Research Papers

The use of deep learning in computer vision tasks such as image classification has led to a rapid increase in the performance of such systems. Due to this substantial increment in the utility of these systems, the use of artificial intelligence in many critical tasks has exploded. In the medical domain, medical image classification systems are being adopted due to their high accuracy and near parity with human physicians in many tasks. However, these artificial intelligence systems are extremely complex and are considered black boxes by scientists, due to the difficulty in interpreting what exactly led to the predictions made by these models. When these systems are being used to assist high-stakes decision-making, it is extremely important to be able to understand, verify and justify the conclusions reached by the model. The research techniques being used to gain insight into the black-box models are in the field of explainable artificial intelligence (XAI). In this paper, we evaluated three different XAI methods across two convolutional neural network models trained to classify lung cancer from histopathological images. We visualized the outputs and analyzed the performance of these methods, in order to better understand how to apply explainable artificial intelligence in the medical domain.

Sixth-generation (6G) networks anticipate intelligently supporting a massive number of coexisting and heterogeneous slices associated with various vertical use cases. Such a context urges the adoption of artificial intelligence (AI)-driven zero-touch management and orchestration (MANO) of the end-to-end (E2E) slices under stringent service level agreements (SLAs). Specifically, the trustworthiness of the AI black-boxes in real deployment can be achieved by explainable AI (XAI) tools to build transparency between the interacting actors in the slicing ecosystem, such as tenants, infrastructure providers and operators. Inspired by the turbo principle, this paper presents a novel iterative explainable federated learning (FL) approach where a constrained resource allocation model and an \emph{explainer} exchange -- in a closed loop (CL) fashion -- soft attributions of the features as well as inference predictions to achieve a transparent and SLA-aware zero-touch service management (ZSM) of 6G network slices at RAN-Edge setup under non-independent identically distributed (non-IID) datasets. In particular, we quantitatively validate the faithfulness of the explanations via the so-called attribution-based \emph{confidence metric} that is included as a constraint in the run-time FL optimization task. In this respect, Integrated-Gradient (IG) as well as Input $\times$ Gradient and SHAP are used to generate the attributions for the turbo explainable FL (TEFL), wherefore simulation results under different methods confirm its superiority over an unconstrained Integrated-Gradient \emph{post-hoc} FL baseline.

Seeking Interpretability and Explainability in Binary Activated Neural Networks

Benjamin Leblanc, Pascal Germain

We study the use of binary activated neural networks as interpretable and explainable predictors in the context of regression tasks on tabular data; more specifically, we provide guarantees on their expressiveness, present an approach based on the efficient computation of SHAP values for quantifying the relative importance of the features, hidden neurons and even weights. As the model's simplicity is instrumental in achieving interpretability, we propose a greedy algorithm for building compact binary activated networks. This approach doesn't need to fix an architecture for the network in advance: it is built one layer at a time, one neuron at a time, leading to predictors that aren't needlessly complex for a given task.

LUCID: Exposing Algorithmic Bias through Inverse Design

Carmen Mazijn, Carina Prunkl, Andres Algaba +2

AI systems can create, propagate, support, and automate bias in decision-making processes. To mitigate biased decisions, we both need to understand the origin of the bias and define what it means for an algorithm to make fair decisions. Most group fairness notions assess a model's equality of outcome by computing statistical metrics on the outputs. We argue that these output metrics encounter intrinsic obstacles and present a complementary approach that aligns with the increasing focus on equality of treatment. By Locating Unfairness through Canonical Inverse Design (LUCID), we generate a canonical set that shows the desired inputs for a model given a preferred output. The canonical set reveals the model's internal logic and exposes potential unethical biases by repeatedly interrogating the decision-making process. We evaluate LUCID on the UCI Adult and COMPAS data sets and find that some biases detected by a canonical set differ from those of output metrics. The results show that by shifting the focus towards equality of treatment and looking into the algorithm's internal workings, the canonical sets are a valuable addition to the toolbox of algorithmic fairness evaluation.

On Interactive Explanations as Non-Monotonic Reasoning

Guilherme Paulino-Passos, Francesca Toni

Recent work shows issues of consistency with explanations, with methods generating local explanations that seem reasonable instance-wise, but that are inconsistent across instances. This suggests not only that instance-wise explanations can be unreliable, but mainly that, when interacting with a system via multiple inputs, a user may actually lose confidence in the system. To better analyse this issue, in this work we treat explanations as objects that can be subject to reasoning and present a formal model of the interactive scenario between user and system, via sequences of inputs, outputs, and explanations. We argue that explanations can be thought of as committing to some model behaviour (even if only prima facie), suggesting a form of entailment, which, we argue, should be thought of as non-monotonic. This allows: 1) to solve some considered inconsistencies in explanation, such as via a specificity relation; 2) to consider properties from the non-monotonic reasoning literature and discuss their desirability, gaining more insight on the interactive explanation scenario.

