Political polarization in the US is on the rise. This polarization negatively affects the public sphere by contributing to the creation of ideological echo chambers. In this paper, we focus on addressing one of the factors that contributes to this polarity, polarized media. We introduce a framework for depolarizing news articles. Given an article on a certain topic with a particular ideological slant (eg., liberal or conservative), the framework first detects polar language in the article and then generates a new article with the polar language replaced with neutral expressions. To detect polar words, we train a multi-attribute-aware word embedding model that is aware of ideology and topics on 360k full-length media articles. Then, for text generation, we propose a new algorithm called Text Annealing Depolarization Algorithm (TADA). TADA retrieves neutral expressions from the word embedding model that not only decrease ideological polarity but also preserve the original argument of the text, while maintaining grammatical correctness. We evaluate our framework by comparing the depolarized output of our model in two modes, fully-automatic and semi-automatic, on 99 stories spanning 11 topics. Based on feedback from 161 human testers, our framework successfully depolarized 90.1% of paragraphs in semi-automatic mode and 78.3% of paragraphs in fully-automatic mode. Furthermore, 81.2% of the testers agree that the non-polar content information is well-preserved and 79% agree that depolarization does not harm semantic correctness when they compare the original text and the depolarized text. Our work shows that data-driven methods can help to locate political polarity and aid in the depolarization of articles.

More than 200 engineers and other workers have formed a union at Google, a breakthrough in labor organizing in Silicon Valley where workers have clashed with executives over workplace culture, diversity and ethics. Across half a dozen Google offices in the U.S. and Canada, 226 workers signed cards to form the Alphabet Workers Union, the group said on Monday. They are supported by the Communications Workers of America, which represents workers in telecommunications and media. The new union won't have collective bargaining rights and represents only a small fraction of Google's workforce. Google, which is owned by Alphabet Inc., has faced employee outcry over issues including sexual harassment, its work with the Pentagon and the company's treatment of its massive contract workforce.

A group of 226 engineers and other Google workers have formed a union, according to an article and opinion piece in the New York Times. Called the Alphabet Workers Union, it is affiliated with the Communications Workers of America and was organized in secret over the last year or so. "We are joining together -- temps, vendors, contractors, and full-time employees -- to create a unified worker voice," wrote the Parul Koul and Chewy Shaw, the executive chair and vice chair of the Alphabet Workers Union. "We want Alphabet to be a company where workers have a meaningful say in decisions that affect us and the societies we live in." The union represents a small minority of the company's 260,000 strong employee and contractor workforce.

Abstract dialectical frameworks (ADFs) are a recently introduced powerful generalization of Dung’s popular abstract argumentation frameworks (AFs). Inspired by similar work for AFs, we introduce a decomposition scheme for ADFs, which proceeds along the ADF’s strongly connected components. We find that, for several semantics, the decomposition-based version coincides with the original semantics, whereas for others, it gives rise to a new semantics. These new semantics allow us to deal with pertinent problems such as odd-length negative cycles in a more general setting, that for instance also encompasses logic programs. We perform an exhaustive analysis of the computational complexity of these new, so-called naive-based semantics. The results are quite interesting, for some of them involve little-known classes of the so-called Boolean hierarchy (another hierarchy in between classes of the polynomial hierarchy). Furthermore, in credulous and sceptical entailment, the complexity can be different depending on whether we check for truth or falsity of a specific statement.

Tucked away in the 4,517-page annual defense bill awaiting signature is an overlooked piece of legislation on artificial intelligence (AI). It doesn't make every military weapon system autonomous or require brigades of robotic infantry. Instead, it's a sensible, 63-page plan establishing a civilian-led initiative to coordinate and accelerate investments in "trustworthy" artificial intelligence systems across the federal government. In passing this legislation, the United States Congress has demonstrated that it collectively realizes that AI will be transformative, and that urgent research and development is needed to ensure the United States remains the world leader in AI. Make no mistake, the "National Artificial Intelligence Initiative Act of 2020," also dubbed as "Division E" of the National Defense Authorization Act (NDAA), is the closest thing to a national strategy on AI from the United States to be formally endorsed by Congress.

