Research Papers

Several policy options exist, or have been proposed, to further responsible artificial intelligence (AI) development and deployment. Institutions, including U.S. government agencies, states, professional societies, and private and public sector businesses, are well positioned to implement these policies. However, given limited resources, not all policies can or should be equally prioritized. We define and review nine suggested policies for furthering responsible AI, rank each policy on potential use and impact, and recommend prioritization relative to each institution type. We find that pre-deployment audits and assessments and post-deployment accountability are likely to have the highest impact but also the highest barriers to adoption. We recommend that U.S. government agencies and companies highly prioritize development of pre-deployment audits and assessments, while the U.S. national legislature should highly prioritize post-deployment accountability. We suggest that U.S. government agencies and professional societies should highly prioritize policies that support responsible AI research and that states should highly prioritize support of responsible AI education. We propose that companies can highly prioritize involving community stakeholders in development efforts and supporting diversity in AI development. We advise lower levels of prioritization across institutions for AI ethics statements and databases of AI technologies or incidents. We recognize that no one policy will lead to responsible AI and instead advocate for strategic policy implementation across institutions.

In healthcare, AI systems support clinicians and patients in diagnosis, treatment, and monitoring, but many systems' poor explainability remains challenging for practical application. Overcoming this barrier is the goal of explainable AI (XAI). However, an explanation can be perceived differently and, thus, not solve the black-box problem for everyone. The domain of Human-Centered AI deals with this problem by adapting AI to users. We present a user-centered persona concept to evaluate XAI and use it to investigate end-users preferences for various explanation styles and contents in a mobile health stress monitoring application. The results of our online survey show that users' demographics and personality, as well as the type of explanation, impact explanation preferences, indicating that these are essential features for XAI design. We subsumed the results in three prototypical user personas: power-, casual-, and privacy-oriented users. Our insights bring an interactive, human-centered XAI closer to practical application.

Privacy and Transparency in Graph Machine Learning: A Unified Perspective

Megha Khosla

Graph Machine Learning (GraphML), whereby classical machine learning is generalized to irregular graph domains, has enjoyed a recent renaissance, leading to a dizzying array of models and their applications in several domains. With its growing applicability to sensitive domains and regulations by governmental agencies for trustworthy AI systems, researchers have started looking into the issues of transparency and privacy of graph learning. However, these topics have been mainly investigated independently. In this position paper, we provide a unified perspective on the interplay of privacy and transparency in GraphML. In particular, we describe the challenges and possible research directions for a formal investigation of privacy-transparency tradeoffs in GraphML.

Novel data sensing and AI technologies are finding practical use in the analysis of crisis resilience, revealing the need to consider how responsible artificial intelligence (AI) practices can mitigate harmful outcomes and protect vulnerable populations. In this paper, we present a responsible AI roadmap that is embedded in the Crisis Information Management Circle. This roadmap includes six propositions to highlight and address important challenges and considerations specifically related to responsible AI for crisis resilience management. We cover a wide spectrum of interwoven challenges and considerations pertaining to the responsible collection, analysis, sharing, and use of information such as equity, fairness, biases, explainability and transparency, accountability, privacy and security, inter-organizational coordination, and public engagement. Through examining issues around AI systems for crisis resilience management, we dissect the inherent complexities of information management and decision-making in crises and highlight the urgency of responsible AI research and practice. The ideas laid out in this paper are the first attempt in establishing a roadmap for researchers, practitioners, developers, emergency managers, humanitarian organizations, and public officials to address important considerations for responsible AI pertaining to crisis resilience management.

This paper summarizes and evaluates various approaches, methods, and techniques for pursuing fairness in artificial intelligence (AI) systems. It examines the merits and shortcomings of these measures and proposes practical guidelines for defining, measuring, and preventing bias in AI. In particular, it cautions against some of the simplistic, yet common, methods for evaluating bias in AI systems, and offers more sophisticated and effective alternatives. The paper also addresses widespread controversies and confusions in the field by providing a common language among different stakeholders of high-impact AI systems. It describes various trade-offs involving AI fairness, and provides practical recommendations for balancing them. It offers techniques for evaluating the costs and benefits of fairness targets, and defines the role of human judgment in setting these targets. This paper provides discussions and guidelines for AI practitioners, organization leaders, and policymakers, as well as various links to additional materials for a more technical audience. Numerous real-world examples are provided to clarify the concepts, challenges, and recommendations from a practical perspective.

Adversarial patches are images designed to fool otherwise well-performing neural network-based computer vision models. Although these attacks were initially conceived of and studied digitally, in that the raw pixel values of the image were perturbed, recent work has demonstrated that these attacks can successfully transfer to the physical world. This can be accomplished by printing out the patch and adding it into scenes of newly captured images or video footage. In this work we further test the efficacy of adversarial patch attacks in the physical world under more challenging conditions. We consider object detection models trained on overhead imagery acquired through aerial or satellite cameras, and we test physical adversarial patches inserted into scenes of a desert environment. Our main finding is that it is far more difficult to successfully implement the adversarial patch attacks under these conditions than in the previously considered conditions. This has important implications for AI safety as the real-world threat posed by adversarial examples may be overstated.

A Human-Centric Assessment Framework for AI

Sascha Saralajew, Ammar Shaker, Zhao Xu +5

With the rise of AI systems in real-world applications comes the need for reliable and trustworthy AI. An essential aspect of this are explainable AI systems. However, there is no agreed standard on how explainable AI systems should be assessed. Inspired by the Turing test, we introduce a human-centric assessment framework where a leading domain expert accepts or rejects the solutions of an AI system and another domain expert. By comparing the acceptance rates of provided solutions, we can assess how the AI system performs compared to the domain expert, and whether the AI system's explanations (if provided) are human-understandable. This setup -- comparable to the Turing test -- can serve as a framework for a wide range of human-centric AI system assessments. We demonstrate this by presenting two instantiations: (1) an assessment that measures the classification accuracy of a system with the option to incorporate label uncertainties; (2) an assessment where the usefulness of provided explanations is determined in a human-centric manner.

