Handling missing values is a common challenge in biostatistical analyses, typically addressed by imputation methods. We propose a novel, fast, and easy-to-use imputation method called missing value imputation with adversarial random forests (MissARF), based on generative machine learning, that provides both single and multiple imputation. MissARF employs adversarial random forest (ARF) for density estimation and data synthesis. To impute a missing value of an observation, we condition on the non-missing values and sample from the estimated conditional distribution generated by ARF. Our experiments demonstrate that MissARF performs comparably to state-of-the-art single and multiple imputation methods in terms of imputation quality and fast runtime with no additional costs for multiple imputation.

Instrumental variables (IVs) are widely used to estimate causal effects in the presence of unobserved confounding between exposure and outcome. An IV must affect the outcome exclusively through the exposure and be unconfounded with the outcome. We present a framework for relaxing either or both of these strong assumptions with tuneable and interpretable budget constraints. Our algorithm returns a feasible set of causal effects that can be identified exactly given relevant covariance parameters. The feasible set may be disconnected but is a finite union of convex subsets. We discuss conditions under which this set is sharp, i.e., contains all and only effects consistent with the background assumptions and the joint distribution of observable variables. Our method applies to a wide class of semiparametric models, and we demonstrate how its ability to select specific subsets of instruments confers an advantage over convex relaxations in both linear and nonlinear settings. We also adapt our algorithm to form confidence sets that are asymptotically valid under a common statistical assumption from the Mendelian randomization literature.

This article looks at how reasoning works in current Large Language Models (LLMs) that function using the token-completion method. It examines their stochastic nature and their similarity to human abductive reasoning. The argument is that these LLMs create text based on learned patterns rather than performing actual abductive reasoning. When their output seems abductive, this is largely because they are trained on human-generated texts that include reasoning structures. Examples are used to show how LLMs can produce plausible ideas, mimic commonsense reasoning, and give explanatory answers without being grounded in truth, semantics, verification, or understanding, and without performing any real abductive reasoning. This dual nature, where the models have a stochastic base but appear abductive in use, has important consequences for how LLMs are evaluated and applied. They can assist with generating ideas and supporting human thinking, but their outputs must be critically assessed because they cannot identify truth or verify their explanations. The article concludes by addressing five objections to these points, noting some limitations in the analysis, and offering an overall evaluation.

Several competing narratives drive the contemporary AI ethics discourse. At the two extremes are sociotechnical dogmatism, which holds that society is full of inefficiencies and imperfections that can only be solved by better technology; and sociotechnical skepticism, which highlights the unacceptable risks AI systems pose. While both narratives have their merits, they are ultimately reductive and limiting. As a constructive synthesis, we introduce and defend sociotechnical pragmatism—a narrative that emphasizes the central role of context and human agency in designing and evaluating emerging technologies. In doing so, we offer two novel contributions. First, we demonstrate how ethical and epistemological considerations are intertwined in the AI ethics discourse by tracing the dialectical interplay between dogmatic and skeptical narratives across disciplines. Second, we show through examples how sociotechnical pragmatism does more to promote fair and transparent AI than dogmatic or skeptical alternatives. By spelling out the assumptions that underpin sociotechnical pragmatism, we articulate a robust stance for policymakers and scholars who seek to enable societies to reap the benefits of AI while managing the associated risks through feasible, effective, and proportionate governance.

Instrumental variables (IVs) are a popular and powerful tool for estimating causal effects in the presence of unobserved confounding. However, classical approaches rely on strong assumptions such as the $\textit{exclusion criterion}$, which states that instrumental effects must be entirely mediated by treatments. This assumption often fails in practice. When IV methods are improperly applied to data that do not meet the exclusion criterion, estimated causal effects may be badly biased. In this work, we propose a novel solution that provides $\textit{partial}$ identification in linear systems given a set of $\textit{leaky instruments}$, which are allowed to violate the exclusion criterion to some limited degree. We derive a convex optimization objective that provides provably sharp bounds on the average treatment effect under some common forms of information leakage, and implement inference procedures to quantify the uncertainty of resulting estimates. We demonstrate our method in a set of experiments with simulated data, where it performs favorably against the state of the art. An accompanying $\texttt{R}$ package, $\texttt{leakyIV}$, is available from $\texttt{CRAN}$.

