Research Papers

Using Stratified Sampling to Improve LIME Image Explanations

Muhammad Rashid, Elvio G. Amparore, Enrico Ferrari +1

We investigate the use of a stratified sampling approach for LIME Image, a popular model-agnostic explainable AI method for computer vision tasks, in order to reduce the artifacts generated by typical Monte Carlo sampling. Such artifacts are due to the undersampling of the dependent variable in the synthetic neighborhood around the image being explained, which may result in inadequate explanations due to the impossibility of fitting a linear regressor on the sampled data. We then highlight a connection with the Shapley theory, where similar arguments about undersampling and sample relevance were suggested in the past. We derive all the formulas and adjustment factors required for an unbiased stratified sampling estimator. Experiments show the efficacy of the proposed approach.

This paper explores the pressing issue of risk assessment in Large Language Models (LLMs) as they become increasingly prevalent in various applications. Focusing on how reward models, which are designed to fine-tune pretrained LLMs to align with human values, perceive and categorize different types of risks, we delve into the challenges posed by the subjective nature of preference-based training data. By utilizing the Anthropic Red-team dataset, we analyze major risk categories, including Information Hazards, Malicious Uses, and Discrimination/Hateful content. Our findings indicate that LLMs tend to consider Information Hazards less harmful, a finding confirmed by a specially developed regression model. Additionally, our analysis shows that LLMs respond less stringently to Information Hazards compared to other risks. The study further reveals a significant vulnerability of LLMs to jailbreaking attacks in Information Hazard scenarios, highlighting a critical security concern in LLM risk assessment and emphasizing the need for improved AI safety measures.

Ovarian cancer (OC) is one of the most prevalent types of cancer in women. Early and accurate diagnosis is crucial for the survival of the patients. However, the majority of women are diagnosed in advanced stages due to the lack of effective biomarkers and accurate screening tools. While previous studies sought a common biomarker, our study suggests different biomarkers for the premenopausal and postmenopausal populations. This can provide a new perspective in the search for novel predictors for the effective diagnosis of OC. Lack of explainability is one major limitation of current AI systems. The stochastic nature of the ML algorithms raises concerns about the reliability of the system as it is difficult to interpret the reasons behind the decisions. To increase the trustworthiness and accountability of the diagnostic system as well as to provide transparency and explanations behind the predictions, explainable AI has been incorporated into the ML framework. SHAP is employed to quantify the contributions of the selected biomarkers and determine the most discriminative features. A hybrid decision support system has been established that can eliminate the bottlenecks caused by the black-box nature of the ML algorithms providing a safe and trustworthy AI tool. The diagnostic accuracy obtained from the proposed system outperforms the existing methods as well as the state-of-the-art ROMA algorithm by a substantial margin which signifies its potential to be an effective tool in the differential diagnosis of OC.

Playing Large Games with Oracles and AI Debate

Xinyi Chen, Angelica Chen, Dean Foster +1

We consider regret minimization in repeated games with a very large number of actions. Such games are inherent in the setting of AI Safety via Debate \cite{irving2018ai}, and more generally games whose actions are language-based. Existing algorithms for online game playing require per-iteration computation polynomial in the number of actions, which can be prohibitive for large games. We thus consider oracle-based algorithms, as oracles naturally model access to AI agents. With oracle access, we characterize when internal and external regret can be minimized efficiently. We give a novel efficient algorithm for simultaneous external and internal regret minimization whose regret depends logarithmically on the number of actions. We conclude with experiments in the setting of AI Safety via Debate that shows the benefit of insights from our algorithmic analysis.