Computable Artificial General Intelligence

Michael Timothy Bennett

Artificial general intelligence (AGI) may herald our extinction, according to AI safety research. Yet claims regarding AGI must rely upon mathematical formalisms -- theoretical agents we may analyse or attempt to build. AIXI appears to be the only such formalism supported by proof that its behaviour is optimal, a consequence of its use of compression as a proxy for intelligence. Unfortunately, AIXI is incomputable and claims regarding its behaviour highly subjective. We argue that this is because AIXI formalises cognition as taking place in isolation from the environment in which goals are pursued (Cartesian dualism). We propose an alternative, supported by proof and experiment, which overcomes these problems. Integrating research from cognitive science with AI, we formalise an enactive model of learning and reasoning to address the problem of subjectivity. This allows us to formulate a different proxy for intelligence, called weakness, which addresses the problem of incomputability. We prove optimal behaviour is attained when weakness is maximised. This proof is supplemented by experimental results comparing weakness and description length (the closest analogue to compression possible without reintroducing subjectivity). Weakness outperforms description length, suggesting it is a better proxy. Furthermore we show that, if cognition is enactive, then minimisation of description length is neither necessary nor sufficient to attain optimal performance, undermining the notion that compression is closely related to intelligence. However, there remain open questions regarding the implementation of scale-able AGI. In the short term, these results may be best utilised to improve the performance of existing systems. For example, our results explain why Deepmind's Apperception Engine is able to generalise effectively, and how to replicate that performance by maximising weakness.

From mean-field to complex topologies: network effects on the algorithmic bias model

Valentina Pansanella, Giulio Rossetti, Letizia Milli

Nowadays, we live in a society where people often form their opinion by accessing and discussing contents shared on social networking websites. While these platforms have fostered information access and diffusion, they represent optimal environments for the proliferation of polluted contents, which is argued to be one of the co-causes of polarization/radicalization. Moreover, recommendation algorithms - intended to enhance platform usage - are likely to augment such phenomena, generating the so called Algorithmic Bias. In this work, we study the impact that different network topologies have on the formation and evolution of opinion in the context of a recent opinion dynamic model which includes bounded confidence and algorithmic bias. Mean-field, scale-free and random topologies, as well as networks generated by the Lancichinetti-Fortunato-Radicchi benchmark, are compared in terms of opinion fragmentation/polarization and time to convergence.

Can counterfactual explanations of AI systems' predictions skew lay users' causal intuitions about the world? If so, can we correct for that?

Marko Tesic, Ulrike Hahn

Counterfactual (CF) explanations have been employed as one of the modes of explainability in explainable AI-both to increase the transparency of AI systems and to provide recourse. Cognitive science and psychology, however, have pointed out that people regularly use CFs to express causal relationships. Most AI systems are only able to capture associations or correlations in data so interpreting them as casual would not be justified. In this paper, we present two experiment (total N = 364) exploring the effects of CF explanations of AI system's predictions on lay people's causal beliefs about the real world. In Experiment 1 we found that providing CF explanations of an AI system's predictions does indeed (unjustifiably) affect people's causal beliefs regarding factors/features the AI uses and that people are more likely to view them as causal factors in the real world. Inspired by the literature on misinformation and health warning messaging, Experiment 2 tested whether we can correct for the unjustified change in causal beliefs. We found that pointing out that AI systems capture correlations and not necessarily causal relationships can attenuate the effects of CF explanations on people's causal beliefs.

Artificial intelligence (AI) systems power the world we live in. Deep neural networks (DNNs) are able to solve tasks in an ever-expanding landscape of scenarios, but our eagerness to apply these powerful models leads us to focus on their performance and deprioritises our ability to understand them. Current research in the field of explainable AI tries to bridge this gap by developing various perturbation or gradient-based explanation techniques. For images, these techniques fail to fully capture and convey the semantic information needed to elucidate why the model makes the predictions it does. In this work, we develop a new form of explanation that is radically different in nature from current explanation methods, such as Grad-CAM. Perception visualization provides a visual representation of what the DNN perceives in the input image by depicting what visual patterns the latent representation corresponds to. Visualizations are obtained through a reconstruction model that inverts the encoded features, such that the parameters and predictions of the original models are not modified. Results of our user study demonstrate that humans can better understand and predict the system's decisions when perception visualizations are available, thus easing the debugging and deployment of deep models as trusted systems.

Local Explanations for Reinforcement Learning

Ronny Luss, Amit Dhurandhar, Miao Liu

Many works in explainable AI have focused on explaining black-box classification models. Explaining deep reinforcement learning (RL) policies in a manner that could be understood by domain users has received much less attention. In this paper, we propose a novel perspective to understanding RL policies based on identifying important states from automatically learned meta-states. The key conceptual difference between our approach and many previous ones is that we form meta-states based on locality governed by the expert policy dynamics rather than based on similarity of actions, and that we do not assume any particular knowledge of the underlying topology of the state space. Theoretically, we show that our algorithm to find meta-states converges and the objective that selects important states from each meta-state is submodular leading to efficient high quality greedy selection. Experiments on four domains (four rooms, door-key, minipacman, and pong) and a carefully conducted user study illustrate that our perspective leads to better understanding of the policy. We conjecture that this is a result of our meta-states being more intuitive in that the corresponding important states are strong indicators of tractable intermediate goals that are easier for humans to interpret and follow.