The Fairness, Accountability, and Transparency (FAccT) literature tends to focus on bias as a problem that requires ex post solutions (e.g. fairness metrics), rather than addressing the underlying social and technical conditions that (re)produce it. In this article, we propose a complementary strategy that uses genealogy as a constructive, epistemic critique to explain algorithmic bias in terms of the conditions that enable it. We focus on XAI feature attributions (Shapley values) and counterfactual approaches as potential tools to gauge these conditions and offer two main contributions. One is constructive: we develop a theoretical framework to classify these approaches according to their relevance for bias as evidence of social disparities. We draw on Pearl’s ladder of causation (Causality: models, reasoning, and inference. Cambridge University Press, Cambridge, 2000, Causality, 2nd edn. Cambridge University Press, Cambridge, 2009. https://doi.org/10.1017/CBO9780511803161 ) to order these XAI approaches concerning their ability to answer fairness-relevant questions and identify fairness-relevant solutions. The other contribution is critical: we evaluate these approaches in terms of their assumptions about the role of protected characteristics in discriminatory outcomes. We achieve this by building on Kohler-Hausmann’s (Northwest Univ Law Rev 113(5):1163–1227, 2019) constructivist theory of discrimination. We derive three recommendations for XAI practitioners to develop and AI policymakers to regulate tools that address algorithmic bias in its conditions and hence mitigate its future occurrence.

Unsupervised learning algorithms are widely used for many important statistical tasks with numerous applications in science and industry. Yet despite their prevalence, they have attracted remarkably little philosophical scrutiny to date. This stands in stark contrast to supervised and reinforcement learning algorithms, which have been widely studied and critically evaluated, often with an emphasis on ethical concerns. In this article, I analyze three canonical unsupervised learning problems: clustering, abstraction, and generative modeling. I argue that these methods raise unique epistemological and ontological questions, providing data-driven tools for discovering natural kinds and distinguishing essence from contingency. This analysis goes some way toward filling the lacuna in contemporary philosophical discourse on unsupervised learning, as well as bringing conceptual unity to a heterogeneous field more often described by what it is not (i.e., supervised or reinforcement learning) than by what it is . I submit that unsupervised learning is not just a legitimate subject of philosophical inquiry but perhaps the most fundamental branch of all AI. However, an uncritical overreliance on unsupervised methods poses major epistemic and ethical risks. I conclude by advocating for a pragmatic, error-statistical approach that embraces the opportunities and mitigates the challenges posed by this powerful class of algorithms.

Prospects for the discovery of robust and reproducible biomarkers have improved considerably with the development of sensitive omics platforms that can enable measurement of biological molecules at an unprecedented scale. With technical barriers to success lowering, the challenge is now moving into the analytical domain. Genome-wide discovery presents a problem of scale and multiple testing as standard statistical methods struggle to distinguish signal from noise in increasingly complex biological systems. Machine learning and AI methods are good at finding answers in large datasets, but they have a tendency to overfit solutions. It may be possible to find a local answer or mechanism in a specific patient sample or small group of samples, but this may not generalise to wider patient populations due to the high likelihood of false discovery. The rise of explainable AI offers to improve the opportunity for true discovery by providing explanations for predictions that can be explored mechanistically before proceeding to costly and time-consuming validation studies. This review aims to introduce some of the basic concepts of machine learning and AI for biomarker discovery with a focus on post hoc explanation of predictions. To illustrate this, we consider how explainable AI has already been used successfully, and we explore a case study that applies AI to biomarker discovery in rheumatoid arthritis, demonstrating the accessibility of tools for AI and machine learning. We use this to illustrate and discuss some of the potential challenges and solutions that may enable AI to critically interrogate disease and response mechanisms.

