Despite the impressive performance of Artificial Intelligence (AI) systems, their robustness remains elusive and constitutes a key issue that impedes large-scale adoption. Besides, robustness is interpreted differently across domains and contexts of AI. In this work, we systematically survey recent progress to provide a reconciled terminology of concepts around AI robustness. We introduce three taxonomies to organize and describe the literature both from a fundamental and applied point of view: (1) methods and approaches that address robustness in different phases of the machine learning pipeline; (2) methods improving robustness in specific model architectures, tasks, and systems; and in addition, (3) methodologies and insights around evaluating the robustness of AI systems, particularly the tradeoffs with other trustworthiness properties. Finally, we identify and discuss research gaps and opportunities and give an outlook on the field. We highlight the central role of humans in evaluating and enhancing AI robustness, considering the necessary knowledge they can provide, and discuss the need for better understanding practices and developing supportive tools in the future.

Human intervention is claimed to safeguard decision-subjects’ rights in algorithmic decision-making and contribute to their fairness perceptions. However, how decision-subjects perceive hybrid decision-maker configurations (i.e., combining humans and algorithms) is unclear. We address this gap through a mixed-methods study in an algorithmic policy enforcement context. Through qualitative interviews (Study 1; N1 = 21), we identify three characteristics (i.e., decision-maker’s profile, model type, input data provenance) that affect how decision-subjects perceive decision-makers’ ability, benevolence, and integrity (ABI). Through a quantitative study (Study 2; N2 = 223), we then systematically evaluate the individual and combined effects of these characteristics on decision-subjects’ perceptions towards decision-makers, and fairness perceptions. We found that only decision-maker’s profile contributes to perceived ability, benevolence, and integrity. Interestingly, the effect of decision-maker’s profile on fairness perceptions was mediated by perceived ability and integrity. Our findings have design implications for ensuring effective human intervention as a protection against harmful algorithmic decisions.

Contestability, i.e., a property that makes AI systems open to human intervention throughout their lifecycles, has been claimed to be essential for counteracting algorithmic harms. By enabling decision subjects to influence algorithmic outputs, contestable AI systems aim to safeguard decision subjects' rights to autonomy and dignity. Despite the interest and relevance of contestability in HCI, little is known about whether and how elements of contestable AI systems can empower decision subjects in algorithmic decision-making. In this dissertation, we aim to generate empirical insights into decision subjects' needs for and fairness perceptions towards contestable AI systems in decision-making. By focusing on decision subjects, this dissertation leads to a set of recommendations for organizations setting up algorithmic decision-making processes. These recommendations encourage organizations to account for the interests of those impacted by algorithmic decisions from the early stages of the design process.

Public sector organizations increasingly use artificial intelligence to augment, support, and automate decision-making. However, such public AI can potentially infringe on citizens’ right to autonomy. Contestability is a system quality that protects against this by ensuring systems are open and responsive to disputes throughout their life cycle. While a growing body of work is investigating contestable AI by design, little of this knowledge has so far been evaluated with practitioners. To make explicit the guiding ideas underpinning contestable AI research, we construct the generative metaphor of the Agonistic Arena, inspired by the political theory of agonistic pluralism. Combining this metaphor and current contestable AI guidelines, we develop an infographic supporting the early-stage concept design of public AI system contestability mechanisms. We evaluate this infographic in five workshops paired with focus groups with a total of 18 practitioners, yielding ten concept designs. Our findings outline the mechanisms for contestability derived from these concept designs. Building on these findings, we subsequently evaluate the efficacy of the Agonistic Arena as a generative metaphor for the design of public AI and identify two competing metaphors at play in this space: the Black Box and the Sovereign.

Appropriate trust, trust which aligns with system trustworthiness, in Artificial Intelligence (AI) systems has become an important area of research. However, there remains debate in the community about how to design for appropriate trust. This debate is a result of the complex nature of trust in AI, which can be difficult to understand and evaluate, as well as the lack of holistic approaches to trust. In this paper, we aim to clarify some of this debate by operationalising appropriate trust within the context of the Human-Centred AI Design (HCD) process. To do so, we organised three workshops with 13 participants total from design and development backgrounds. We carried out design activities to stimulate discussion on appropriate trust in the HCD process. This paper aims to help researchers and practitioners understand appropriate trust in AI through a design lens by illustrating how it interacts with the HCD process.

Fairness toolkits are developed to support machine learning (ML) practitioners in using algorithmic fairness metrics and mitigation methods. Past studies have investigated practical challenges for toolkit usage, which are crucial to understanding how to support practitioners. However, the extent to which fairness toolkits impact practitioners’ practices and enable reflexivity around algorithmic harms remains unclear (i.e., distributive unfairness beyond algorithmic fairness, and harms that are not related to the outputs of ML systems). Little is currently understood about the root factors that fragment practices when using fairness toolkits and how practitioners reflect on algorithmic harms. Yet, a deeper understanding of these facets is essential to enable the design of support tools for practitioners. To investigate the impact of toolkits on practices and identify factors that shape these practices, we carried out a qualitative study with 30 ML practitioners with varying backgrounds. Through a mixed within and between-subjects design, we tasked the practitioners with developing an ML model, and analyzed their reported practices to surface potential factors that lead to differences in practices. Interestingly, we found that fairness toolkits act as double-edge swords — with potentially positive and negative impacts on practices. Our findings showcase a plethora of human and organizational factors that play a key role in the way toolkits are envisioned and employed. These results bear implications for the design of future toolkits and educational training for practitioners and call for the creation of new policies to handle the organizational constraints faced by practitioners.

Recent research claims that information cues and system attributes of algorithmic decision-making processes affect decision subjects' fairness perceptions. However, little is still known about how these factors interact. This paper presents a user study (N = 267) investigating the individual and combined effects of explanations, human oversight, and contestability on informational and procedural fairness perceptions for high- and low-stakes decisions in a loan approval scenario. We find that explanations and contestability contribute to informational and procedural fairness perceptions, respectively, but we find no evidence for an effect of human oversight. Our results further show that both informational and procedural fairness perceptions contribute positively to overall fairness perceptions but we do not find an interaction effect between them. A qualitative analysis exposes tensions between information overload and understanding, human involvement and timely decision-making, and accounting for personal circumstances while maintaining procedural consistency. Our results have important design implications for algorithmic decision-making processes that meet decision subjects' standards of justice.

In an effort to regulate Machine Learning-driven (ML) systems, current auditing processes mostly focus on detecting harmful algorithmic biases. While these strategies have proven to be impactful, some values outlined in documents dealing with ethics in ML-driven systems are still underrepresented in auditing processes. Such unaddressed values mainly deal with contextual factors that cannot be easily quantified. In this paper, we develop a value-based assessment framework that is not limited to bias auditing and that covers prominent ethical principles for algorithmic systems. Our framework presents a circular arrangement of values with two bipolar dimensions that make common motivations and potential tensions explicit. In order to operationalize these high-level principles, values are then broken down into specific criteria and their manifestations. However, some of these value-specific criteria are mutually exclusive and require negotiation. As opposed to some other auditing frameworks that merely rely on ML researchers' and practitioners' input, we argue that it is necessary to include stakeholders that present diverse standpoints to systematically negotiate and consolidate value and criteria tensions. To that end, we map stakeholders with different insight needs, and assign tailored means for communicating value manifestations to them. We, therefore, contribute to current ML auditing practices with an assessment framework that visualizes closeness and tensions between values and we give guidelines on how to operationalize them, while opening up the evaluation and deliberation process to a wide range of stakeholders.