness is the question of how we justify enforcing a certain fairness metric from a moral perspective. When fairness metrics are proposed, they are typically argued for by highlighting their mathematical properties. Rarely are the moral assumptions beneath the metric explained. Our aim in this paper is to consider the moral aspects associated with the statistical fairness criterion of independence (statistical parity). To this end, we consider previous work, which discusses the two worldviews "What You See Is What You Get" (WYSIWYG) and "We're All Equal" (WAE) and by doing so provides some guidance for clarifying the possible assumptions in the design of algorithms. We present an extension of this work, which centers on morality. The most natural moral extension is that independence needs to be fulfilled if and only if differences in predictive features (e.g. high school grades and standardized test scores are predictive of performance at university) between socio-demographic groups are caused by unjust social disparities or measurement errors. Through two counterexamples, we demonstrate that this extension is not universally true. This means that the question of whether independence should be used or not cannot be satisfactorily answered by only considering the justness of differences in the predictive features., Comment: 11 pages, accepted to ACM FAccT 2021

Counterfactual explanations are a prominent example of post-hoc interpretability methods in the explainable Artificial Intelligence research domain. They provide individuals with alternative scenarios and a set of recommendations to achieve a sought-after machine learning model outcome. Recently, the literature has identified desiderata of counterfactual explanations, such as feasibility, actionability and sparsity that should support their applicability in real-world contexts. However, we show that the literature has neglected the problem of the time dependency of counterfactual explanations. We argue that, due to their time dependency and because of the provision of recommendations, even feasible, actionable and sparse counterfactual explanations may not be appropriate in real-world applications. This is due to the possible emergence of what we call "unfortunate counterfactual events." These events may occur due to the retraining of machine learning models whose outcomes have to be explained via counterfactual explanation. Series of unfortunate counterfactual events frustrate the efforts of those individuals who successfully implemented the recommendations of counterfactual explanations. This negatively affects people's trust in the ability of institutions to provide machine learning-supported decisions consistently. We introduce an approach to address the problem of the emergence of unfortunate counterfactual events that makes use of histories of counterfactual explanations. In the final part of the paper we propose an ethical analysis of two distinct strategies to cope with the challenge of unfortunate counterfactual events. We show that they respond to an ethically responsible imperative to preserve the trustworthiness of credit lending organizations, the decision models they employ, and the social-economic function of credit lending., Comment: 11 pages, 1 figure

In this paper, we provide a philosophical account of the value of creative systems for individuals and society. We characterize creativity in very broad philosophical terms, encompassing natural, existential, and social creative processes, such as natural evolution and entrepreneurship, and explain why creativity understood in this way is instrumental for advancing human well-being in the long term. We then explain why current mainstream AI tends to be anti-creative, which means that there are moral costs of employing this type of AI in human endeavors, although computational systems that involve creativity are on the rise. In conclusion, there is an argument for ethics to be more hospitable to creativity-enabling AI, which can also be in a trade-off with other values promoted in AI ethics, such as its explainability and accuracy., Comment: 4 pages, 1 figure, Eleventh International Conference on Computational Creativity, ICCC'20

With this paper, we aim to put an issue on the agenda of AI ethics that in our view is overlooked in the current discourse. The current discussions are dominated by topics suchas trustworthiness and bias, whereas the issue we like to focuson is counter to the debate on trustworthiness. We fear that the overuse of currently dominant AI systems that are driven by short-term objectives and optimized for avoiding error leads to a society that loses its diversity and flexibility needed for true progress. We couch our concerns in the discourse around the term anti-fragility and show with some examples what threats current methods used for decision making pose for society., Comment: 7th Swiss Conference on Data Science (accepted as Poster)

We map the recently proposed notions of algorithmic fairness to economic models of Equality of opportunity (EOP)---an extensively studied ideal of fairness in political philosophy. We formally show that through our conceptual mapping, many existing definition of algorithmic fairness, such as predictive value parity and equality of odds, can be interpreted as special cases of EOP. In this respect, our work serves as a unifying moral framework for understanding existing notions of algorithmic fairness. Most importantly, this framework allows us to explicitly spell out the moral assumptions underlying each notion of fairness, and interpret recent fairness impossibility results in a new light. Last but not least and inspired by luck egalitarian models of EOP, we propose a new family of measures for algorithmic fairness. We illustrate our proposal empirically and show that employing a measure of algorithmic (un)fairness when its underlying moral assumptions are not satisfied, can have devastating consequences for the disadvantaged group's welfare.