Recommending the best course of action for an individual is a major application of individual-level causal effect estimation. This application is often needed in safety-critical domains such as healthcare, where estimating and communicating uncertainty to decision-makers is crucial. We introduce a practical approach for integrating uncertainty estimation into a class of state-of-the-art neural network methods used for individual-level causal estimates. We show that our methods enable us to deal gracefully with situations of "no-overlap", common in high-dimensional data, where standard applications of causal effect approaches fail. Further, our methods allow us to handle covariate shift, where test distribution differs to train distribution, common when systems are deployed in practice. We show that when such a covariate shift occurs, correctly modeling uncertainty can keep us from giving overconfident and potentially harmful recommendations. We demonstrate our methodology with a range of state-of-the-art models. Under both covariate shift and lack of overlap, our uncertainty-equipped methods can alert decisions makers when predictions are not to be trusted while outperforming their uncertainty-oblivious counterparts.

Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task within a target distribution. A previous theoretical study has argued that this remarkable performance is because the meta-training protocol incentivises agents to behave Bayes-optimally. We empirically investigate this claim on a number of prediction and bandit tasks. Inspired by ideas from theoretical computer science, we show that meta-learned and Bayes-optimal agents not only behave alike, but they even share a similar computational structure, in the sense that one agent system can approximately simulate the other. Furthermore, we show that Bayes-optimal agents are fixed points of the meta-learning dynamics. Our results suggest that memory-based meta-learning might serve as a general technique for numerically approximating Bayes-optimal agents - that is, even for task distributions for which we currently don't possess tractable models.

"Most professional scientists aim to be the first to publish their findings, because it is through dissemination that the work realises its value." So wrote mathematician James Ellis in 1987. By contrast, he went on, "the fullest value of cryptography is realised by minimising the information available to potential adversaries." Ellis, like Alan Turing, and so many of the driving forces in the development of computers and the Internet, worked in government signals intelligence, or SIGINT. Today, this covers COMINT (harvested from communications such as phone calls) and ELINT (from electronic emissions, such as radar and other electromagnetic radiation).

The thesis explores the role machine learning methods play in creating intuitive computational models of neural processing. Combined with interpretability techniques, machine learning could replace human modeler and shift the focus of human effort to extracting the knowledge from the ready-made models and articulating that knowledge into intuitive descroptions of reality. This perspective makes the case in favor of the larger role that exploratory and data-driven approach to computational neuroscience could play while coexisting alongside the traditional hypothesis-driven approach. We exemplify the proposed approach in the context of the knowledge representation taxonomy with three research projects that employ interpretability techniques on top of machine learning methods at three different levels of neural organization. The first study (Chapter 3) explores feature importance analysis of a random forest decoder trained on intracerebral recordings from 100 human subjects to identify spectrotemporal signatures that characterize local neural activity during the task of visual categorization. The second study (Chapter 4) employs representation similarity analysis to compare the neural responses of the areas along the ventral stream with the activations of the layers of a deep convolutional neural network. The third study (Chapter 5) proposes a method that allows test subjects to visually explore the state representation of their neural signal in real time. This is achieved by using a topology-preserving dimensionality reduction technique that allows to transform the neural data from the multidimensional representation used by the computer into a two-dimensional representation a human can grasp. The approach, the taxonomy, and the examples, present a strong case for the applicability of machine learning methods to automatic knowledge discovery in neuroscience.

Trust is a central component of the interaction between people and AI, in that 'incorrect' levels of trust may cause misuse, abuse or disuse of the technology. But what, precisely, is the nature of trust in AI? What are the prerequisites and goals of the cognitive mechanism of trust, and how can we cause these prerequisites and goals, or assess whether they are being satisfied in a given interaction? This work aims to answer these questions. We discuss a model of trust inspired by, but not identical to, sociology's interpersonal trust (i.e., trust between people). This model rests on two key properties of the vulnerability of the user and the ability to anticipate the impact of the AI model's decisions. We incorporate a formalization of 'contractual trust', such that trust between a user and an AI is trust that some implicit or explicit contract will hold, and a formalization of 'trustworthiness' (which detaches from the notion of trustworthiness in sociology), and with it concepts of 'warranted' and 'unwarranted' trust. We then present the possible causes of warranted trust as intrinsic reasoning and extrinsic behavior, and discuss how to design trustworthy AI, how to evaluate whether trust has manifested, and whether it is warranted. Finally, we elucidate the connection between trust and XAI using our formalization.

