Representatives of 66% of companies predict that, in the coming years, they will increasingly make decisions using artificial intelligence and machine learning. In order to do so responsibly, they should consider ethical aspects, and the obstacle to such actions may be data bias. A survey conducted by Progress reveals that the boards of most firms understand the significance of this phenomenon and believe it’s prevalent in their businesses, but they encounter problems with its effective neutralization.
Artificial Intelligence is rapidly taking over various industries – from healthcare and finance to e-commerce and production. Companies are increasingly using AI to make decisions based on rules related to, for example, determining creditworthiness or customer segmentation. Benefits such as improved work efficiency, cost reduction, and accelerated business development are noticeable to business owners and consumers who get more efficient and faster service.
However, it is worth remembering that the processes behind the use of AI have a certain flaw – the data used to power these systems is not neutral. There is always some form of created bias, which results from the nature of the medium from which the data was obtained. Artificial intelligence reflects and reinforces the prejudices of its creators, raising ethical concerns about privacy, security, stereotypes, and objective assessment.
– Data bias occurs when, due to errors present in it, a certain group is favored at the expense of another. Usually, the result is the algorithm making unfair decisions because the available data does not accurately reflect the attitudes of the entire represented population. Bias causes differences between the model’s predicted and actual values. These prejudices can be based on stereotypes, not on specific knowledge of individuals or circumstances – says Niklas Enge, Regional Director of Nordics and Poland at Progress.
How to prevent data bias?
The study conducted by Progress revealed that 78% of people responsible for making business and IT-related decisions believe that various types of bias present in the data will become a greater problem with the increase in the use of AI/ML. However, only 13% currently deal with this phenomenon and have developed a constant process of assessing its scale. The biggest barriers respondents see are lack of awareness of data bias, understanding how to identify biases, and lack of access to expert resources, such as consultations with data scientists.
To counteract the phenomenon of data bias, one should apply an approach that will result directly from the policy and organizational culture in the company. Artificial intelligence and machine learning are increasingly integrated with business operations. Such an approach should include establishing ethical standards, best practices for data collection and model development, regular model assessment, ongoing monitoring, and cooperation between all parties involved in its use.
Source: Progress’ “Data Bias: The Hidden Risk of AI” study.