Despite the popularity of feature importance (FI) measures in interpretable machine learning, the statistical adequacy of these methods is rarely discussed. From a statistical perspective, a major distinction is between analysing a variable’s importance before and after adjusting for covariates—i.e., between marginal and conditional measures. Our work draws attention to this rarely acknowledged, yet crucial distinction and showcases its implications. We find that few methods are available for testing conditional FI and practitioners have hitherto been severely restricted in method application due to mismatched data requirements. Most real-world data exhibits complex feature dependencies and incorporates both continuous and categorical features (i.e., mixed data). Both properties are oftentimes neglected by conditional FI measures. To fill this gap, we propose to combine the conditional predictive impact (CPI) framework with sequential knockoff sampling. The CPI enables conditional FI measurement that controls for any feature dependencies by sampling valid knockoffs—hence, generating synthetic data with similar statistical properties—for the data to be analysed. Sequential knockoffs were deliberately designed to handle mixed data and thus allow us to extend the CPI approach to such datasets. We demonstrate through numerous simulations and a real-world example that our proposed workflow controls type I error, achieves high power, and is in-line with results given by other conditional FI measures, whereas marginal FI metrics can result in misleading interpretations. Our findings highlight the necessity of developing statistically adequate, specialized methods for mixed data.

Organisations that design and deploy artificial intelligence (AI) systems increasingly commit themselves to high-level, ethical principles. However, there still exists a gap between principles and practices in AI ethics. One major obstacle organisations face when attempting to operationalise AI Ethics is the lack of a well-defined material scope. Put differently, the question to which systems and processes AI ethics principles ought to apply remains unanswered. Of course, there exists no universally accepted definition of AI, and different systems pose different ethical challenges. Nevertheless, pragmatic problem-solving demands that things should be sorted so that their grouping will promote successful actions for some specific end. In this article, we review and compare previous attempts to classify AI systems for the purpose of implementing AI governance in practice. We find that attempts to classify AI systems proposed in previous literature use one of three mental models: the Switch, i.e., a binary approach according to which systems either are or are not considered AI systems depending on their characteristics; the Ladder, i.e., a risk-based approach that classifies systems according to the ethical risks they pose; and the Matrix, i.e., a multi-dimensional classification of systems that take various aspects into account, such as context, input data, and decision-model. Each of these models for classifying AI systems comes with its own set of strengths and weaknesses. By conceptualising different ways of classifying AI systems into simple mental models, we hope to provide organisations that design, deploy, or regulate AI systems with the vocabulary needed to demarcate the material scope of their AI governance frameworks.

One of the goals of causal inference is to generalize from past experiments and observational data to novel conditions. While it is in principle possible to eventually learn a mapping from a novel experimental condition to an outcome of interest, provided a sufficient variety of experiments is available in the training data, coping with a large combinatorial space of possible interventions is hard. Under a typical sparse experimental design, this mapping is ill-posed without relying on heavy regularization or prior distributions. Such assumptions may or may not be reliable, and can be hard to defend or test. In this paper, we take a close look at how to warrant a leap from past experiments to novel conditions based on minimal assumptions about the factorization of the distribution of the manipulated system, communicated in the well-understood language of factor graph models. A postulated interventional factor model (IFM) may not always be informative, but it conveniently abstracts away a need for explicitly modeling unmeasured confounding and feedback mechanisms, leading to directly testable claims. Given an IFM and datasets from a collection of experimental regimes, we derive conditions for identifiability of the expected outcomes of new regimes never observed in these training data. We implement our framework using several efficient algorithms, and apply them on a range of semi-synthetic experiments.

