Advancements in machine learning have fuelled the popularity of using AI decision algorithms in procedures such as bail hearings, medical diagnoses and recruitment. Academic articles, policy texts, and popularizing books alike warn that such algorithms tend to be opaque: they do not provide explanations for their outcomes. Building on a causal account of transparency and opacity as well as recent work on the value of causal explanation, I formulate a moral concern for opaque algorithms that is yet to receive a (...) systematic treatment in the literature: when such algorithms are used in life-changing decisions, they can obstruct us from effectively shaping our lives according to our goals and preferences, thus undermining our autonomy. I argue that this concern deserves closer attention as it furnishes the call for transparency in algorithmic decision-making with both new tools and new challenges. (shrink)

Deep learning AI systems have proven a wide capacity to take over human-related activities such as car driving, medical diagnosing, or elderly care, often displaying behaviour with unpredictable consequences, including negative ones. This has raised the question whether highly autonomous AI may qualify as morally responsible agents. In this article, we develop a set of four conditions that an entity needs to meet in order to be ascribed moral responsibility, by drawing on Aristotelian ethics and contemporary philosophical research. We encode (...) these conditions and generate a flowchart that we call the Moral Responsibility Test. This test can be used as a tool both to evaluate whether an entity is a morally responsible agent and to inform human moral decision-making over the influencing variables of the context of action. We apply the test to the case of Artificial Moral Advisors and conclude that this form of AI cannot qualify as morally responsible agents. We further discuss the implications for the use of AMAs as moral enhancement and show that using AMAs to offload human responsibility is inadequate. We argue instead that AMAs could morally enhance users if they are interpreted as enablers for moral knowledge of the contextual variables surrounding human moral decision-making, with the implication that such a use might actually enlarge human moral responsibility. (shrink)

Automated algorithms are silently making crucial decisions about our lives, but most of the time we have little understanding of how they work. To counter this hidden influence, there have been increasing calls for algorithmic transparency. Much ink has been spilled over the informational account of algorithmic transparency—about how much information should be revealed about the inner workings of an algorithm. But few studies question the power structure beneath the informational disclosure of the algorithm. As a result, the information disclosure (...) itself can be a means of manipulation used by a group of people to advance their own interests. Instead of concentrating on information disclosure, this paper examines algorithmic transparency from the perspective of power, explaining how algorithmic transparency under a disciplinary power structure can be a technique of normalizing people’s behavior. The informational disclosure of an algorithm can not only set up some de facto norms, but also build a scientific narrative of its algorithm to justify those norms. In doing so, people would be internally motivated to follow those norms with less critical analysis. This article suggests that we should not simply open the black box of an algorithm without challenging the existing power relations. (shrink)

Institutions are increasingly employing algorithms to provide performance feedback to individuals by tracking productivity, conducting performance appraisals, and developing improvement plans, compared to traditional human managers. However, this shift has provoked considerable debate over the effectiveness and fairness of algorithmic feedback. This study investigates the effects of negative performance feedback (NPF) on the attitudes, cognition and behavior of medical researchers, comparing NPF from algorithms versus humans. Two scenario-based experimental studies were conducted with a total sample of 660 medical researchers (algorithm (...) group: N1 = 411; human group: N2 = 249). Study 1 analyzes the differences in scientific misconduct, moral disengagement, and algorithmic attitudes between the two sources of NPF. The findings reveal that NPF from algorithms shows higher levels of moral disengagement, scientific misconduct, and negative attitudes towards algorithms compared to NPF from humans. Study 2, grounded in trait activation theory, investigates how NPF from algorithms triggers individual’s egoism and algorithm aversion, potentially leading to moral disengagement and scientific misconduct. Results indicate that algorithm aversion triggers individuals’ egoism, and their interaction enhances moral disengagement, which in turn leads to increased scientific misconduct among researchers. This relationship is also moderated by algorithmic transparency. The study concludes that while algorithms can streamline performance evaluations, they pose significant risks to scientific misconduct of researchers if not properly designed. These findings extend our understanding of NPF by highlighting the emotional and cognitive challenges algorithms face in decision-making processes, while also underscoring the importance of balancing technological efficiency with moral considerations to promote a healthy research environment. Moreover, managerial implications include integrating human oversight in algorithmic NPF processes and enhancing transparency and fairness to mitigate negative impacts on medical researchers’ attitudes and behaviors. (shrink)