When might human input help (or not) when assessing risk in fairness-related domains? Dressel and Farid asked Mechanical Turk workers to evaluate a subset of individuals in the ProPublica COMPAS data set for risk of recidivism, and concluded that COMPAS predictions were no more accurate or fair than predictions made by humans. We delve deeper into this claim in this paper. We construct a Human Risk Score based on the predictions made by multiple Mechanical Turk workers on the same individual, study the agreement and disagreement between COMPAS and Human Scores on subgroups of individuals, and construct hybrid Human+AI models to predict recidivism. Our key finding is that on this data set, human and COMPAS decision making differed, but not in ways that could be leveraged to significantly improve ground truth prediction. We present the results of our analyses and suggestions for how machine and human input may have complementary strengths to address challenges in the fairness domain.

In this work, we ask two questions: 1. Can we predict the type of community interested in a news article using only features from the article content? and 2. How well do these models generalize over time? To answer these questions, we compute well-studied content-based features on over 60K news articles from 4 communities on reddit.com. We train and test models over three different time periods between 2015 and 2017 to demonstrate which features degrade in performance the most due to concept drift. Our models can classify news articles into communities with high accuracy, ranging from 0.81 ROC AUC to 1.0 ROC AUC. However, while we can predict the community-specific popularity of news articles with high accuracy, practitioners should approach these models carefully. Predictions are both community-pair dependent and feature group dependent. Moreover, these feature groups generalize over time differently, with some only degrading slightly over time, but others degrading greatly. Therefore, we recommend that community-interest predictions are done in a hierarchical structure, where multiple binary classifiers can be used to separate community pairs, rather than a traditional multi-class model. Second, these models should be retrained over time based on accuracy goals and the availability of training data.

Smart devices of everyday use (such as smartphones and wearables) are increasingly integrated with sensors that provide immense amounts of information about a person's daily life such as behavior and context. The automatic and unobtrusive sensing of behavioral context can help develop solutions for assisted living, fitness tracking, sleep monitoring, and several other fields. Towards addressing this issue, we raise the question: can a machine learn to recognize a diverse set of contexts and activities in a real-life through joint learning from raw multi-modal signals (e.g. accelerometer, gyroscope and audio etc.)? In this paper, we propose a multi-stream temporal convolutional network to address the problem of multi-label behavioral context recognition. A four-stream network architecture handles learning from each modality with a contextualization module which incorporates extracted representations to infer a user's context. Our empirical evaluation suggests that a deep convolutional network trained end-to-end achieves an optimal recognition rate. Furthermore, the presented architecture can be extended to include similar sensors for performance improvements and handles missing modalities through multi-task learning without any manual feature engineering on highly imbalanced and sparsely labeled dataset.

We developed a novel statistical method to identify structural differences between networks characterized by structural equation models. We propose to reparameterize the model to separate the differential structures from common structures, and then design an algorithm with calibration and construction stages to identify these differential structures. The calibration stage serves to obtain consistent prediction by building the L2 regularized regression of each endogenous variables against pre-screened exogenous variables, correcting for potential endogeneity issue. The construction stage consistently selects and estimates both common and differential effects by undertaking L1 regularized regression of each endogenous variable against the predicts of other endogenous variables as well as its anchoring exogenous variables. Our method allows easy parallel computation at each stage. Theoretical results are obtained to establish nonasymptotic error bounds of predictions and estimates at both stages, as well as the consistency of identified common and differential effects. Our studies on synthetic data demonstrated that our proposed method performed much better than independently constructing the networks. A real data set is analyzed to illustrate the applicability of our method.

Deep Learning has a hierarchical network architecture to represent the complicated feature of input patterns. The adaptive structural learning method of Deep Belief Network (DBN) has been developed. The method can discover an optimal number of hidden neurons for given input data in a Restricted Boltzmann Machine (RBM) by neuron generation-annihilation algorithm, and generate a new hidden layer in DBN by the extension of the algorithm. In this paper, the proposed adaptive structural learning of DBN was applied to the comprehensive medical examination data for the cancer prediction. The prediction system shows higher classification accuracy (99.8% for training and 95.5% for test) than the traditional DBN. Moreover, the explicit knowledge with respect to the relation between input and output patterns was extracted from the trained DBN network by C4.5. Some characteristics extracted in the form of IF-THEN rules to find an initial cancer at the early stage were reported in this paper.

