Fraud Detection: This is Why Artificial Intelligence Needs Human Intelligence

Artificial intelligence (AI) will obviously be something big in the future, including in the field of fraud detection, for example in e-commerce, banking and other business types. AI is a valuable tool against fraud, but this technology still needs human input and insight in creating comprehensive solutions to deliver the best results. Here are some reasons why artificial intelligence needs human intelligence.

Automatic screening can lead to false decline

The algorithm is very useful for quickly identifying potential fraud, but other factors can lead to the false machine identification. For example, consumers who make purchase transactions while travelling abroad. The transaction may be declined by the bank because it was marked as potential fraud.

The false decline is certainly hurting the merchant. MasterCard and Javelin found that 32% of consumers who experience false decline never shop at the merchant again. That way, there is potential for future income loss and the potential loss on attracting new customers due to a rejection.

Taking into consideration the cost of lost purchases in the future as well as higher relative costs to attract new customers rather than retaining existing ones, such false decline are likely to have a major impact on merchants. To prevent false identification and make AI machines more contextual in detecting potential fraud, businesses must combine AI algorithms with data collected by human analysts.

AI machine learns a variety of fraud patterns from human analysts.

Different business segments constitute different patterns of fraud. For example, botnet fraud patterns are highest in the digital business segment – digital banking, online applications, and so on. The pattern of friendly fraud tends to be high on the luxury goods segment. An effective algorithm will take these patterns and their change into account in each market segment and geography. How can algorithm take it into account? Certainly, with input and insight from experienced analysts in each segment.

In essence, human analysts have an in-depth knowledge of the client, of a segment landscape where there are certain fraud patterns, have the ability to communicate directly with parties involved in the false decline, and have intuition and experience in recognizing new patterns of fraud. Ultimately, human intelligence is still needed to improve AI machines.

Source:

http://dataconomy.com/2016/05/machine-learning-fraud-artificial-intelligence-isnt-enough/