Jumio explained the risk of bias embedded in biometric machine learning models and artificial intelligence-based systems, and suggested practical methods to reduce these biases to minimize the impact on people’s daily lives.
In an article titled “5 Practical Ways To Reduce AI Bias In Online Identity VerificationJumio also looked at the different forms of AI bias and the demographic traits they manifest through. The article points out that facial recognition processes are one of the most prone to racial bias.
According to Jumio, customers should be able to ask questions of potential service providers to better understand how they deal with demographic biases.
The five things companies should find out from biometrics vendors are the size and representativeness of the vendor’s biometric training database, where the data is coming from creating the training datasets, how the datasets are tagged, the type of quality controls governing the tagging process, and the diversity of the team that develops the algorithms, Jumio says.
The blog post says that if all of these are taken into account, the biases that are often unintentional in AI algorithms will be significantly reduced. He also cites the 2020 Gartner Market Guide for Identity Proofing and Affirmation which suggests that “by 2022, more than 95% of tenders for document-centric identity verification will contain clear requirements regarding minimizing demographic bias, an increase from at less than 15% today â.
Jumio postulates that many leading identity verification solutions leverage AI and machine learning biometrics to assess the digital identity of remote users – and, unfortunately, these algorithms are also susceptible to demographic biases that include race, age, sex and other characteristics.
Jumio notes that while demographic bias is largely an unconscious act, as many providers of facial recognition solutions don’t necessarily know when they create the algorithm that it will likely produce incorrect results at uneven rates, the The impact of bias on a business can be far-reaching as it could tarnish their image or even cause litigation from users.
Meanwhile, Jumio says he is making efforts in his day-to-day work to address concerns about his AI algorithms and processes. In this regard, it has developed a new reference guide to achieve this goal and that includes what it says are its large and representative datasets, realistic data production, quality control and governance, among others. .
precision | IA | algorithms | biometric identification | biometrics | digital identity | facial recognition | identity verification | coaching