Author: Steve Durbin
Publisher: World Economic Forum
Publication Year: 2022
Summary: The following article discusses how although artificial intelligence (AI) promises to improve and streamline business operations and everyday life, there are proportional increasing concerns about the implementation of the technology. In order to counteract possible negative effects of AI, data scientists need to account for “inbuilt prejudices” that arise from training data or human bias. Additionally, bias can occur in the usage or implementation of the product when they are used unintendedly. Durbin proposes 5 different risk mitigation strategies. The first is to translate ethics into metrics. When ethical standards can be translated into clear principles or metrics it helps make ethical goals tangible and easier to follow. Second is to become familiar with the sources of bias in algorithms, especially within industries, professionals can learn to preempt many of these biases. Next, is to balance autonomy with human oversight: these algorithms should be the end all be all. Fourth, is to empower employees and elevate responsible AI. Finally AI can be leverage to tackled discrimination via procedural checks.