Our world is a digital place. The more time passes, the more interactions we will have online. For businesses, this presents a problem. How are they to be truly sure who they are doing business with? This is especially important for banks and financial institutions that musts uphold strict compliance standards. To solve this dilemma, a variety of identity verification solutions have been introduced to the market. In this article we will discuss why automated ID checks provide better match results than ID verification solutions that rely on manual review.
Types of ID Verification: Automated ID Checks vs Human Review Checks
In digital environments, you cannot be sure who you are interacting with. Someone could say they are John Smith from Johnsonville, Missouri when in fact they are Alexander Petrov from Moscow, Russia. Without some sort of check though, there is no way to find out that John is Alexander online.
This is where identity verification comes in. In online situations where knowing your customer is important, identity verification solutions can help solve that problem. The most robust identity verification services will ask customers for a photo of a government issued ID, extract the personally identifiable information (PII) from that document, crosscheck that information with information stored in databases from the customer’s country of origin and then perform a biometric test to match the user with the ID.
Not all identity verification providers perform these tasks the same way though. Some, like IDMERIT, will perform automated identity checks. These checks will scan and verify a customer’s ID and then extract the PII from the ID and automatically run it through a secure API into a database from their country of origin. This happens instantaneously and does not endanger the PII.
Other providers will take photo of the national ID and send it overseas for manual review. These providers rely on trained staff to verify that an ID is real. From there they can check the PII in databases from customer’s country of origin. These checks are fast, but not instantaneous. They also endanger the PII being checked, which we’ve discussed previously on this blog.
The Problem with Human Review
Automated ID checks provide better match results than those done with manual review. Automated identity verification relies on artificial intelligence (AI) and machine learning (ML) to create robust identity verification software. The reason for this is that in manual review, errors are prone to happen.
- Humans have to worry about implicit bias
Humans are subject to perform implicit bias, which the Kirwan Institute for the Study of Race and Ethnicity describes as, “the attitudes or stereotypes that affect our understanding, actions and decisions in an unconscious manner. [It is] activated involuntarily without awareness or intentional control. It can be either positive or negative and everyone is susceptible.” They state that implicit bias presents unique challenges to effective law enforcement because it can alter where investigators look for evidence and how they analyze it without their awareness or ability to compensate.
Automated solutions that use biometric technology to fully authenticate an individual remove these issues associated with manual review because they are performed by a sophisticated software that does not have implicit bias programed into its system.
2. Humans have limits to how much they can learn
There are 195 countries in the world. Each country has different identification requirements, sometimes having a variety of IDs for different reasons. This means that anyone who must validate IDs must do so for each identity document within these 195 countries. This is not only unrealistic to expect from people, it is unlikely. Sure, a person can be trained to identify valid IDs for all 195 countries, but that does not guarantee they will be 100% accurate when doing so.
Say, for example, a person performing human ID review is given a US passport to review. They will likely be able to do this easily because US passports are a common form of identification for many people. In contrast, if a person is given a Latvian ID for review, this will likely be harder for them to identify. Latvia is a small country, so it has less IDs floating around for identification. A person could be trained in validating Latvian IDs, but they may not have any real-world experience doing so.
Automated ID verification solutions take away the guesswork from validating IDs. Computer programs do not have a limit to how much information they can learn. If they are programmed to do a certain task, they will do it. This means automated ID validation programs have the capacity to accurately verify IDs from all 195 countries around the world. They also can use machine learning to perform checks with even more accuracy as IDs change over time, or criminals adopt new ways to create fake identification documents.
Conclusion
Artificial intelligence is not perfect yet is continuously being improved with machine learning and human intervention. In the identity space, AI has created automated ID verification programs that offer better match results than those that use human review. During human ID review, match results can be skewed by implicit bias and how much knowledge a person has about the ID they are checking. Plus, human review puts PII at risk. For identity verification, it is likely that your company will achieve better match results with automated ID checks than those that rely on human review.
