IEEE 7003-2024

Computer algorithms and analytics are playing an increasingly influential role in government, business and society. They underpin information services and autonomous intelligent systems (AIS) including but not limited to artificial intelligence applications that involve symbolic and subsymbolic technologies and their hybridization. These technologies are having a direct and significant impact on human lives across a broad socioeconomic, political and cultural spectrum. Algorithms enable the exploitation of vast and varied data sources from public and private spheres to support human decision-making and actions that serve the diverse interests of the societies and economies in which they operate. However, alongside the benefits, their use is not without maleficent risk. This standard describes processes and methodologies to help users address issues of bias in the creation of algorithms and models. Elements include but are not limited to: criteria for the selection of data sets; guidelines on establishing and communicating the application boundaries for which the AIS has been designed and validated to guard against unwanted consequences arising from out-of-bounds application of an AIS; suggestions for user expectation management to help mitigate unwanted bias due to incorrect interpretation of systems outputs by users (e.g., correlation vs. causation).

This standard is designed to provide individuals or organizations creating an AIS, certification-oriented processes and methodologies to produce clearly articulated accountability and clarity around how an AIS targets, assesses and influences the stakeholders of said AIS. This standard enables AIS creators to define, measure and communicate how bias is used in the AIS to users and regulatory authorities, and show that best practices were used in the design, development, testing and evaluation of the AIS to avoid unwanted differential impact on stakeholders.

New IEEE Standard - Active. The processes and methodologies to help users address issues of bias in the creation of algorithms are described in this standard. Elements include but are not limited to: criteria for the selection of validation data sets for bias quality control, guidelines on establishing and communicating the application boundaries for which the algorithm has been designed and validated to guard against unintended consequences arising from out-of-bound application of algorithms, and suggestions for user expectation management to help mitigate bias due to incorrect interpretation of systems outputs by users (e.g., correlation vs. causation). (The PDF of this standard is available at no cost at https://ieeexplore.ieee.org/browse/standards/get-program/page/series?id=93)