Society’s capacity for algorithmic problem-solving has never been greater. Artificial Intelligence is now applied across more domains than ever, a consequence of powerful abstractions, abundant data, and accessible software. As capabilities have expanded, so have risks, with models often deployed without fully understanding their potential impacts. Interpretable and interactive machine learning aims to make complex models more transparent and controllable, enhancing user agency. This review synthesizes key principles from the growing literature in this field. We first introduce precise vocabulary for discussing interpretability, like the distinction between glass box and explainable models. We then explore connections to classical statistical and design principles, like parsimony and the gulfs of interaction. Basic explainability techniques – including learned embeddings, integrated gradients, and concept bottlenecks – are illustrated with a simple case study. We also review criteria for objectively evaluating interpretability approaches. Throughout, we underscore the importance of considering audience goals when designing interactive data-driven systems. Finally, we outline open challenges and discuss the potential role of data science in addressing them. Code to reproduce all examples can be found at https://go.wisc.edu/3k1ewe.
Pub. online:2 May 2024Type:Data Science In ActionOpen Access
Journal:Journal of Data Science
Volume 22, Issue 2 (2024): Special Issue: 2023 Symposium on Data Science and Statistics (SDSS): “Inquire, Investigate, Implement, Innovate”, pp. 191–207
Abstract
Attention Deficit Hyperactivity Disorder (ADHD) is a frequent neurodevelopmental disorder in children that is commonly diagnosed subjectively. The objective detection of ADHD based on neuroimaging data has been a complex problem with low ranges of accuracy, possibly due to (among others) complex diagnostic processes, the high number of features considered and imperfect measurements in data collection. Hence, reliable neuroimaging biomarkers for detecting ADHD have been elusive. To address this problem we consider a recently proposed multi-model selection method called Sparse Wrapper AlGorithm (SWAG), which is a greedy algorithm that combines screening and wrapper approaches to create a set of low-dimensional models with good predictive power. While preserving the previous levels of accuracy, SWAG provides a measure of importance of brain regions for identifying ADHD. Our approach also provides a set of equally-performing and simple models which highlight the main feature combinations to be analyzed and the interactions between them. Taking advantage of the network of models resulting from this approach, we confirm the relevance of the frontal and temporal lobes as well as highlight how the different regions interact to detect the presence of ADHD. In particular, these results are fairly consistent across different learning mechanisms employed within the SWAG (i.e. logistic regression, linear and radial-kernel support vector machines) thereby providing population-level insights, as well as delivering feature combinations that are smaller and often perform better than those that would be used if employing their original versions directly.
Abstract: Hyperplane fitting factor rotations perform better than conventional rotations in attaining simple structure for complex configurations. Hyperplane rotations are reviewed and then compared using familiar exam es from the literature selected to vary in complexity. Included is a new method for fitting hyperplanes, hypermax, which updates the work of Horst (1941) and Derflinger and Kaiser (1989). Hypercon, a method for confirmatory target rotation, is a natural extension. These performed very well when compared with selected hyperplane and conventional rotations. The concluding sections consider the pros and cons of each method.
Factor analysis (FA) is the most commonly used pattern recognition methodology in social and health research. A technique that may help to better retrieve true information from FA is the rotation of the information axes. The purpose of this study was to evaluate whether the selection of rotation type affects the repeatability of the patterns derived from FA, under various scenarios of random error introduced, based on simulated data from the Standard Normal distribution. It was observed that when applying promax non - orthogonal rotation, the results were more repeatable as compared to the orthogonal rotation, irrespective of the level of random error introduced in the model.
Abstract: In maximum likelihood exploratory factor analysis, the estimates of unique variances can often turn out to be zero or negative, which makes no sense from a statistical point of view. In order to overcome this difficulty, we employ a Bayesian approach by specifying a prior distribution for the variances of unique factors. The factor analysis model is estimated by EM algorithm, for which we provide the expectation and maximization steps within a general framework of EM algorithms. Crucial issues in Bayesian factor analysis model are the choice of adjusted parameters including the number of factors and also the hyper-parameters for the prior distribution. The choice of these parameters can be viewed as a model selection and evaluation problem. We derive a model selection criterion for evaluating a Bayesian factor analysis model. Monte Carlo simulations are conducted to investigate the effectiveness of the proposed procedure. A real data example is also given to illustrate our procedure. We observe that our modeling procedure prevents the occurrence of improper solutions and also chooses the appropriate number of factors objectively.