Do you know how AI systems work? Here’s why explainable AI is essential.

Five benefits of explainable AI ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌   ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ 

TODAY'S NEWS. TOMORROW'S INSIGHTS. A daily newsletter from McKinsey & Company   Edited by Belinda Yu   Editor, Atlanta Exploring explainable AI In the news • Bikinis and business suits. Images in data sets used to train AI systems reflect human bias, researchers have found. When provided a picture of a woman's face and asked to fill in the rest of the image, one computer vision model auto-completed more than half of the images with bikinis or revealing attire. Forty-two percent of men's faces, however, were auto-completed with business attire such as suits. But even though computer vision models are involved in making critical decisions (for example, in evaluating job candidates), there's no standardized process for transparency that gives people insight into how AI is used. [Brookings] • Four features of AI. To be successful, AI systems must gain the trust of consumers, employees, and other stakeholders. But AI models have four features that may make it difficult for people to trust them, MIT researchers find. One is that it's difficult to understand if an AI model is accurate and ethical, because it depends on complicated mathematical and statistical calculations. By establishing practices for tracing decisions, such as using charts and other visuals, developers can make clearer how the models actually work. [MIT] Organizations that establish digital trust among consumers are more likely to see their annual revenue and EBIT grow at rates of 10% or more. On McKinsey.com • Explainability is essential. Companies increasingly depend on AI to make decisions, and consumers need to know that these decisions are fair, accurate, and trustworthy. Firms that show how their AI-driven recommendations are derived (that is, "explainability") win over regulators and the public, while adding to the company's bottom line, say McKinsey senior partner Alex Singla, Andreas Kremer, and colleagues. Companies that attribute at least 20% of EBIT to their use of AI are more likely to follow best practices that enable explainability. • Opening the black box. Mastering explainability in AI is more important than ever, but it's also getting harder. Too often, advanced machine learning models remain a black box. The more sophisticated an AI system is, the tougher it is to understand. AI systems constantly learn from the data that they ingest, sometimes delivering outputs within fractions of a second. To demystify AI models, firms need to establish a governance framework and invest in the right tools and talent. See five ways that explainable AI can benefit organizations. See the value of explainable AI Was this forwarded to you? Sign up here. Or send us feedback—we'd love to hear from you. Follow our thinking This email contains information about McKinsey's research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy. You received this email because you subscribed to the On Point newsletter. Manage subscriptions | Unsubscribe Copyright © 2022 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007