In the domain of Mobility Data Science, the intricate task of interpreting models trained on trajectory data, and elucidating the spatio-temporal movement of entities, has persistently posed significant challenges. Conventional XAI techniques, although brimming with potential, frequently overlook the distinct structure and nuances inherent within trajectory data. Observing this deficiency, we introduced a comprehensive framework that harmonizes pivotal XAI techniques: LIME (Local Interpretable Model-agnostic Explanations), SHAP (SHapley Additive exPlanations), Saliency maps, attention mechanisms, direct trajectory visualization, and Permutation Feature Importance (PFI). Unlike conventional strategies that deploy these methods singularly, our unified approach capitalizes on the collective efficacy of these techniques, yielding deeper and more granular insights for models reliant on trajectory data. In crafting this synthesis, we effectively address the multifaceted essence of trajectories, achieving not only amplified interpretability but also a nuanced, contextually rich comprehension of model decisions. To validate and enhance our framework, we undertook a survey to gauge preferences and reception among various user demographics. Our findings underscored a dichotomy: professionals with academic orientations, particularly those in roles like Data Scientist, IT Expert, and ML Engineer, showcased a profound, technical understanding and often exhibited a predilection for amalgamated methods for interpretability. Conversely, end-users or individuals less acquainted with AI and Data Science showcased simpler inclinations, such as bar plots indicating timestep significance or visual depictions pinpointing pivotal segments of a vessel's trajectory.

Local Concept Embeddings for Analysis of Concept Distributions in Vision DNN Feature Spaces

Georgii Mikriukov, Gesina Schwalbe, Korinna Bade

Insights into the learned latent representations are imperative for verifying deep neural networks (DNNs) in critical computer vision (CV) tasks. Therefore, state-of-the-art supervised Concept-based eXplainable Artificial Intelligence (C-XAI) methods associate user-defined concepts like ``car'' each with a single vector in the DNN latent space (concept embedding vector). In the case of concept segmentation, these linearly separate between activation map pixels belonging to a concept and those belonging to background. Existing methods for concept segmentation, however, fall short of capturing implicitly learned sub-concepts (e.g., the DNN might split car into ``proximate car'' and ``distant car''), and overlap of user-defined concepts (e.g., between ``bus'' and ``truck''). In other words, they do not capture the full distribution of concept representatives in latent space. For the first time, this work shows that these simplifications are frequently broken and that distribution information can be particularly useful for understanding DNN-learned notions of sub-concepts, concept confusion, and concept outliers. To allow exploration of learned concept distributions, we propose a novel local concept analysis framework. Instead of optimizing a single global concept vector on the complete dataset, it generates a local concept embedding (LoCE) vector for each individual sample. We use the distribution formed by LoCEs to explore the latent concept distribution by fitting Gaussian mixture models (GMMs), hierarchical clustering, and concept-level information retrieval and outlier detection. Despite its context sensitivity, our method's concept segmentation performance is competitive to global baselines. Analysis results are obtained on three datasets and six diverse vision DNN architectures, including vision transformers (ViTs).

The number of local model-agnostic explanation techniques proposed has grown rapidly recently. One main reason is that the bar for developing new explainability techniques is low due to the lack of optimal evaluation measures. Without rigorous measures, it is hard to have concrete evidence of whether the new explanation techniques can significantly outperform their predecessors. Our study proposes a new taxonomy for evaluating local explanations: robustness, evaluation using ground truth from synthetic datasets and interpretable models, model randomization, and human-grounded evaluation. Using this proposed taxonomy, we highlight that all categories of evaluation methods, except those based on the ground truth from interpretable models, suffer from a problem we call the "blame problem." In our study, we argue that this category of evaluation measure is a more reasonable method for evaluating local model-agnostic explanations. However, we show that even this category of evaluation measures has further limitations. The evaluation of local explanations remains an open research problem.

A Novel Explainable Artificial Intelligence Model in Image Classification problem

Quoc Hung Cao, Truong Thanh Hung Nguyen, Vo Thanh Khang Nguyen +1

In recent years, artificial intelligence is increasingly being applied widely in many different fields and has a profound and direct impact on human life. Following this is the need to understand the principles of the model making predictions. Since most of the current high-precision models are black boxes, neither the AI scientist nor the end-user deeply understands what's going on inside these models. Therefore, many algorithms are studied for the purpose of explaining AI models, especially those in the problem of image classification in the field of computer vision such as LIME, CAM, GradCAM. However, these algorithms still have limitations such as LIME's long execution time and CAM's confusing interpretation of concreteness and clarity. Therefore, in this paper, we propose a new method called Segmentation - Class Activation Mapping (SeCAM) that combines the advantages of these algorithms above, while at the same time overcoming their disadvantages. We tested this algorithm with various models, including ResNet50, Inception-v3, VGG16 from ImageNet Large Scale Visual Recognition Challenge (ILSVRC) data set. Outstanding results when the algorithm has met all the requirements for a specific explanation in a remarkably concise time.

