With the growing use of AI in highly consequential domains, the quantification and removal of bias with respect to protected attributes, such as gender, race, etc., is becoming increasingly important. While quantifying bias is essential, sometimes the needs of a business (e.g., hiring) may require the use of certain features that are critical in a way that any bias that can be explained by them might need to be exempted. E.g., a standardized test-score may be a critical feature that should be weighed strongly in hiring even if biased, whereas other features, such as zip code may be used only to the extent that they do not discriminate. In this work, we propose a novel information-theoretic decomposition of the total bias (in a counterfactual sense) into a non-exempt component that quantifies the part of the bias that cannot be accounted for by the critical features, and an exempt component which quantifies the remaining bias. This decomposition allows one to check if the bias arose purely due to the critical features (inspired from the business necessity defense of disparate impact law) and also enables selective removal of the non-exempt component if desired. We arrive at this decomposition through examples that lead to a set of desirable properties (axioms) that any measure of non-exempt bias should satisfy. We demonstrate that our proposed counterfactual measure satisfies all of them. Our quantification bridges ideas of causality, Simpson's paradox, and a body of work from information theory called Partial Information Decomposition. We also obtain an impossibility result showing that no observational measure of non-exempt bias can satisfy all of the desirable properties, which leads us to relax our goals and examine observational measures that satisfy only some of these properties. We then perform case studies to show how one can train models while reducing non-exempt bias.

Artificial intelligence (AI) is often depicted as harming humanity in science fiction. The Stanley Kubrick-directed sci-fi masterpiece "2001: A Space Odyssey" found the HAL 9000 killing off its space crew, while the "Terminator" franchise centers on AI-powered machines rising up and warring with humans. However, a major real-world use case for artificial intelligence has been medicine. Learn how AI is being implemented in the medical world! Machine learning is the practice in artificial intelligence of training a system to learn and iteratively improve without the need for additional programming.

In this article, we will look at toxic speech detection, the problem of text moderation and understand the different challenges that one might encounter trying to automate the process. We look at several NLP and deep learning approaches to solve the problem and finally implement a toxic speech classifier using BERT embeddings. As of June 2019 there are now over 4.4 billion internet users. According to the latest Domo Data Never Sleeps report, Twitter users send 511,200 tweets per minute. While that happens, TikTok gets banned in Indonesia, Discord sees an increasing number of neo-Nazi posts, tech and film celebrity accounts get hacked so hackers can spurt out several racist slurs and hate speech volumes rise in India on facebook due to the controversial Citizenship Amendment Act (CAA). Social media continues to be used by several to incite violence, spread hate and target minorities based on religion, sex, race and disabilities.

Amazon announced on Wednesday it was implementing a "one-year moratorium" on police use of Rekognition, its facial-recognition technology. Lawmakers and civil liberties groups have expressed growing alarm over the tool's potential for misuse by law enforcement for years, particularly against communities of color. Now, weeks into worldwide protests against police brutality and racism sparked by the killing of George Floyd, Amazon appears to have acknowledged these concerns. In a short blog post about the decision, the tech giant said it hopes the pause "might give Congress enough time to implement appropriate rules" for the use of facial-recognition technology, which is largely unregulated in the US. Critics have said that the tech could easily be abused by the government, and they cite studies showing tools like Rekognition misidentify people of color at higher rates than white people.

But on Wednesday, June 10, Amazon shocked civil rights activists and researchers when it announced that it would place a one-year moratorium on police use of Rekognition. The move followed IBM's decision to discontinue its general-purpose face recognition system. The next day, Microsoft announced that it would stop selling its system to police departments until federal law regulates the technology. While Amazon made the smallest concession of the three companies, it is also the largest provider of the technology to law enforcement. The decision is the culmination of two years of research and external pressure to demonstrate Rekognition's technical flaws and its potential for abuse. "It's incredible that Amazon's actually responding within this current conversation around racism," said Deborah Raji, an AI accountability researcher who coauthored a foundational study on the racial biases and inaccuracies built into the company's technology.

