Strategic framework for the ethical use of face biometrics proposed
The World Economic Forum has published its Framework for Responsible Limits on Facial Recognition, a white paper that the organization says is the first framework for ethical and safe use of facial recognition.
The goal is to help create standards for engineers and policy makers that protect human rights while making it less risky for new players to enter the market. Standards can make investment in technology and technology businesses safer, which draws more entrepreneurs and investors.
The forum is an international organization seeking a maximum of public-sector and private-sector cooperation. Members identify concerns and opportunities that could benefit from coordinated public and private attention.
Four primary steps, identified by the forum’s artificial-intelligence team, are designed to ensure the responsible design and use of facial recognition.
First, organizations are encouraged to assemble a variety of stakeholders including politicians, recognition system designers, regulators, academics and non-governmental communities.
This group will create “principles for action,” which define what is responsible use of face recognition. Next, the work group should design methodologies unique to each use case in order to support the product teams developing systems.
Third, group members should create (partly based on the results of the first step) an assessment questionnaire written to determine if the system designed, indeed, complies with the principles of action — if it is responsible in terms of ethics and safety. Finally, the work group needs to employ a thrusted third party to create an audit framework that can be used to validate compliance with the principles of action.
There are more technical standards applied to facial recognition. One, ISO/IEC 19794-5 sets out rules for digital-image attributes. This standard sets out preferred or required data formats, scene constraints like a person’s pose or expression, photographic properties such as lighting and focus, and digital-image attributes including image resolution and size.