Data bias is becoming an increasingly pressing issue for businesses that leverage artificial intelligence and machine learning, but many organizations struggle to address it effectively.
Two-thirds of executives think there is currently data bias in their organizations, according to a global study sponsored by Progress and conducted by Insight Avenue.
Data bias has become a more prominent concern for companies using artificial intelligence (AI) and machine learning (ML) to analyze and make sense of their data. With everything going digital, companies now have access to a wealth of information—in most cases, too much information to know where to start.
AI and ML can draw actionable insights from big data, helping companies make better business decisions.
It’s no wonder, then, that more businesses have begun to use and rely on AI: The study showed that 66% of organizations anticipate becoming more reliant on AI and/or ML for decision making.
While the intention is to make businesses smarter and more efficient, the use of AI has also come with some inadvertent consequences—data bias being a major one.
Decisions made with biased data can negatively impact finance, IT, digital, operations, sales and strategy. Even worse, data bias can lead to poor customer experiences, damage companies’ reputations and set back inclusion and diversity efforts.
“Every day, bias can negatively impact business operations and decision making—from governance and lost customer trust to financial implications and potential legal and ethical exposure,” said John Ainsworth, EVP and General Manager, Application and Data Platform, Progress.
“We put our customers at the center of everything we do and as we explore all that AI/ML can do, we want to ensure our customers are armed with the right information to make the best decisions to drive their business forward,” he added.
At Progress, we wanted to get a sense of how widespread data bias is, the actions businesses are taking to prevent and address bias, the barriers of addressing it and the implications of unchecked bias. In partnership with U.K.-based Insight Avenue, we commissioned a worldwide survey of 640 business and IT leaders who use data to make decisions or are planning to use AI or ML to support decision-making. All executives led companies with 500+ employees.
The study revealed that while 78% believe data bias will become a bigger problem as AI/ML use increases, just 13% are currently addressing data bias and have an ongoing evaluation process to weed it out. Additionally, more than half of respondents consider lack of awareness of potential biases as a barrier to addressing data bias.
Read more highlights from the study below or download a copy of the study to get a complete picture of the state of data bias in business.
Recognizing the Threat of Data Bias
While companies vary in their strategies to address potential data bias, businesses are aware of the risk and consequences data bias can bring.
77% of respondents acknowledge that they need to be doing more to understand and address bias in their organization, and 76% say there are wider social implications if companies do not adequately address the issue of data bias.
Most leaders (78%) are aware that as AI becomes more widely used, the problem will only intensify. With that in mind, 67% of execs believe their organization had evaluated technology to tackle data bias, and 40% said data bias was a consideration when evaluating AI/ML vendors.
It will likely take a combination of people, tools, training and policy to avoid data bias: 76% recognize that data bias is best addressed centrally across the organization instead of having siloed departments handle the issue.
Where Organizations Can Improve
When it comes to combating data bias, organizations have several obstacles to overcome before they can make progress. Top barriers to addressing data bias include a lack of awareness of potential biases (51%), a lack of understanding about how to identify bias (43%) and a lack of expert resources (31%).
Just 9% of respondents said they don’t see data bias as an issue, indicating that inaction can be attributed to struggles with planning and execution, rather than a failure to recognize the threat of data bias.
77% of respondents said their organizations still need to do more to understand data bias. Execs believe the most effective measures will be technology and tools (65%), more training (59%) and adjusting their strategy and vision (49%).
How Businesses Can Address and Avoid Data Bias
As more organizations begin to rely on AI and ML, the need to address potential data bias becomes more urgent. Companies will need to have plans and processes in place to identify and prevent data bias, and entities will need to recognize how it can threaten every aspect of a business.
Organizations will need to look at all parts of a project, from hiring and team diversity to training and technology. Data bias can impact day-to-day decisions at any company, and it can have a detrimental effect on its victims. The ones who lead technology and training efforts will need to ensure their work promotes equity and fairness in the workplace.