The eBook makes a strong case for facial recognition technology in itself, noting that it has important applications in law enforcement – from identifying the victims of human trafficking to helping to resolve police investigations – and major benefits in consumer-facing areas like financial transactions, airport screening, and securing mobile devices. But it also makes the case for implementing certain ethical safeguards in the development of facial recognition technology, acknowledging that there are legitimate concerns about issues like racial bias and privacy protection.
This is a matter of particular interest to Oosto, which has been vocal about the need for strong ethical standards when it comes to facial recognition and AI. And the company recently demonstrated its commitment to this ideal with its announcement of a strategic shift that will see Oosto strive to expand the use cases of its own technology beyond surveillance and security and into new areas like smart cities, healthcare, and payments.
This pivot was announced alongside a name change from its previous moniker, “AnyVision”, and a new partnership with Carnegie Mellon University’s CyLab Biometric Research Center.
Now, Oosto is laying out its ethical views and practices in detail in its new eBook. The company says it is guided by six key principles governing the ethical development of its facial recognition technology: fairness, transparency, accountability, non-discrimination, lawful surveillance, and notice and consent.
In practice, adherence to those principles means undertaking a number of important measures. For example, Oosto has built into its solution a “GDPR mode” (referring to the European Union’s data privacy regulation) that is designed to effectively blur the faces of any individual detected by its facial recognition technology who is not explicitly on a given organization’s watchlist.
The company also specifically avoids using the Fitzpatrick Skin Type scale in its solution, opting instead of an AI-driven system that Oosto says can capture millions of skin tone categories. This, together with highly granular color coding, “ensures a higher facial recognition accuracy and dramatically improves our ability to reduce demographic bias,” the company says.
There are many more examples like these in the full eBook, which is available for free online. For many in the biometrics industry and those who are considering integrating or deploying such technology, it’s a valuable resource as the discourse over ethics and AI continues to evolve.
November 1, 2021 – by Doug OGorden