In recent years, the problem of addressing fairness in Machine Learning (ML) and automatic decision-making has attracted a lot of attention in the scientific communities dealing with Artificial Intelligence. A plethora of different definitions of fairness in ML have been proposed, that consider different notions of what is a "fair decision" in situations impacting individuals in the population. The precise differences, implications and "orthogonality" between these notions have not yet been fully analyzed in the literature. In this work, we try to make some order out of this zoo of definitions.

Despite the enormous success of neural networks, they are still hard to interpret and often overfit when applied to low-sample-size (LSS) datasets. To tackle these obstacles, we propose a framework for training locally sparse neural networks where the local sparsity is learned via a sample-specific gating mechanism that identifies the subset of most relevant features for each measurement. The sample-specific sparsity is predicted via a \textit{gating} network, which is trained in tandem with the \textit{prediction} network. By learning these subsets and weights of a prediction model, we obtain an interpretable neural network that can handle LSS data and can remove nuisance variables, which are irrelevant for the supervised learning task. Using both synthetic and real-world datasets, we demonstrate that our method outperforms state-of-the-art models when predicting the target function with far fewer features per instance.

Identifying academic plagiarism is a pressing problem, among others, for research institutions, publishers, and funding organizations. Detection approaches proposed so far analyze lexical, syntactical, and semantic text similarity. These approaches find copied, moderately reworded, and literally translated text. However, reliably detecting disguised plagiarism, such as strong paraphrases, sense-for-sense translations, and the reuse of non-textual content and ideas, is an open research problem. The thesis addresses this problem by proposing plagiarism detection approaches that implement a different concept: analyzing non-textual content in academic documents, specifically citations, images, and mathematical content. To validate the effectiveness of the proposed detection approaches, the thesis presents five evaluations that use real cases of academic plagiarism and exploratory searches for unknown cases. The evaluation results show that non-textual content elements contain a high degree of semantic information, are language-independent, and largely immutable to the alterations that authors typically perform to conceal plagiarism. Analyzing non-textual content complements text-based detection approaches and increases the detection effectiveness, particularly for disguised forms of academic plagiarism. To demonstrate the benefit of combining non-textual and text-based detection methods, the thesis describes the first plagiarism detection system that integrates the analysis of citation-based, image-based, math-based, and text-based document similarity. The system's user interface employs visualizations that significantly reduce the effort and time users must invest in examining content similarity.

Technological advances in information sharing have raised concerns about data protection. Privacy policies contain privacy-related requirements about how the personal data of individuals will be handled by an organization or a software system (e.g., a web service or an app). In Europe, privacy policies are subject to compliance with the General Data Protection Regulation (GDPR). A prerequisite for GDPR compliance checking is to verify whether the content of a privacy policy is complete according to the provisions of GDPR. Incomplete privacy policies might result in large fines on violating organization as well as incomplete privacy-related software specifications. Manual completeness checking is both time-consuming and error-prone. In this paper, we propose AI-based automation for the completeness checking of privacy policies. Through systematic qualitative methods, we first build two artifacts to characterize the privacy-related provisions of GDPR, namely a conceptual model and a set of completeness criteria. Then, we develop an automated solution on top of these artifacts by leveraging a combination of natural language processing and supervised machine learning. Specifically, we identify the GDPR-relevant information content in privacy policies and subsequently check them against the completeness criteria. To evaluate our approach, we collected 234 real privacy policies from the fund industry. Over a set of 48 unseen privacy policies, our approach detected 300 of the total of 334 violations of some completeness criteria correctly, while producing 23 false positives. The approach thus has a precision of 92.9% and recall of 89.8%. Compared to a baseline that applies keyword search only, our approach results in an improvement of 24.5% in precision and 38% in recall.

In this paper we present the results of our experiments in training and deploying a self-supervised retrieval-based chatbot trained with contrastive learning for assisting customer support agents. In contrast to most existing research papers in this area where the focus is on solving just one component of a deployable chatbot, we present an end-to-end set of solutions to take the reader from an unlabelled chatlogs to a deployed chatbot. This set of solutions includes creating a self-supervised dataset and a weakly labelled dataset from chatlogs, as well as a systematic approach to selecting a fixed list of canned responses. We present a hierarchical-based RNN architecture for the response selection model, chosen for its ability to cache intermediate utterance embeddings, which helped to meet deployment inference speed requirements. We compare the performance of this architecture across 3 different learning objectives: self-supervised contrastive learning, binary classification, and multi-class classification. We find that using a self-supervised contrastive learning model outperforms training the binary and multi-class classification models on a weakly labelled dataset. Our results validate that the self-supervised contrastive learning approach can be effectively used for a real-world chatbot scenario.

