Explainable Artificial Intelligence and Machine Learning: A reality rooted perspective Frank Emmert-Streib 1,2, Olli Yli-Harja 2, and Matthias Dehmer 3 1 Predictive Society and Data Analytics Lab, Faculty of Information Technology and Communication Sciences, Tampere University, Tampere, Finland 2 Institute of Biosciences and Medical Technology, Tampere University of Technology, Tampere, Finland 3 Institute for Intelligent Production, Faculty for Management, University of Applied Sciences Upper Austria, Steyr Campus, 4040 Steyr, Austria January 26, 2020 Abstract We are used to the availability of big data generated in nearly all fields of science as a consequence of technological progress. However, the analysis of such data possess vast challenges. One of these relates to the explainability of artificial intelligence (AI) or machine learning methods. Currently, many of such methods are non-transparent with respect to their working mechanism and for this reason are called black box models, most notably deep learning methods. However, it has been realized that this constitutes severe problems for a number of fields including the health sciences and criminal justice and arguments have been brought forward in favor of an explainable AI. In this paper, we do not assume the usual perspective presenting explainable AI as it should be, but rather we provide a discussion what explainable AI can be . The difference is that we do not present wishful thinking but reality grounded properties in relation to a scientific theory beyond physics. 1 Introduction Artificial intelligence (AI) and machine learning (ML) have achieved great successes in a number of different learning tasks including image recognition and speech processing [1-3].

A negotiation process by 2 agents e1 and e2 can be interleaved by another negotiation process between, say, e1 and e3. The interleaving may alter the resource allocation assumed at the inception of the first negotiation process. Existing proposals for argumentation-based negotiations have focused primarily on two-agent bilateral negotiations, but scarcely on the concurrency of multi-agent negotiations. To fill the gap, we present a novel argumentation theory, basing its development on abstract persuasion argumentation (which is an abstract argumentation formalism with a dynamic relation). Incorporating into it numerical information and a mechanism of handshakes among members of the dynamic relation, we show that the extended theory adapts well to concurrent multi-agent negotiations over scarce resources.

Explainable artificially intelligent (XAI) systems form part of sociotechnical systems, e.g., human+AI teams tasked with making decisions. Yet, current XAI systems are rarely evaluated by measuring the performance of human+AI teams on actual decision-making tasks. We conducted two online experiments and one in-person think-aloud study to evaluate two currently common techniques for evaluating XAI systems: (1) using proxy, artificial tasks such as how well humans predict the AI's decision from the given explanations, and (2) using subjective measures of trust and preference as predictors of actual performance. The results of our experiments demonstrate that evaluations with proxy tasks did not predict the results of the evaluations with the actual decision-making tasks. Further, the subjective measures on evaluations with actual decision-making tasks did not predict the objective performance on those same tasks. Our results suggest that by employing misleading evaluation methods, our field may be inadvertently slowing its progress toward developing human+AI teams that can reliably perform better than humans or AIs alone.

Lack of understanding of the decisions made by model-based AI systems is an important barrier for their adoption. We examine counterfactual explanations as an alternative for explaining AI decisions. The counterfactual approach defines an explanation as a set of the system's data inputs that causally drives the decision (meaning that removing them changes the decision) and is irreducible (meaning that removing any subset of the inputs in the explanation does not change the decision). We generalize previous work on counterfactual explanations, resulting in a framework that (a) is model-agnostic, (b) can address features with arbitrary data types, (c) is able explain decisions made by complex AI systems that incorporate multiple models, and (d) is scalable to large numbers of features. We also propose a heuristic procedure to find the most useful explanations depending on the context. We contrast counterfactual explanations with another alternative: methods that explain model predictions by weighting features according to their importance (e.g., SHAP, LIME). This paper presents two fundamental reasons why explaining model predictions is not the same as explaining the decisions made using those predictions, suggesting we should carefully consider whether importance-weight explanations are well-suited to explain decisions made by AI systems. Specifically, we show that (1) features that have a large importance weight for a model prediction may not actually affect the corresponding decision, and (2) importance weights are insufficient to communicate whether and how features influence system decisions. We demonstrate this using several examples, including three detailed studies using real-world data that compare the counterfactual approach with SHAP and illustrate various conditions under which counterfactual explanations explain data-driven decisions better than feature importance weights.