Are machine-learning algorithms biased, wrong, and racist? Essentially rule-based structures for making decisions, machine-learning algorithms play an increasingly large role in our lives. They suggest what we should read and watch, whom we should date, and whether or not we are detained while awaiting trial. Their promise is huge–they can better detect cancers. But they can also discriminate based on the color of our skin or the zip code we live in.

This week IBM, Amazon, and Microsoft all said they would halt sales of facial recognition to US police and called on Congress to impose rules on use of the technology. A police reform bill introduced in the House of Representatives Monday by prominent Democrats in response to weeks of protest over racist policing practices would do just that. But some privacy advocates say its restrictions aren't tight enough and could legitimize the way police use facial recognition today. "We're concerned," says Neema Guliani, senior legislative counsel for the ACLU in Washington, DC, citing evidence that many facial recognition algorithms are less accurate on darker skin tones. She urges a federal ban on facial recognition "unless and until it can be used in a way that respects civil liberties;" Guliani says it's not clear that that is possible.

Amazon announced Wednesday that it is pausing police use of its facial recognition software for one year following nationwide pressure on tech companies to address potential bias. While Amazon did not specify a reason for its decision, racial injustice has been at the forefront of ongoing protests in the wake of the death of George Floyd, who died May 25 after a white Minneapolis police officer pressed his knee into the handcuffed black man's neck for nearly nine minutes. "We've advocated that governments should put in place stronger regulations to govern the ethical use of facial recognition technology, and in recent days, Congress appears ready to take on this challenge," Amazon said in a statement posted to the company's blog website. Researchers have long criticized the technology for producing inaccurate results for people with darker skin, while other studies have shown technological bias against minorities and young people. Nicole Ozer, technology and civil liberties director with the American Civil Liberties Union of Northern California, said in a statement that the organization was "glad the company is finally recognizing the dangers face recognition poses to Black and Brown communities and civil rights more broadly," but that it was not enough to combat the threat to "our civil rights and civil liberties."

Apple's Siri is supporting the Black Lives Matter movement by providing users who say'All Lives Matter' with a link to learn more about human rights initiative. When speaking the phrase to Siri, it will respond, 'All Lives Matter' is often used in response to phrase'Black Lives Matter,' but it does not represent the same concerns,' and then the technology prompts users to visit BlackLivesMatter.com. The update is to align with other businesses and organizations that are showing solidarity for the movement with worldwide protests following the death of George Floyd who was killed while in police custody last month. Apple also joins Amazon and Google, which have also updated their smart voice assistance to explain the Black Lives Matter movement to users. Apple's Siri is supporting the Black Lives Matter movement by providing users who say'All Lives Matter' with a link to learn more about human rights initiative Floyd was killed on May 25 in Minneapolis, Minnesota when Officer Derek Chauvin knelt on his neck until he lost consciousness – autopsies have since deemed the death a homicide.

Many important decisions in societies such as school admissions, hiring, or elections are based on the selection of top-ranking individuals from a larger pool of candidates. This process is often subject to biases, which typically manifest as an under-representation of certain groups among the selected or accepted individuals. The most common approach to this issue is debiasing, for example via the introduction of quotas that ensure proportional representation of groups with respect to a certain, often binary attribute. Cases include quotas for women on corporate boards or ethnic quotas in elections. This, however, has the potential to induce changes in representation with respect to other attributes. For the case of two correlated binary attributes we show that quota-based debiasing based on a single attribute can worsen the representation of already underrepresented groups and decrease overall fairness of selection. We use several data sets from a broad range of domains from recidivism risk assessments to scientific citations to assess this effect in real-world settings. Our results demonstrate the importance of including all relevant attributes in debiasing procedures and that more efforts need to be put into eliminating the root causes of inequalities as purely numerical solutions such as quota-based debiasing might lead to unintended consequences.