Abstract Over one thousand blood pressure (BP) loci have been identified by genetic association studies. However, determination of causal genes remains a bottleneck for further translational discovery. Here we triage genes identified by a BP genome-wide association study (GWAS) using optimized machine learning (ML) methodologies. We investigated regression models with nested cross-validation, benchmarking fourteen models (tree-based, ensemble and generalized linear models) using multi-omic features and 293 training genes. The top-performing model was extreme gradient boosting (0.897 predicted r2) that prioritized 794 genes. These genes showed significantly more intolerance to variation and were more often termed as essential. 27/794 genes showed evidence of direct interaction with blood pressure medications potentially highlighting opportunities for genetic stratification of response. Notably some BP drug mechanisms were not well represented in GWAS, while 51 genes showed no interaction with known BP drugs, highlighting possible target and repositioning opportunities. This study exploits ML to prioritize signals within BP-GWAS associations based on similarities with established BP-drug interacting genes, streamlining identification of genes underpinning BP that could inform disease management and drug discovery.

Shapley values have achieved great popularity in explainable artificial intelligence. However, with standard sampling methods, resulting feature attributions are susceptible to adversarial attacks. This originates from target function evaluations at extrapolated data points, which are easily detectable and hence, enable models to behave accordingly. In this paper, we introduce a novel strategy for increased robustness against adversarial attacks of both local and global explanations: Knockoff imputed Shapley values. Our approach builds on the model-X knockoff methodology, which generates synthetic data that preserves statistical properties of the original samples. This enables researchers to flexibly choose an appropriate model to generate on-manifold data for the calculation of Shapley values upfront, instead of having to estimate a large number of conditional densities or make strong parametric assumptions. Through real and simulated data experiments, we demonstrate the effectiveness of knockoff imputation against adversarial attacks.

Online controlled experiments, also known as A/B tests, have become ubiquitous. While many practical challenges in running experiments at scale have been thoroughly discussed, the ethical dimension of A/B testing has been neglected. This article fills this gap in the literature by introducing a new, soft ethics and governance framework that explicitly recognizes how the rise of an experimentation culture in industry settings brings not only unprecedented opportunities to businesses but also significant responsibilities. More precisely, the article (a) introduces a set of principles to encourage ethical and responsible experimentation to protect users, customers, and society; (b) argues that ensuring compliance with the proposed principles is a complex challenge unlikely to be addressed by resorting to a one-solution response; (c) discusses the relevance and effectiveness of several mechanisms and policies in educating, governing, and incentivizing companies conducting online controlled experiments; and (d) offers a list of prompting questions specifically designed to help and empower practitioners by stimulating specific ethical deliberations and facilitating coordination among different groups of stakeholders.

