Around 8,000 technologists, half of whom are software engineers, work at RELX. Annually, the company spends $1.4bn on technology. The combination of our rich data assets, technology infrastructure and knowledge of how to use next generation technologies, such as machine learning and natural language processing, allows us to create effective solutions for our customers. Helping research chemists with Elsevier's Reaxys Reaxys enables the shortest path to chemistry research answers, supporting the early stages of drug development in the pharmaceutical industry, exploratory chemistry research in academia, and product development in industries such as chemicals and oil & gas. The amount of chemical information published each year is increasing exponentially, making it more and more challenging for research chemists to quickly find targeted and actionable information to help support their research.

With the advent of GDPR, the domain of explainable AI and model interpretability has gained added impetus. Methods to extract and communicate visibility into decision-making models have become legal requirement. Two specific types of explanations, contrastive and counterfactual have been identified as suitable for human understanding. In this paper, we propose a model agnostic method and its systemic implementation to generate these explanations using shapely additive explanations (SHAP). We discuss a generative pipeline to create contrastive explanations and use it to further to generate counterfactual datapoints. This pipeline is tested and discussed on the IRIS, Wine Quality & Mobile Features dataset. Analysis of the results obtained follows.

There has been rapidly growing interest in the use of algorithms for employment assessment, especially as a means to address or mitigate bias in hiring. Yet, to date, little is known about how these methods are being used in practice. How are algorithmic assessments built, validated, and examined for bias? In this work, we document and assess the claims and practices of companies offering algorithms for employment assessment, using a methodology that can be applied to evaluate similar applications and issues of bias in other domains. In particular, we identify vendors of algorithmic pre-employment assessments (i.e., algorithms to screen candidates), document what they have disclosed about their development and validation procedures, and evaluate their techniques for detecting and mitigating bias. We find that companies' formulation of "bias" varies, as do their approaches to dealing with it. We also discuss the various choices vendors make regarding data collection and prediction targets, in light of the risks and trade-offs that these choices pose. We consider the implications of these choices and we raise a number of technical and legal considerations.

The Fair Housing Act ("FHA"), enacted more than fifty years ago, prohibits discriminatory practices in housing. The FHA makes it illegal to "make unavailable or deny . . . In many jurisdictions, it is also illegal to discriminate on the basis of income (e.g. But recent technological advancements have raised new questions about the statute's reach--both in terms of which entities may be liable for violating the FHA and what new technologies may run afoul of the statute's prohibitions. For example, companies that use, facilitate, or support digital advertising need to be particularly cognizant of the FHA's purview.

The vast scope of GDPR has raised fresh challenges -- chief among them is the complex interaction between AI and the GDPR. In particular, this shines a spotlight on Article 22, which concerns automated profiling and decision-making, where the incorrect use of personal data can have huge ramifications for the individuals concerned. The problem is that existing AI system logic takes automated decisions without user consent. Since data is the engine behind AI, Article 22 impacts every industry hoping to leverage the power of technology to drive efficiencies through automated means. In an increasingly data-reliant business landscape, how can organisations reconcile the advent of disruptive technologies and their inherent risks while remaining fully compliant?

Algorithms are coming to the San Francisco District Attorney's Office. In a bid to combat implicit bias, artificial intelligence-backed software will redact race information from police incident reports before prosecutors make their initial charging decision.

Alphabet, the parent company of Google, failed to pass several proposals to address sexual harassment, antitrust issues and diversity policies at its annual shareholder meeting, despite hundreds of employees protesting outside the event. The annual meeting comes as Alphabet faces growing pressure from shareholders and employees, including over its handling of sexual harassment allegations, ethical concerns surrounding its artificial intelligence systems, its widespread reliance on and treatment of contract workers and its operations in China. But the company's board voted down all 13 proposals shareholders had put forward, which encompassed a wide range of social concerns surrounding the company. They included efforts to change employment practicesand end forced arbitration. Google removed its policy of forced arbitration regarding sexual assault claims for full-time employees in November 2018, following a walkout of roughly 20,000 employees over the company's responses to sexual misconduct.

The increasing prevalence and accessibility of artificial intelligence (AI) has allowed more companies to use AI to analyze data and engage with customers. While AI-powered software is already being utilized to carry out simple legal tasks, recent technological advancements are enabling AI to take responsibility for a significant amount of legal work. According to a 2017 report, 39 percent of Thomas Reuters in-house counsel agree that AI will become commonplace within the legal profession over the next decade. So, what exactly is AI, and how can it be effectively utilized in litigation? The term "artificial intelligence" is used to describe how computers perform tasks that are typically regarded as requiring human intelligence.

Companies and entire industries are looking to harness data analytics to make more accurate and effective decisions, within and across organizations. Such real-time and accurate insights have enabled boards and their management to be more effective in conducting their duties. Artificial intelligence (AI) mimics the learning function of the human brain, which means it could be deliberately or accidently corrupted and even adopt human biases, potentially resulting in mistakes and unethical decisions. Control of AI systems by the wrong hands is also a concern. Any AI system failure could have profound ramifications on security, decision-making and credibility, and may lead to costly litigation, reputational damage, regulatory scrutiny, and reduced stakeholder trust and profitability.

This paper provides a comprehensive survey of Machine Learning Testing (ML testing) research. It covers 128 papers on testing properties (e.g., correctness, robustness, and fairness), testing components (e.g., the data, learning program, and framework), testing workflow (e.g., test generation and test evaluation), and application scenarios (e.g., autonomous driving, machine translation). The paper also analyses trends concerning datasets, research trends, and research focus, concluding with research challenges and promising research directions in ML testing.