Despite its importance to experimental design, statistical power (the probability that, given a real effect, an experiment will reject the null hypothesis) has largely been ignored by the NLP community. Underpowered experiments make it more difficult to discern the difference between statistical noise and meaningful model improvements, and increase the chances of exaggerated findings. By meta-analyzing a set of existing NLP papers and datasets, we characterize typical power for a variety of settings and conclude that underpowered experiments are common in the NLP literature. In particular, for several tasks in the popular GLUE benchmark, small test sets mean that most attempted comparisons to state of the art models will not be adequately powered. Similarly, based on reasonable assumptions, we find that the most typical experimental design for human rating studies will be underpowered to detect small model differences, of the sort that are frequently studied. For machine translation, we find that typical test sets of 2000 sentences have approximately 75% power to detect differences of 1 BLEU point. To improve the situation going forward, we give an overview of best practices for power analysis in NLP and release a series of notebooks to assist with future power analyses.

A robot that can perform colonoscopies may make the procedure simpler and less unpleasant. Pietro Valdastri at the University of Leeds in the UK and his colleagues have developed a robotic arm that uses a machine learning algorithm to move a flexible probe along the colon. The probe is a magnetic endoscope, a tube with a camera lens at the tip, that the robot controls via a magnet external to the body. The system can either work autonomously or be controlled by a human operator using a joystick, which pushes the endoscope tip further along the colon. The system also keeps track of the location and orientation of the endoscope inside the colon.

A wearable device consisting of two mini-cameras mounted on earphones can recognise your facial expressions and read your lips, even if your mouth is covered. The tool – called C-Face – was developed by Cheng Zhang at Cornell University in Ithaca, New York, and his colleagues. It looks at the sides of the wearer's head and uses machine learning to accurately visualise facial expressions by analysing small changes in cheek contour lines. "With previous technology to reconstruct facial expression, you had to put a camera in front of you. But that brings a lot of limitations," says Zhang. "Right now, many people are wearing a face mask, and standard facial tracking will not work. Our technology still works because it doesn't rely on what your face looks like."

Diagnosis of Parkinson's disease (PD) is commonly based on medical observations and assessment of clinical signs, including the characterization of a variety of motor symptoms. However, traditional diagnostic approaches may suffer from subjectivity as they rely on the evaluation of movements that are sometimes subtle to human eyes and therefore difficult to classify, leading to possible misclassification. In the meantime, early non-motor symptoms of PD may be mild and can be caused by many other conditions. Therefore, these symptoms are often overlooked, making diagnosis of PD at an early stage challenging. To address these difficulties and to refine the diagnosis and assessment procedures of PD, machine learning methods have been implemented for the classification of PD and healthy controls or patients with similar clinical presentations (e.g., movement disorders or other Parkinsonian syndromes). To provide a comprehensive overview of data modalities and machine learning methods that have been used in the diagnosis and differential diagnosis of PD, in this study, we conducted a systematic literature review of studies published until February 14, 2020, using the PubMed and IEEE Xplore databases. A total of 209 studies were included, extracted for relevant information and presented in this systematic review, with an investigation of their aims, sources of data, types of data, machine learning methods and associated outcomes. These studies demonstrate a high potential for adaptation of machine learning methods and novel biomarkers in clinical decision making, leading to increasingly systematic, informed diagnosis of PD.

Language is a valuable source of clinical information in Alzheimer's Disease, as it declines concurrently with neurodegeneration. Consequently, speech and language data have been extensively studied in connection with its diagnosis. This paper summarises current findings on the use of artificial intelligence, speech and language processing to predict cognitive decline in the context of Alzheimer's Disease, detailing current research procedures, highlighting their limitations and suggesting strategies to address them. We conducted a systematic review of original research between 2000 and 2019, registered in PROSPERO (reference CRD42018116606). An interdisciplinary search covered six databases on engineering (ACM and IEEE), psychology (PsycINFO), medicine (PubMed and Embase) and Web of Science. Bibliographies of relevant papers were screened until December 2019. From 3,654 search results 51 articles were selected against the eligibility criteria. Four tables summarise their findings: study details (aim, population, interventions, comparisons, methods and outcomes), data details (size, type, modalities, annotation, balance, availability and language of study), methodology (pre-processing, feature generation, machine learning, evaluation and results) and clinical applicability (research implications, clinical potential, risk of bias and strengths/limitations). While promising results are reported across nearly all 51 studies, very few have been implemented in clinical research or practice. We concluded that the main limitations of the field are poor standardisation, limited comparability of results, and a degree of disconnect between study aims and clinical applications. Attempts to close these gaps should support translation of future research into clinical practice.