Governments are increasingly turning to algorithmic risk assessments when making important decisions, believing that these algorithms will improve public servants' ability to make policy-relevant predictions and thereby lead to more informed decisions. Yet because many policy decisions require balancing risk-minimization with competing social goals, evaluating the impacts of risk assessments requires considering how public servants are influenced by risk assessments when making policy decisions rather than just how accurately these algorithms make predictions. Through an online experiment with 2,140 lay participants simulating two high-stakes government contexts, we provide the first large-scale evidence that risk assessments can systematically alter decision-making processes by increasing the salience of risk as a factor in decisions and that these shifts could exacerbate racial disparities. These results demonstrate that improving human prediction accuracy with algorithms does not necessarily improve human decisions and highlight the need to experimentally test how government algorithms are used by human decision-makers.

The rapidly increasing amount of available data in many fields of application has triggered the development of distributed methods for data analysis. Distributed methods, besides being able to speed up computations considerably, can reduce local memory requirements and can also help in protecting privacy, by refraining from storing a whole dataset in a single central location. Moreover, distributed methods occur naturally when data is by construction observed and processed at multiple locations, such as for instance in astronomy, meteorology, seismology, military radar or air traffic control systems. The information theoretic aspects of distributed statistical methods have only been studied rigorously relatively recently. Most work up till now has focussed on distributed methods for estimating a signal in the normal-means model under bandwidth, or communication restrictions (see for instance [23, 6, 4, 7]) and, related to that, on deriving minimix lower bounds and optimal distributed estimation strategies in the context of nonparametric regression, density estimation and Gaussian signal-in-white-noise models (e.g.

Recent research from Accuity indicates that banks struggle with the time required for staff to perform manual repetitive compliance tasks. Labor still accounts for the largest cost of compliance functions. Management tends to view compliance functions as cost centers, rather than profit generators, while regulatory changes drive bank behavior. Most people today understand the power of a data-driven approach when it comes to their recent online shopping behavior: we purchase goods online and our preferred vendor will offer us previously bought goods that might complement our purchase. The more data this algorithm has, the easier it can make predictions and recommendations for users.

Timnit Gebru, one of Google's most prominent researchers on ethics and computer vision, says she was fired this week after sending an email to Google Brain Women and Allies, an internal resource group at the company. The email alludes to Google censoring one of Gebru's research papers without talking to her about it, as well as the poor treatment of those who advocate for underrepresented people at the company. The email was published in full on the outlet Platformer. After sending the email, Gebru had an exchange with managers and privately threatened to quit unless certain undisclosed conditions were met. Instead, Gebru says she was immediately fired, she told OneZero's Will Oremus.

Waymo has long kept details about its industry-leading self-driving technology under wraps. The company has done millions of miles of testing in Arizona and California--including thousands of miles with no one behind the wheel. But until last month, almost everyone who experienced those driverless rides was bound by a strict nondisclosure agreement. This story originally appeared on Ars Technica, a trusted source for technology news, tech policy analysis, reviews, and more. Ars is owned by WIRED's parent company, Condé Nast.

Machine Learning has seen tremendous growth recently, which has led to a larger adoption of ML systems for educational assessments, credit risk, healthcare, employment, criminal justice, to name a few. Trustworthiness of ML and NLP systems is a crucial aspect and requires guarantee that the decisions they make are fair and robust. Aligned with this, we propose a framework GYC, to generate a set of counterfactual text samples, which are crucial for testing these ML systems. Our main contributions include a) We introduce GYC, a framework to generate counterfactual samples such that the generation is plausible, diverse, goal-oriented, and effective, b) We generate counterfactual samples, that can direct the generation towards a corresponding condition such as named-entity tag, semantic role label, or sentiment. Our experimental results on various domains show that GYC generates counterfactual text samples exhibiting the above four properties. %The generated counterfactuals can then be fed complementary to the existing data augmentation for improving the debiasing algorithms performance as compared to existing counterfactuals generated by token substitution. GYC generates counterfactuals that can act as test cases to evaluate a model and any text debiasing algorithm.

Algorithms are now routinely used to make consequential decisions that affect human lives. Examples include college admissions, medical interventions or law enforcement. While algorithms empower us to harness all information hidden in vast amounts of data, they may inadvertently amplify existing biases in the available datasets. This concern has sparked increasing interest in fair machine learning, which aims to quantify and mitigate algorithmic discrimination. Indeed, machine learning models should undergo intensive tests to detect algorithmic biases before being deployed at scale. In this paper, we use ideas from the theory of optimal transport to propose a statistical hypothesis test for detecting unfair classifiers. Leveraging the geometry of the feature space, the test statistic quantifies the distance of the empirical distribution supported on the test samples to the manifold of distributions that render a pre-trained classifier fair. We develop a rigorous hypothesis testing mechanism for assessing the probabilistic fairness of any pre-trained logistic classifier, and we show both theoretically as well as empirically that the proposed test is asymptotically correct. In addition, the proposed framework offers interpretability by identifying the most favorable perturbation of the data so that the given classifier becomes fair.

Survival analysis methods are required for censored data in which the outcome of interest is generally the time until an event (onset of disease, death, etc.), but the exact time of the event is unknown (censored) for some individuals. When a lower bound for these missing values is known (for example, a patient is known to be alive until at least time t) the data is said to be right-censored. A common survival analysis technique is Cox proportional hazards regression (Cox, 1972) which models the hazard rate for an event as a linear combination of covariate effects. Although this model is widely used and easily interpreted, its parametric nature makes it unable to identify nonlinear effects or interactions between covariates (Bou-Hamad et al., 2011). Recursive partitioning techniques (also referred to as trees) are a popular alternative to parametric models. When applied to survival data, survival tree algorithms partition the covariate space into smaller and smaller regions (nodes) containing observations with homogeneous survival outcomes.

Outcry is growing within Google over the treatment of the AI ethics researcher Timnit Gebru, with Gebru's colleagues challenging the company's account of her exit in an open letter. In a letter posted on Monday on Medium, Gebru's colleagues disputed an executive's claim that she had resigned and called internal research policies into question. "Dr Gebru did not resign, despite what Jeff Dean (Senior Vice President and head of Google Research), has publicly stated," the letter reads before going into detail about the events that led to Gebru's dismissal. Gebru, a Black female scientist who is highly respected in her field, said on Twitter last week that she had been fired after sending an email to an internal company group for women and allies, expressing frustration over discrimination at Google and a dispute over one of her papers that was retracted after initially being approved for publication. The paper in question examined the ethical issues associated with AI language technology and reportedly mentions Google's own software, which is important to the company's business model development.

Tinder parent company Match Group is partnering with one of the largest anti-sexual violence groups in the US to audit how it handles reports of sexual assault across its many dating platforms. The Rape, Abuse & Incest National Network (RAINN) will "conduct a comprehensive review of sexual misconduct reporting, moderation and response across Match Group's dating platforms and to work together to improve current safety systems and tools," the company said on Monday. The first phase of the review will focus on Tinder, Hinge and Plenty of Fish before moving on to Match's other platforms -- the company owns around 40 other dating brands altogether. The partnership, which Axios was the first to report on, will continue through 2021, with recommend changes rolling out "shortly thereafter." This is an important move for Match, even if there aren't many details at the moment. While you frequently hear of horror stories, it's difficult to put an exact number to the incidents of sexual assault that happen through Tinder and other online dating platforms.