AI-based text-to-image generation has undergone a significant leap in the production of visually comprehensive and aesthetic imagery over the past year, to the point where differentiating between a man-made piece of art and an AI-generated image is becoming more difficult. Generative Models such as Stable Diffusion, Midjourney and others are expected to affect several major industries in technological and ethical aspects. Striking the balance between raising human standard of life and work vs exploiting one group of people to enrich another is a complex and crucial part of the discussion. Due to the rapid growth of this technology, the way in which its models operate, and gray area legalities, visual and artistic domains - including the video game industry, are at risk of being taken over from creators by AI infrastructure owners. This paper is a literature review examining the concerns facing both AI developers and users today, including identity theft, data laundering and more. It discusses legalization challenges and ethical concerns, and concludes with how AI generative models can be tremendously useful in streamlining the process of visual creativity in both static and interactive media given proper regulation. Keywords: AI text-to-image generation, Midjourney, Stable Diffusion, AI Ethics, Game Design, Digital Art, Data Laundering

A stable supply of electrical energy is essential for the functioning of our society. Therefore, the electrical power grid's operation and energy and balancing markets are subject to strict regulations. As the external technical, economic, or social influences on the power grid change, these regulations must also be constantly adapted. However, whether these regulatory changes lead to the intended results is not easy to assess. Could eXplainable Artificial Intelligence (XAI) models distinguish regulatory settings and support the understanding of the effects of these changes? In this article, we explore two examples of regulatory changes in the German energy markets for bulk electricity and for reserve power. We explore the splitting of the German-Austrian bidding zone and changes in the pricing schemes of the German balancing energy market. We find that boosted tree models and feedforward neural networks before and after a regulatory change differ in their respective parametrizations. Using Shapley additive explanations, we reveal model differences, e.g. in terms of feature importances, and identify key features of these distinct models. With this study, we demonstrate how XAI can be applied to investigate system changes in power systems.

Thinking Upstream: Ethics and Policy Opportunities in AI Supply Chains

David Gray Widder, Richmond Wong

After children were pictured sewing its running shoes in the early 1990s, Nike at first disavowed the "working conditions in its suppliers' factories", before public pressure led them to take responsibility for ethics in their upstream supply chain. In 2023, OpenAI responded to criticism that Kenyan workers were paid less than $2 per hour to filter traumatic content from its ChatGPT model by stating in part that it had outsourced the work to a subcontractor, who managed workers' payment and mental health concerns. In this position paper, we argue that policy interventions for AI Ethics must consider AI as a supply chain problem, given how the political economy and intra-firm relations structure AI production, in particular examining opportunities upstream.

Monotonicity for AI ethics and society: An empirical study of the monotonic neural additive model in criminology, education, health care, and finance

Dangxing Chen, Luyao Zhang

Algorithm fairness in the application of artificial intelligence (AI) is essential for a better society. As the foundational axiom of social mechanisms, fairness consists of multiple facets. Although the machine learning (ML) community has focused on intersectionality as a matter of statistical parity, especially in discrimination issues, an emerging body of literature addresses another facet -- monotonicity. Based on domain expertise, monotonicity plays a vital role in numerous fairness-related areas, where violations could misguide human decisions and lead to disastrous consequences. In this paper, we first systematically evaluate the significance of applying monotonic neural additive models (MNAMs), which use a fairness-aware ML algorithm to enforce both individual and pairwise monotonicity principles, for the fairness of AI ethics and society. We have found, through a hybrid method of theoretical reasoning, simulation, and extensive empirical analysis, that considering monotonicity axioms is essential in all areas of fairness, including criminology, education, health care, and finance. Our research contributes to the interdisciplinary research at the interface of AI ethics, explainable AI (XAI), and human-computer interactions (HCIs). By evidencing the catastrophic consequences if monotonicity is not met, we address the significance of monotonicity requirements in AI applications. Furthermore, we demonstrate that MNAMs are an effective fairness-aware ML approach by imposing monotonicity restrictions integrating human intelligence.