Automatic evaluations for natural language generation (NLG) conventionally rely on token-level or embedding-level comparisons with the text references. This is different from human language processing, for which visual imaginations often improve comprehension. In this work, we propose ImaginE, an imagination-based automatic evaluation metric for natural language generation. With the help of CLIP and DALL-E, two cross-modal models pre-trained on large-scale image-text pairs, we automatically generate an image as the embodied imagination for the text snippet and compute the imagination similarity using contextual embeddings. Experiments spanning several text generation tasks demonstrate that adding imagination with our ImaginE displays great potential in introducing multi-modal information into NLG evaluation, and improves existing automatic metrics' correlations with human similarity judgments in many circumstances.

We study fair classification in the presence of an omniscient adversary that, given an $\eta$, is allowed to choose an arbitrary $\eta$-fraction of the training samples and arbitrarily perturb their protected attributes. The motivation comes from settings in which protected attributes can be incorrect due to strategic misreporting, malicious actors, or errors in imputation; and prior approaches that make stochastic or independence assumptions on errors may not satisfy their guarantees in this adversarial setting. Our main contribution is an optimization framework to learn fair classifiers in this adversarial setting that comes with provable guarantees on accuracy and fairness. Our framework works with multiple and non-binary protected attributes, is designed for the large class of linear-fractional fairness metrics, and can also handle perturbations besides protected attributes. We prove near-tightness of our framework's guarantees for natural hypothesis classes: no algorithm can have significantly better accuracy and any algorithm with better fairness must have lower accuracy. Empirically, we evaluate the classifiers produced by our framework for statistical rate on real-world and synthetic datasets for a family of adversaries.

We study the behavior of linear discriminant functions for binary classification in the infinite-imbalance limit, where the sample size of one class grows without bound while the sample size of the other remains fixed. The coefficients of the classifier minimize an expected loss specified through a weight function. We show that for a broad class of weight functions, the intercept diverges but the rest of the coefficient vector has a finite limit under infinite imbalance, extending prior work on logistic regression. The limit depends on the left tail of the weight function, for which we distinguish three cases: bounded, asymptotically polynomial, and asymptotically exponential. The limiting coefficient vectors reflect robustness or conservatism properties in the sense that they optimize against certain worst-case alternatives. In the bounded and polynomial cases, the limit is equivalent to an implicit choice of upsampling distribution for the minority class. We apply these ideas in a credit risk setting, with particular emphasis on performance in the high-sensitivity and high-specificity regions.

An integrated clinical environment (ICE) enables the connection and coordination of the internet of medical things around the care of patients in hospitals. However, ransomware attacks and their spread on hospital infrastructures, including ICE, are rising. Often the adversaries are targeting multiple hospitals with the same ransomware attacks. These attacks are detected by using machine learning algorithms. But the challenge is devising the anti-ransomware learning mechanisms and services under the following conditions: (1) provide immunity to other hospitals if one of them got the attack, (2) hospitals are usually distributed over geographical locations, and (3) direct data sharing is avoided due to privacy concerns. In this regard, this paper presents a federated distributed integrated clinical environment, aka. FedDICE. FedDICE integrates federated learning (FL), which is privacy-preserving learning, to SDN-oriented security architecture to enable collaborative learning, detection, and mitigation of ransomware attacks. We demonstrate the importance of FedDICE in a collaborative environment with up to four hospitals and four popular ransomware families, namely WannaCry, Petya, BadRabbit, and PowerGhost. Our results find that in both IID and non-IID data setups, FedDICE achieves the centralized baseline performance that needs direct data sharing for detection. However, as a trade-off to data privacy, FedDICE observes overhead in the anti-ransomware model training, e.g., 28x for the logistic regression model. Besides, FedDICE utilizes SDN's dynamic network programmability feature to remove the infected devices in ICE.

Automated high-stake decision-making such as medical diagnosis requires models with high interpretability and reliability. As one of the interpretable and reliable models with good prediction ability, we consider Sparse High-order Interaction Model (SHIM) in this study. However, finding statistically significant high-order interactions is challenging due to the intrinsic high dimensionality of the combinatorial effects. Another problem in data-driven modeling is the effect of "cherry-picking" a.k.a. selection bias. Our main contribution is to extend the recently developed parametric programming approach for selective inference to high-order interaction models. Exhaustive search over the cherry tree (all possible interactions) can be daunting and impractical even for a small-sized problem. We introduced an efficient pruning strategy and demonstrated the computational efficiency and statistical power of the proposed method using both synthetic and real data.