People now see social media sites as their sole source of information due to their popularity. The Majority of people get their news through social media. At the same time, fake news has grown exponentially on social media platforms in recent years. Several artificial intelligence-based solutions for detecting fake news have shown promising results. On the other hand, these detection systems lack explanation capabilities, i.e., the ability to explain why they made a prediction. This paper highlights the current state of the art in explainable fake news detection. We discuss the pitfalls in the current explainable AI-based fake news detection models and present our ongoing research on multi-modal explainable fake news detection model.

Existing methods of explainable AI and interpretable ML cannot explain change in the values of an output variable for a statistical unit in terms of the change in the input values and the change in the "mechanism" (the function transforming input to output). We propose two methods based on counterfactuals for explaining unit-level changes at various input granularities using the concept of Shapley values from game theory. These methods satisfy two key axioms desirable for any unit-level change attribution method. Through simulations, we study the reliability and the scalability of the proposed methods. We get sensible results from a case study on identifying the drivers of the change in the earnings for individuals in the US.

Explanation as Question Answering based on a Task Model of the Agent's Design

Ashok Goel, Harshvardhan Sikka, Vrinda Nandan +3

We describe a stance towards the generation of explanations in AI agents that is both human-centered and design-based. We collect questions about the working of an AI agent through participatory design by focus groups. We capture an agent's design through a Task-Method-Knowledge model that explicitly specifies the agent's tasks and goals, as well as the mechanisms, knowledge and vocabulary it uses for accomplishing the tasks. We illustrate our approach through the generation of explanations in Skillsync, an AI agent that links companies and colleges for worker upskilling and reskilling. In particular, we embed a question-answering agent called AskJill in Skillsync, where AskJill contains a TMK model of Skillsync's design. AskJill presently answers human-generated questions about Skillsync's tasks and vocabulary, and thereby helps explain how it produces its recommendations.

As we discussed in Part I of this topic, there is a clear desire to model and comprehend human behavior. Given the popular presupposition of human reasoning as the standard for learning and decision-making, there have been significant efforts and a growing trend in research to replicate these innate human abilities in artificial systems. In Part I, we discussed learning methods which generate a model of behavior from exploration of the system and feedback based on the exhibited behavior as well as topics relating to the use of or accounting for beliefs with respect to applicable skills or mental states of others. In this work, we will continue the discussion from the perspective of methods which focus on the assumed cognitive abilities, limitations, and biases demonstrated in human reasoning. We will arrange these topics as follows (i) methods such as cognitive architectures, cognitive heuristics, and related which demonstrate assumptions of limitations on cognitive resources and how that impacts decisions and (ii) methods which generate and utilize representations of bias or uncertainty to model human decision-making or the future outcomes of decisions.

The increased predictive power of machine learning models comes at the cost of increased complexity and loss of interpretability, particularly in comparison to parametric statistical models. This trade-off has led to the emergence of eXplainable AI (XAI) which provides methods, such as local explanations (LEs) and local variable attributions (LVAs), to shed light on how a model use predictors to arrive at a prediction. These provide a point estimate of the linear variable importance in the vicinity of a single observation. However, LVAs tend not to effectively handle association between predictors. To understand how the interaction between predictors affects the variable importance estimate, we can convert LVAs into linear projections and use the radial tour. This is also useful for learning how a model has made a mistake, or the effect of outliers, or the clustering of observations. The approach is illustrated with examples from categorical (penguin species, chocolate types) and quantitative (soccer/football salaries, house prices) response models. The methods are implemented in the R package cheem, available on CRAN.

Effective Representation to Capture Collaboration Behaviors between Explainer and User

Arjun Akula, Song-Chun Zhu

An explainable AI (XAI) model aims to provide transparency (in the form of justification, explanation, etc) for its predictions or actions made by it. Recently, there has been a lot of focus on building XAI models, especially to provide explanations for understanding and interpreting the predictions made by deep learning models. At UCLA, we propose a generic framework to interact with an XAI model in natural language.

Characterizing Human Explanation Strategies to Inform the Design of Explainable AI for Building Damage Assessment

Donghoon Shin, Sachin Grover, Kenneth Holstein +1

Explainable AI (XAI) is a promising means of supporting human-AI collaborations for high-stakes visual detection tasks, such as damage detection tasks from satellite imageries, as fully-automated approaches are unlikely to be perfectly safe and reliable. However, most existing XAI techniques are not informed by the understandings of task-specific needs of humans for explanations. Thus, we took a first step toward understanding what forms of XAI humans require in damage detection tasks. We conducted an online crowdsourced study to understand how people explain their own assessments, when evaluating the severity of building damage based on satellite imagery. Through the study with 60 crowdworkers, we surfaced six major strategies that humans utilize to explain their visual damage assessments. We present implications of our findings for the design of XAI methods for such visual detection contexts, and discuss opportunities for future research.