Background: When starting treatment with factor VIII (FVIII) concentrate in individuals with severe hemophilia A, minimizing the risk of FVIII inhibitor development is an important concern for clinicians, patients and their families. Despite increased understanding of some of the genetic and environmental risk factors for inhibitor formation from epidemiological studies, identification of previously untreated patients (PUPs) at risk of developing inhibitors remains a challenge. Prediction of inhibitor development may help optimize care in PUPs with hemophilia A. RNA sequencing (RNA-Seq) with next-generation sequencing (NGS) could be used to investigate the underlying mechanisms of inhibitor development. Aims: Toevaluate peripheral blood transcriptomic profiles prior to and during early FVIII treatment in severe hemophilia A patients who did or did not develop inhibitors to simoctocog alfain the prospective NuProtect study. Methods: The NuProtect study (NCT01712438; EudraCT 2012-002554-23) was a multicenter, multinational, open-label, non-controlled, phase III study that enrolled male PUPs with severe hemophilia A (FVIII:C <1%) of any age and ethnicity. Patients received simoctocog alfa for prophylaxis or on-demand treatment, as well as for surgical prophylaxis. RNA-Seq was performed on blood samples collected from baseline and every 3-4 exposure days (ED) up to 20 EDs or until inhibitor development to assess messenger RNA (mRNA) expression profiles. Bioinformatics analyses were performed for the following timepoints: baseline, early treatment and a later timepoint on treatment. Principal component analysis (PCA) was performed to identify differences in global mRNA expression profiles between inhibitor and non-inhibitor patients. Gene ontology (GO) analyses were performed to detect differences in GO clusters between inhibitor and non-inhibitor patients. Results: Within this first analysis of the full dataset, transcriptomic profiles from 78 of 108 severe hemophilia A PUPs were analyzed (inhibitors n=22, non-inhibitors n=56). Median (range) age at ED1 was 13 (0.5-146) months; 44.9% of patients had the intron 22 inversion F8 gene mutation. Although the PCA indicated no strong global signal for differential expression profiles between non-inhibitor and inhibitor patients, GO analyses demonstrated differences between these patient groups. Inhibitor patients demonstrated increased gene expression of metabolic processes (GO:0008152) and cellular processes (GO:0009987) at baseline. Upregulation of immune system processes (GO:0002376) was seen in inhibitor patients at the early timepoint compared with non-inhibitor patients. Expression of ribosomal proteins, including RPL17 and RPS8 were significantly upregulated at baseline in inhibitor patients (p<0.05). Inhibitor patients further demonstrated upregulation of proteins involved in adaptive immune responses: IGHV1-69-2 at all timepoints (p<0.01) and IGHV2-70D at baseline and the early timepoint (p<0.01) compared with non-inhibitor patients. Upregulation of the B-cell related factors DPP4 at baseline and the early timepoint (p<0.05) and CD20 at the late timepoint (p<0.05) was observed in inhibitor patients compared with non-inhibitor patients. Conclusion: Differential transcriptomic profiles were seen in severe hemophilia A patients who developed inhibitors compared with those who do not, even before commencing FVIII treatment. Inhibitor patients demonstrated alterations in ribosomal protein expression that could disrupt protein synthesis efficiency, protein folding and quality control processes, and trigger cellular stress responses. Inhibitor patients also showed an upregulation of B-cell mediated factors involved in adaptive immune mechanisms. Additional analyses are ongoing to evaluate dynamic changes in gene expression during commencement of FVIII replacement to further identify patients at risk of inhibitor development.

We propose methods for density estimation and data synthesis using a novel form of unsupervised random forests. Inspired by generative adversarial networks, we implement a recursive procedure in which trees gradually learn structural properties of the data through alternating rounds of generation and discrimination. The method is provably consistent under minimal assumptions. Unlike classic tree-based alternatives, our approach provides smooth (un)conditional densities and allows for fully synthetic data generation. We achieve comparable or superior performance to state-of-the-art probabilistic circuits and deep learning models on various tabular data benchmarks while executing about two orders of magnitude faster on average. An accompanying $\texttt{R}$ package, $\texttt{arf}$, is available on $\texttt{CRAN}$.

Researchers in explainable artificial intelligence have developed numerous methods for helping users understand the predictions of complex supervised learning models. By contrast, explaining the $\textit{uncertainty}$ of model outputs has received relatively little attention. We adapt the popular Shapley value framework to explain various types of predictive uncertainty, quantifying each feature's contribution to the conditional entropy of individual model outputs. We consider games with modified characteristic functions and find deep connections between the resulting Shapley values and fundamental quantities from information theory and conditional independence testing. We outline inference procedures for finite sample error rate control with provable guarantees, and implement efficient algorithms that perform well in a range of experiments on real and simulated data. Our method has applications to covariate shift detection, active learning, feature selection, and active feature-value acquisition.

Necessity and sufficiency are the building blocks of all successful explanations. Yet despite their importance, these notions have been conceptually underdeveloped and inconsistently applied in explainable artificial intelligence (XAI), a fast-growing research area that is so far lacking in firm theoretical foundations. Building on work in logic, probability, and causality, we establish the central role of necessity and sufficiency in XAI, unifying seemingly disparate methods in a single formal framework. We provide a sound and complete algorithm for computing explanatory factors with respect to a given context, and demonstrate its flexibility and competitive performance against state of the art alternatives on various tasks.