Algorithmic bias is a widely recognized problem: unfair human assumptions and judgments are often encoded in algorithms, resulting in biased targeting of certain groups. In a field such as advertising, the existence of such biases has been confirmed by IBM teams, using the AI Fairness 360 (AIF360) toolbox developed by Big Blue.
To do away with these unconscious biases, IBM has just released its free Advertising Toolkit for AI Fairness 360, an open-source tool featuring 75 fairness metrics and 13 algorithms to help identify and mitigate bias in sets of data. This toolkit also includes a playbook and sample code for ease of use. It is designed to help organizations better understand the presence and impact of bias on their advertising campaigns, as well as the composition of their audience.
IBM does not intend to stop at this toolbox alone to tackle the biases in the algorithms head-on. The American giant invites companies and organizations, particularly in the field of digital advertising, to sign its Advertising Fairness Pledge, with the aim of improving the image of algorithms and rationalizing their use. “Used correctly, data can help brands personalize consumer engagement and identify the most relevant touchpoints,” said Mark Read, CEO of WPP, a signatory advertising agency of this initiative, in a press release.
Source: ZDNet.com