AI ethics in identity services
As a provider of a range of identity services utilising AI technology, including biometric identification and verification, Onfido are fully aware of the ethical complexities of employing AI. Algorithmic bias, for one, is a recognised phenomenon, and we have invested heavily in efforts to mitigate it and to ensure that our products are fair and inclusive for all data subjects.
Onfido recently participated in the ICO’s Regulatory Sandbox with the aim of ensuring that the research we conducted in relation to reducing algorithmic bias was carried out properly and respected the rights and freedoms of individuals when processing their personal data.
In line with the objectives outlined in our Sandbox Plan, the team trialled and developed methodologies to group and label data, tested the performance of Onfido’s facial recognition, retrained the models, and measured the performance changes to those models. We plan on continuing to do so in the future to constantly improve our AI software’s performance.
To trust AI-enabled services, consumers will require reassurance services in the short and medium term. This will consist of robust standards and requirements to ensure privacy, security and protection against fraud, as well as mitigating issues like racial bias.
There are good reasons, however, that the broad implementation of AI technologies across sectors such as financial services is not just inevitable, but desirable. Digital, AI-assisted identity verification and authentication, to use our specialty as an example, has the potential to expand access to financial services. Even prior to the pandemic, financial services were moving online – bank branch closures have been accelerating in the UK for years.
Consumers need both to maintain that access to online services, but also the certainty that their data and assets are safe and secure. For that, providers need to check ID. The only feasible way of doing this at the speed users want is through AI-assisted, biometric verification and authentication such as Onfido’s selfie technology.
AI identity verification and authentication thus facilitates the inclusion of new or difficult to access consumers, enabling them to contribute to the economy while ensuring that they and the businesses they use are protected. With such AI-assisted RegTech in place to ensure that any risk is managed, businesses can open their doors to more users. This promotes digital access to a broad cross-section of society and promotes responsible innovation.
We are therefore clear that to balance the need for AI technologies and concerns about the implementation of AI, a multi-level, risk-based approach is appropriate. This will promote digitalisation, provide opportunities for consumers to use AI products to access financial services and therefore improve financial wellbeing across the board.
As a company, we deal on an everyday basis with ethical concerns around AI. Onfido’s technology has reduced to an absolute minimum the bias in our AI biometric identification, for which we were recognised in the CogX Awards for “Best Innovation in Algorithmic Bias Mitigation” and “Outstanding Leader in Accessibility”, and we were Highly Commended in the SC Europe Awards 2020 for “Best Use of Machine Learning”.
Matt Peake, Global Head of Policy, Onfido | Document ID & Facial Biometrics Verification