Predictions obtained by, e.g., artificial neural networks have a high accuracy but humans often perceive the models as black boxes. Insights about the decision making are mostly opaque for humans. Particularly understanding the decision making in highly sensitive areas such as healthcare or fifinance, is of paramount importance. The decision-making behind the black boxes requires it to be more transparent, accountable, and understandable for humans. This survey paper provides essential definitions, an overview of the different principles and methodologies of explainable Supervised Machine Learning (SML). We conduct a state-of-the-art survey that reviews past and recent explainable SML approaches and classifies them according to the introduced definitions. Finally, we illustrate principles by means of an explanatory case study and discuss important future directions.

The origin of the word'bias' has never been quite certain. Linguists reckon that the antecedent of bias is the Old French word'biasi' which meant at an angle or oblique. It came to mean'a one-sided tendency of the mind'. In the old English game of bowls, the ball had asymmetrical weight or bias, which made it roll in a curved line. This is how bias came to be the favoured word for having a disproportionate weight in favour of or against an idea or person.

When the UK government decided to cancel school exams due to the coronavirus pandemic, they gave examination regulators Ofqual a challenge: allocate grades to students anyway, and make sure the grades given out this year are equivalent in standard to previous years. Ofqual's solution was to create an algorithm – a computer program designed to predict what grades the students would have received if they had taken exams. Unfortunately, when the computer-generated grades were issued, 40 per cent of A-Level students got lower grades than their teachers had predicted. Promised university places were withdrawn. Lawyers offered to take legal action against Ofqual.

BEGIN ARTICLE PREVIEW:   Artificial Intelligence (AI) is bringing a technological revolution to society. The new emerging digital world carries with it a scary thing: Artificial Intelligence (AI) bias. It is a pressing concern over as AI is becoming extremely powerful and at the same time with a lot of discriminatory thoughts like humans. Human bias is not new. The recent protests across the globe on racial discrimination are a pure example that bias is a major threat to human society. Discrimination is not just related to race, it also concerns gender inequality. Women leaders like New Zealand Prime Minister Jacinda Arden and San Francisco Mayor London Breed are receiving recognition for their rapid action in tackling and controlling Covid-19 s

This work proposes the continuous conditional generative adversarial network (CcGAN), the first generative model for image generation conditional on continuous, scalar conditions (termed regression labels). Existing conditional GANs (cGANs) are mainly designed for categorical conditions (e.g., class labels); conditioning on regression labels is mathematically distinct and raises two fundamental problems: (P1) Since there may be very few (even zero) real images for some regression labels, minimizing existing empirical versions of cGAN losses (a.k.a. empirical cGAN losses) often fails in practice; (P2) Since regression labels are scalar and infinitely many, conventional label input methods are not applicable. The proposed CcGAN solves the above problems, respectively, by (S1) reformulating existing empirical cGAN losses to be appropriate for the continuous scenario; and (S2) proposing a naive label input (NLI) method and an improved label input (ILI) method to incorporate regression labels into the generator and the discriminator. The reformulation in (S1) leads to two novel empirical discriminator losses, termed the hard vicinal discriminator loss (HVDL) and the soft vicinal discriminator loss (SVDL) respectively, and a novel empirical generator loss. The error bounds of a discriminator trained with HVDL and SVDL are derived under mild assumptions in this work. Two new benchmark datasets (RC-49 and Cell-200) and a novel evaluation metric (Sliding Fr\'echet Inception Distance) are also proposed for this continuous scenario. Our experiments on the Circular 2-D Gaussians, RC-49, UTKFace, Cell-200, and Steering Angle datasets show that CcGAN can generate diverse, high-quality samples from the image distribution conditional on a given regression label. Moreover, in these experiments, CcGAN substantially outperforms cGAN both visually and quantitatively.

We can safely defer the discussion about whether artificial intelligence will eventually take over board functions. We cannot, however, defer the discussion about how boards will oversee AI -- a discussion that's relevant whether organizations are developing AI systems or buying AI-powered software. With the technology in increasingly widespread use, it's time for every board to develop a proactive approach for overseeing how AI operates within the context of an organization's overall mission and risk management. According to McKinsey's 2019 global AI survey, although AI adoption is increasing rapidly, overseeing and mitigating its risks remain unresolved and urgent tasks: Just 41% of respondents said that their organizations "comprehensively identify and prioritize" the risks associated with AI deployment. Get monthly email updates on how artificial intelligence and big data are affecting the development and execution of strategy in organizations.

OTTAWA - Artificial intelligence must be regulated to protect Canadians' privacy and human rights, a federal watchdog says. In issuing new recommendations for regulating AI Thursday, Canada's privacy commissioner Daniel Therrien said he is calling for legislation to regulate the use and development of AI systems. Such legislation will help to reap the benefits of AI while upholding individuals' fundamental right to privacy, he said in a statement. Therrien said these changes should entrench privacy as a human right and a necessary element for the exercise of other fundamental rights. AI models analyze and try to predict aspects of human behaviour and interests that can be used to make automated decisions about people.

Everyone knows about the fatal 2018 crash in Tempe, Arizona where an Uber SUV driving in semi-autonomous mode failed to avoid hitting Elaine Herzberg as she was walking her bike across a darkened street. This past September, 2 years later, the "backup" driver was charged with negligent homicide by a Grand Jury. The first issue I will address might seem like the biggest, but only from a marketing and stature perspective. In playing a game of who has advanced AI more than anyone else in current month, tech companies have been allowed to make broad claims in the media that help drive traffic to websites and build market share (which we might call brand politics). Fast and loose with details is par for the course, but as these companies often live from one round of funding to the next, it allows for a culture within the workforce of "just getting results."

A growing number of tech companies are placing their bets on algorithms to reinvent talent acquisition and create a more inclusive workforce. In some cases, this might mean entirely removing traditional aspects of the hiring process. Introduced in the nineties, applicant tracking systems (ATS), were created to help HR professionals organize the surge of applications that resulted from the growing use of the internet. Over the last several decades, ATS became increasingly advanced, using algorithms to sift through thousands of resumes based on various data. The promise was efficiency and blind hiring, but the algorithms have proven to perpetuate structural inequities in hiring.

In this paper, I argue that counterfactual fairness does not constitute a necessary condition for an algorithm to be fair, and subsequently suggest how the constraint can be modified in order to remedy this shortcoming. To this end, I discuss a hypothetical scenario in which counterfactual fairness and an intuitive judgment of fairness come apart. Then, I turn to the question how the concept of discrimination can be explicated in order to examine the shortcomings of counterfactual fairness as a necessary condition of algorithmic fairness in more detail. I then incorporate the insights of this analysis into a novel fairness constraint, causal relevance fairness, which is a modification of the counterfactual fairness constraint that seems to circumvent its shortcomings.