Artificial Intelligence (AI) is influencing the future of virtually every industry and each person on the planet. Artificial intelligence has been set up as the primary driver of growing technologies, for example, robotics, big data, and the Internet of Things (IoT). Moving into 2021, Artificial Intelligence will keep on going about as a principle technological pioneer for years to come. Artificial intelligence is a highlight of the coming "new normal" in our entire lives. Going ahead, AI will be the intelligent core of robotic, automated, and contactless procedures that will shield us all from future outbreaks.

When the present is scary, the future can be virtually unthinkable. But it's at times of great change and uncertainty--and 2020 surely qualifies--that it is most important to try to look ahead, to think about how decisions made right now can reverberate. This year, Future Tense Fiction--a partnership of Future Tense and Arizona State University's Center for Science and the Imagination--published 12 stories that take very different looks at the years to come. In the case of Max Barry's "It Came From Cruden Farm," that future is very near--it's set on Inauguration Day 2021, when a new president learns that the U.S. government has custody of an alien, and it's complicated. Other futures are more distant; as part of our package of three stories on artificial intelligence and governance, "The State Machine," by Yudhanjaya Wijeratne, follows a graduate student trying to learn about the very earliest days of his country being run by A.I. Tobias S. Buckell's "Scar Tissue," Holli Mintzer's "Legal Salvage," and Karl Schroeder's "The Suicide of Our Troubles" all grapple, in very different ways, with legal rights for nonhumans.

Machine learning based systems are reaching society at large and in many aspects of everyday life. This phenomenon has been accompanied by concerns about the ethical issues that may arise from the adoption of these technologies. ML fairness is a recently established area of machine learning that studies how to ensure that biases in the data and model inaccuracies do not lead to models that treat individuals unfavorably on the basis of characteristics such as e.g. race, gender, disabilities, and sexual or political orientation. In this manuscript, we discuss some of the limitations present in the current reasoning about fairness and in methods that deal with it, and describe some work done by the authors to address them. More specifically, we show how causal Bayesian networks can play an important role to reason about and deal with fairness, especially in complex unfairness scenarios. We describe how optimal transport theory can be used to develop methods that impose constraints on the full shapes of distributions corresponding to different sensitive attributes, overcoming the limitation of most approaches that approximate fairness desiderata by imposing constraints on the lower order moments or other functions of those distributions. We present a unified framework that encompasses methods that can deal with different settings and fairness criteria, and that enjoys strong theoretical guarantees. We introduce an approach to learn fair representations that can generalize to unseen tasks. Finally, we describe a technique that accounts for legal restrictions about the use of sensitive attributes.

Training a classifier under fairness constraints has gotten increasing attention in the machine learning community thanks to moral, legal, and business reasons. However, several recent works addressing algorithmic fairness have only focused on simple models such as logistic regression or support vector machines due to non-convex and non-differentiable fairness criteria across protected groups, such as race or gender. Neural networks, the most widely used models for classification nowadays, are precluded and lack theoretical guarantees. This paper aims to fill this missing but crucial part of the literature of algorithmic fairness for neural networks. In particular, we show that overparametrized neural networks could meet the fairness constraints. The key ingredient of building a fair neural network classifier is establishing no-regret analysis for neural networks in the overparameterization regime, which may be of independent interest in the online learning of neural networks and related applications.

As machine learning algorithms grow in popularity and diversify to many industries, ethical and legal concerns regarding their fairness have become increasingly relevant. We explore the problem of algorithmic fairness, taking an information-theoretic view. The maximal correlation framework is introduced for expressing fairness constraints and shown to be capable of being used to derive regularizers that enforce independence and separation-based fairness criteria, which admit optimization algorithms for both discrete and continuous variables which are more computationally efficient than existing algorithms. We show that these algorithms provide smooth performance-fairness tradeoff curves and perform competitively with state-of-the-art methods on both discrete datasets (COMPAS, Adult) and continuous datasets (Communities and Crimes).