Research Papers

In the last ten years, various automated machine learning (AutoM ) systems have been proposed to build end-to-end machine learning (ML) pipelines with minimal human interaction. Even though such automatically synthesized ML pipelines are able to achieve a competitive performance, recent studies have shown that users do not trust models constructed by AutoML due to missing transparency of AutoML systems and missing explanations for the constructed ML pipelines. In a requirements analysis study with 36 domain experts, data scientists, and AutoML researchers from different professions with vastly different expertise in ML, we collect detailed informational needs for AutoML. We propose XAutoML, an interactive visual analytics tool for explaining arbitrary AutoML optimization procedures and ML pipelines constructed by AutoML. XAutoML combines interactive visualizations with established techniques from explainable artificial intelligence (XAI) to make the complete AutoML procedure transparent and explainable. By integrating XAutoML with JupyterLab, experienced users can extend the visual analytics with ad-hoc visualizations based on information extracted from XAutoML. We validate our approach in a user study with the same diverse user group from the requirements analysis. All participants were able to extract useful information from XAutoML, leading to a significantly increased understanding of ML pipelines produced by AutoML and the AutoML optimization itself.

In this paper, we argue that AI ethics must move beyond the concepts of race-based representation and bias, and towards those that probe the deeper relations that impact how these systems are designed, developed, and deployed. Many recent discussions on ethical considerations of bias in AI systems have centered on racial bias. We contend that antiblackness in AI requires more of an examination of the ontological space that provides a foundation for the design, development, and deployment of AI systems. We examine what this contention means from the perspective of the sociocultural context in which AI systems are designed, developed, and deployed and focus on intersections with anti-Black racism (antiblackness). To bring these multiple perspectives together and show an example of antiblackness in the face of attempts at de-biasing, we discuss results from auditing an existing open-source semantic network (ConceptNet). We use this discussion to further contextualize antiblackness in design, development, and deployment of AI systems and suggest questions one may ask when attempting to combat antiblackness in AI systems.

Machine learning algorithms are useful for various predictions tasks, but they can also learn how to discriminate, based on gender, race or other sensitive attributes. This realization gave rise to the field of fair machine learning, which aims to measure and mitigate such algorithmic bias. This manuscript describes the R-package fairadapt, which implements a causal inference pre-processing method. By making use of a causal graphical model and the observed data, the method can be used to address hypothetical questions of the form "What would my salary have been, had I been of a different gender/race?". Such individual level counterfactual reasoning can help eliminate discrimination and help justify fair decisions. We also discuss appropriate relaxations which assume certain causal pathways from the sensitive attribute to the outcome are not discriminatory.

State-of-the-art entity matching (EM) methods are hard to interpret, and there is significant value in bringing explainable AI to EM. Unfortunately, most popular explainability methods do not work well out of the box for EM and need adaptation. In this paper, we identify three challenges of applying local post hoc feature attribution methods to entity matching: cross-record interaction effects, non-match explanations, and variation in sensitivity. We propose our novel model-agnostic and schema-flexible method LEMON that addresses all three challenges by (i) producing dual explanations to avoid cross-record interaction effects, (ii) introducing the novel concept of attribution potential to explain how two records could have matched, and (iii) automatically choosing explanation granularity to match the sensitivity of the matcher and record pair in question. Experiments on public datasets demonstrate that the proposed method is more faithful to the matcher and does a better job of helping users understand the decision boundary of the matcher than previous work. Furthermore, user studies show that the rate at which human subjects can construct counterfactual examples after seeing an explanation from our proposed method increases from 54% to 64% for matches and from 15% to 49% for non-matches compared to explanations from a standard adaptation of LIME.

It is widely accepted that technology is ubiquitous across the planet and has the potential to solve many of the problems existing in the Global South. Moreover, the rapid advancement of artificial intelligence (AI) brings with it the potential to address many of the challenges outlined in the Sustainable Development Goals (SDGs) in ways which were never before possible. However, there are many questions about how such advanced technologies should be managed and governed, and whether or not the emerging ethical frameworks and standards for AI are dominated by the Global North. This research examines the growing body of documentation on AI ethics to examine whether or not there is equality of participation in the ongoing global discourse. Specifically, it seeks to discover if both countries in the Global South and women are underrepresented in this discourse. Findings indicate a dearth of references to both of these themes in the AI ethics documents, suggesting that the associated ethical implications and risks are being neglected. Without adequate input from both countries in the Global South and from women, such ethical frameworks and standards may be discriminatory with the potential to reinforce marginalisation.

CGEMs: A Metric Model for Automatic Code Generation using GPT-3

Aishwarya Narasimhan, Krishna Prasad Agara Venkatesha Rao, Veena M B

Today, AI technology is showing its strengths in almost every industry and walks of life. From text generation, text summarization, chatbots, NLP is being used widely. One such paradigm is automatic code generation. An AI could be generating anything; hence the output space is unconstrained. A self-driving car is driven for 100 million miles to validate its safety, but tests cannot be written to monitor and cover an unconstrained space. One of the solutions to validate AI-generated content is to constrain the problem and convert it from abstract to realistic, and this can be accomplished by either validating the unconstrained algorithm using theoretical proofs or by using Monte-Carlo simulation methods. In this case, we use the latter approach to test/validate a statistically significant number of samples. This hypothesis of validating the AI-generated code is the main motive of this work and to know if AI-generated code is reliable, a metric model CGEMs is proposed. This is an extremely challenging task as programs can have different logic with different naming conventions, but the metrics must capture the structure and logic of the program. This is similar to the importance grammar carries in AI-based text generation, Q&A, translations, etc. The various metrics that are garnered in this work to support the evaluation of generated code are as follows: Compilation, NL description to logic conversion, number of edits needed, some of the commonly used static-code metrics and NLP metrics. These metrics are applied to 80 codes generated using OpenAI's GPT-3. Post which a Neural network is designed for binary classification (acceptable/not acceptable quality of the generated code). The inputs to this network are the values of the features obtained from the metrics. The model achieves a classification accuracy of 76.92% and an F1 score of 55.56%. XAI is augmented for model interpretability.