Tree-based machine learning models such as random forests, decision trees and gradient boosted trees are popular nonlinear predictive models, yet comparatively little attention has been paid to explaining their predictions. Here we improve the interpretability of tree-based models through three main contributions. We apply these tools to three medical machine learning problems and show how combining many high-quality local explanations allows us to represent global structure while retaining local faithfulness to the original model. These tools enable us to (1) identify high-magnitude but low-frequency nonlinear mortality risk factors in the US population, (2) highlight distinct population subgroups with shared risk characteristics, (3) identify nonlinear interaction effects among risk factors for chronic kidney disease and (4) monitor a machine learning model deployed in a hospital by identifying which features are degrading the model's performance over time. Given the popularity of tree-based machine learning models, these improvements to their interpretability have implications across a broad set of domains.

The semantics as to which set of arguments in a given argumentation graph may be acceptable (acceptability semantics) can be characterised in a few different ways. Among them, labelling-based approach allows for concise and flexible determination of acceptability statuses of arguments through assignment of a label indicating acceptance, rejection, or undecided to each argument. In this work, we contemplate a way of broadening it by accommodating may- and must- conditions for an argument to be accepted or rejected, as determined by the number(s) of rejected and accepted attacking arguments. We show that the broadened label-based semantics can be used to express more mild indeterminacy than inconsistency for acceptability judgement when, for example, it may be the case that an argument is accepted and when it may also be the case that it is rejected. We identify that finding which conditions a labelling satisfies for every argument can be an undecidable problem, which has an unfavourable implication to semantics. We propose to address this problem by enforcing a labelling to maximally respect the conditions, while keeping the rest that would necessarily cause non-termination labelled undecided.

A surge of interest in explainable AI (XAI) has led to a vast collection of algorithmic work on the topic. While many recognize the necessity to incorporate explainability features in AI systems, how to address real-world user needs for understanding AI remains an open question. By interviewing 20 UX and design practitioners working on various AI products, we seek to identify gaps between the current XAI algorithmic work and practices to create explainable AI products. To do so, we develop an algorithm-informed XAI question bank in which user needs for explainability are represented as prototypical questions users might ask about the AI, and use it as a study probe. Our work contributes insights into the design space of XAI, informs efforts to support design practices in this space, and identifies opportunities for future XAI work. We also provide an extended XAI question bank and discuss how it can be used for creating user-centered XAI.

Today, AI is being increasingly used to help human experts make decisions in high-stakes scenarios. In these scenarios, full automation is often undesirable, not only due to the significance of the outcome, but also because human experts can draw on their domain knowledge complementary to the model's to ensure task success. We refer to these scenarios as AI-assisted decision making, where the individual strengths of the human and the AI come together to optimize the joint decision outcome. A key to their success is to appropriately \textit{calibrate} human trust in the AI on a case-by-case basis; knowing when to trust or distrust the AI allows the human expert to appropriately apply their knowledge, improving decision outcomes in cases where the model is likely to perform poorly. This research conducts a case study of AI-assisted decision making in which humans and AI have comparable performance alone, and explores whether features that reveal case-specific model information can calibrate trust and improve the joint performance of the human and AI. Specifically, we study the effect of showing confidence score and local explanation for a particular prediction. Through two human experiments, we show that confidence score can help calibrate people's trust in an AI model, but trust calibration alone is not sufficient to improve AI-assisted decision making, which may also depend on whether the human can bring in enough unique knowledge to complement the AI's errors. We also highlight the problems in using local explanation for AI-assisted decision making scenarios and invite the research community to explore new approaches to explainability for calibrating human trust in AI.

June marked the first anniversary of Google's AI Principles, which formally outline our pledge to explore the potential of AI in a respectful, ethical and socially beneficial way. For Google Cloud, they also serve as an ongoing commitment to our customers--the tens of thousands of businesses worldwide who rely on Google Cloud AI every day--to deliver the transformative capabilities they need to thrive while aiming to help improve privacy, security, fairness, and the trust of their users. We strive to build AI aligned with our AI Principles and we're excited to introduce Explainable AI, which helps humans understand how a machine learning model reaches its conclusions. AI can unlock new ways to make businesses more efficient and create new opportunities to delight customers. That said, as with any new data-driven decision making tool, it can be a challenge to bring machine learning models into a business. Machine learning models can identify intricate correlations between enormous numbers of data points.

In the year ahead, we see the cloud, AI, and data management as the megaforces of the data and analytics agenda. And so, picking up where Big on Data bro Andrew Brust left off last week, we're looking at some of the underlying issues that are shaping adoption. In the world of data and analytics, you can't start a conversation today without bringing in cloud and AI. Yesterday in Part I, we hit the cloud checkbox: we explored how the upcoming generation change in enterprise applications will in turn shift the context of how enterprises are going to be evaluating cloud deployment. Today we turn our attention to the core building block – what's happening in databases, and what we expect to become the sleeper issue this year in AI.