Where is your business on the AI ​​adoption curve? Take our AI survey discover.

Technologists, social scientists and others are rightly concerned about artificial intelligence biases. As technology continues to infiltrate digital systems that impact our lives, making sure it doesn’t discriminate based on race, gender, or other factors becomes a top priority.

Improving social justice is also important to the business, but the ability to succeed in a competitive market is just as important. And the fact remains that AI biases are not only detrimental to society, they can also lead to poor decision-making that can actually hurt business processes and profitability.

A bad reputation hurts

Assistant Professor at USC Kalinda Ukanwa recently highlighted the myriad of ways poorly trained algorithms that produce biased results can lead organizations down the wrong path. On the one hand, word of mouth can quickly spread stories of unfair treatment in a given community, resulting in lost opportunities and decreased sales. Additionally, his research has shown that too much reliance on “group-sensitive” algorithms that attempt to discern an individual’s behavior based on an assignment to a particular group can yield short-term, but ultimately, results. falling behind the AI ​​running on a “blind group”.

Another key source of bias induced friction between organizations and both its customers and employees is when managerial interaction becomes necessary, such as in a call center. PLEASANT, a developer of robotic process automation (RPA) for call centers, recently developed a framework to ensure that AI remains useful and friendly to users and employees, which in turn builds loyalty to the brand and a positive buzz on social networks. Among the key points is the need to focus on achieving positive results in any interaction and to train robots to be devoid of race, gender, age or any other biases in order to produce a vision. totally agnostic of humanity.

Data scientists categorize AI biases in several areas, such as sampling and selection bias, but one of the most detrimental to business is predetermination bias, according to author and entrepreneur Jedidiah Yueh. This is where AI (as well as humans) tries to prepare for the future they expect, not necessarily the one they get. Understandable but, in an age when AI itself produces a radically unpredictable future, it is fraught with dangers as it inhibits innovation and the ability to remain flexible in a changing environment. Unfortunately, predetermination is often hardwired into the ETL process itself, so reversing it requires more than changes in AI training.

Exploiting prejudices for good

Business leaders should also avoid the trap of thinking any bias is wrong, says Dr Min Sun, Chief AI Scientist at Appier. In many marketing scenarios, it can be helpful to build bias into AI algorithms if you’re trying to determine buying trends for, say, single, older women. The trick is to make sure that decision makers are aware of the presence of these biases and can visualize the resulting data in the appropriate way. To do this successfully, it is important not to introduce bias into the training model itself, but into the data on which the model is trained.

The main problem companies face when trying to eliminate AI bias is that today’s data governance policies are not geared to this new way of doing things, according to technical author Tom Taulli. Too often AI projects lack the coordination needed to eliminate bias and produce effective ROI, and this usually stems from the isolation that exists between data science and application development teams. While there is always a temptation to automate all functions in a given data process, governance should be an exception as only a practical and intuitive approach can ensure that goals are met in a rapidly changing environment.

With a prejudice so prevalent in AI projects already deployed, business leaders would be wise to take a close look at where and how it is used – not only in the interest of the greater social good, but also for their own. economic reasons. Nowadays, trust is a rare and precious commodity and once lost it is not easy to regain it. The last thing an organization should want is to be tarnished by a biased label caused by a poorly trained AI process.


VentureBeat’s mission is to be a digital public place for technical decision-makers to learn about transformative technology and conduct transactions. Our site provides essential information on data technologies and strategies to guide you in managing your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the topics that interest you
  • our newsletters
  • Closed thought leader content and discounted access to our popular events, such as Transform 2021: Learn more
  • networking features, and more

Become a member


Please enter your comment!
Please enter your name here