Advances in sensor technology have enabled the collection of large-scale datasets. Such datasets can be extremely noisy and often contain a significant amount of outliers that result from sensor malfunction or human operation faults. In order to utilize such data for real-world applications, it is critical to detect outliers so that models built from these datasets will not be skewed by outliers. In this paper, we propose a new outlier detection method that utilizes the correlations in the data (e.g., taxi trip distance vs. trip time). Different from existing outlier detection methods, we build a robust regression model that explicitly models the outliers and detects outliers simultaneously with the model fitting. We validate our approach on real-world datasets against methods specifically designed for each dataset as well as the state of the art outlier detectors. Our outlier detection method achieves better performances, demonstrating the robustness and generality of our method. Last, we report interesting case studies on some outliers that result from atypical events.

Recently, most of the organizations experienced severe downfall due to an undetected malware, Deeplocker, which secretly evaded even the stringent cyber security mechanisms. Deeplocker leverages the AI model to attack the target host by using indicators such as facial recognition, geolocation and voice recognition. This incidence speaks volumes about the big role AI plays in the cybersecurity domain. In fact, some may even go on to say that AI for cybersecurity is no longer a nice to have tech rather a necessity. Large and small organizations and even startups are hugely investing in building AI systems to analyze the huge data trove and in turn, help their cybersecurity professionals to identify possible threats and take precautions or immediate actions to solve it. If AI can be used in getting the systems protected, it can also harm it.

Pit bulls get a bad rap, which is especially vexing given that no one actually knows exactly what a pit bull is. There's no unified definition, because "pit bull" is not a recognized breed. But the label can have devastating consequences for dogs in shelters, who are perceived as less adoptable because of their purported heritage. In recent years, especially with the advent of genetic testing, some researchers have a new idea: just stop labelling mixed-breed dogs altogether. Researchers at Arizona State University decided to do a large-scale analysis of shelter dogs by looking at every pup that came through the doors of two animal shelters, one in Phoenix, AZ and one in San Diego, CA.

With the rising influence of machine learning algorithms on many important aspects of our daily lives, there are growing concerns that biases inherent in data can lead the behavior of these algorithms to discriminate against certain populations [1, 2, 4, 6, 8, 28, 29, 15]. In recent years, substantial research effort has been devoted to the development of mathematical definitions of bias, or its opposite, fairness, in algorithms and in data [15, 18, 26, 23, 19, 32]. In this work, we focus on the fairness scenario where there are multiple protected attributes that we aim to ensure fairness for, and which may potentially overlap with each other, such as gender, race, and sexual orientation. Our guiding principle is intersectionality, the core theoretical framework underlying the thirdwave feminist movement [13]. The principle of intersectionality states that racism, sexism, and other social systems which harm marginalized groups are interlocking in their effects, such that the lived experience of, e.g., black women, is very different than that of, e.g., white women. Intersectionality was defined by Kimberlé Crenshaw in the 1980's [13] and popularized in the 1990's, e.g. by Patricia Hill Collins [10], although the ideas are much older [11, 35]. In the context of machine learning and fairness, intersectionality was recently considered by [7], who studied the impact of the intersection of gender and skin color on computer vision performance, and by [23, 19], who aimed to protect certain subgroups in order to prevent "fairness gerrymandering."

Kearns et al. [2018] recently proposed a notion of rich subgroup fairness intended to bridge the gap between statistical and individual notions of fairness. Rich subgroup fairness picks a statistical fairness constraint (say, equalizing false positive rates across protected groups), but then asks that this constraint hold over an exponentially or infinitely large collection of subgroups defined by a class of functions with bounded VC dimension. They give an algorithm guaranteed to learn subject to this constraint, under the condition that it has access to oracles for perfectly learning absent a fairness constraint. In this paper, we undertake an extensive empirical evaluation of the algorithm of Kearns et al. On four real datasets for which fairness is a concern, we investigate the basic convergence of the algorithm when instantiated with fast heuristics in place of learning oracles, measure the tradeoffs between fairness and accuracy, and compare this approach with the recent algorithm of Agarwal et al. [2018], which implements weaker and more traditional marginal fairness constraints defined by individual protected attributes. We find that in general, the Kearns et al. algorithm converges quickly, large gains in fairness can be obtained with mild costs to accuracy, and that optimizing accuracy subject only to marginal fairness leads to classifiers with substantial subgroup unfairness. We also provide a number of analyses and visualizations of the dynamics and behavior of the Kearns et al. algorithm. Overall we find this algorithm to be effective on real data, and rich subgroup fairness to be a viable notion in practice.