With the rapid development of eXplainable Artificial Intelligence (XAI), perturbation-based XAI algorithms have become quite popular due to their effectiveness and ease of implementation. The vast majority of perturbation-based XAI techniques face the challenge of Out-of-Distribution (OoD) data -- an artifact of randomly perturbed data becoming inconsistent with the original dataset. OoD data leads to the over-confidence problem in model predictions, making the existing XAI approaches unreliable. To our best knowledge, the OoD data problem in perturbation-based XAI algorithms has not been adequately addressed in the literature. In this work, we address this OoD data problem by designing an additional module quantifying the affinity between the perturbed data and the original dataset distribution, which is integrated into the process of aggregation. Our solution is shown to be compatible with the most popular perturbation-based XAI algorithms, such as RISE, OCCLUSION, and LIME. Experiments have confirmed that our methods demonstrate a significant improvement in general cases using both computational and cognitive metrics. Especially in the case of degradation, our proposed approach demonstrates outstanding performance comparing to baselines. Besides, our solution also resolves a fundamental problem with the faithfulness indicator, a commonly used evaluation metric of XAI algorithms that appears to be sensitive to the OoD issue.

Deterministic frequency deviations (DFDs) critically affect power grid frequency quality and power system stability. A better understanding of these events is urgently needed as frequency deviations have been growing in the European grid in recent years. DFDs are partially explained by the rapid adjustment of power generation following the intervals of electricity trading, but this intuitive picture fails especially before and around noonday. In this article, we provide a detailed analysis of DFDs and their relation to external features using methods from explainable Artificial Intelligence. We establish a machine learning model that well describes the daily cycle of DFDs and elucidate key interdependencies using SHapley Additive exPlanations (SHAP). Thereby, we identify solar ramps as critical to explain patterns in the Rate of Change of Frequency (RoCoF).

Brittle AI, Causal Confusion, and Bad Mental Models: Challenges and Successes in the XAI Program

Jeff Druce, James Niehaus, Vanessa Moody +2

The advances in artificial intelligence enabled by deep learning architectures are undeniable. In several cases, deep neural network driven models have surpassed human level performance in benchmark autonomy tasks. The underlying policies for these agents, however, are not easily interpretable. In fact, given their underlying deep models, it is impossible to directly understand the mapping from observations to actions for any reasonably complex agent. Producing this supporting technology to "open the black box" of these AI systems, while not sacrificing performance, was the fundamental goal of the DARPA XAI program. In our journey through this program, we have several "big picture" takeaways: 1) Explanations need to be highly tailored to their scenario; 2) many seemingly high performing RL agents are extremely brittle and are not amendable to explanation; 3) causal models allow for rich explanations, but how to present them isn't always straightforward; and 4) human subjects conjure fantastically wrong mental models for AIs, and these models are often hard to break. This paper discusses the origins of these takeaways, provides amplifying information, and suggestions for future work.

A central issue addressed by the rapidly growing research area of eXplainable Artificial Intelligence (XAI) is to provide methods to give explanations for the behaviours of Machine Learning (ML) non-interpretable models after the training. Recently, it is becoming more and more evident that new directions to create better explanations should take into account what a good explanation is to a human user. This paper suggests taking advantage of developing an XAI framework that allows producing multiple explanations for the response of image a classification system in terms of potentially different middle-level input features. To this end, we propose an XAI framework able to construct explanations in terms of input features extracted by auto-encoders. We start from the hypothesis that some autoencoders, relying on standard data representation approaches, could extract more salient and understandable input properties, which we call here \textit{Middle-Level input Features} (MLFs), for a user with respect to raw low-level features. Furthermore, extracting different types of MLFs through different type of autoencoders, different types of explanations for the same ML system behaviour can be returned. We experimentally tested our method on two different image datasets and using three different types of MLFs. The results are encouraging. Although our novel approach was tested in the context of image classification, it can potentially be used on other data types to the extent that auto-encoders to extract humanly understandable representations can be applied.

Shapley Values (SV) are widely used in explainable AI, but their estimation and interpretation can be challenging, leading to inaccurate inferences and explanations. As a starting point, we remind an invariance principle for SV and derive the correct approach for computing the SV of categorical variables that are particularly sensitive to the encoding used. In the case of tree-based models, we introduce two estimators of Shapley Values that exploit the tree structure efficiently and are more accurate than state-of-the-art methods. Simulations and comparisons are performed with state-of-the-art algorithms and show the practical gain of our approach. Finally, we discuss the limitations of Shapley Values as a local explanation. These methods are available as a Python package.

After the tremendous advances of deep learning and other AI methods, more attention is flowing into other properties of modern approaches, such as interpretability, fairness, etc. combined in frameworks like Responsible AI. Two research directions, namely Explainable AI and Uncertainty Quantification are becoming more and more important, but have been so far never combined and jointly explored. In this paper, I show how both research areas provide potential for combination, why more research should be done in this direction and how this would lead to an increase in trustability in AI systems.