The benefits provided by AI and machine learning are pretty well established. The technology can help businesses automate processes, gain insight through data analysis, and engage with customers and employees. And it can help them satisfy ever-changing market demands, streamline operational costs, and remain competitive in an increasingly fast-paced digital world.
HR’s Role in Understanding and Mitigating AI Bias
Today many major cloud providers even offer AI features within their service packages, democratizing the technology for businesses that might otherwise struggle to afford expensive in-house AI engineers and data scientists.
For HR teams, the value of AI is undeniably clear. When a single job listing results in hundreds or even thousands of applicants, manually reviewing each résumé is a monumental and often unrealistic task. By leveraging AI and machine learning technologies, HR teams gain the ability to evaluate applicants at scale and make hiring recommendations far more efficient.
The ramifications of AI-induced bias in HR are significant
While AI offers fairly obvious benefits for HR groups, it also introduces pretty serious challenges and potential pitfalls. With any AI system, one of the most difficult (yet critical) aspects you must address head-on is ensuring that it’s free of bias.
This is particularly crucial for AI systems for HR, as any AI-induced bias can result in companies discriminating against qualified candidates — often unknowingly.
Remember when Amazon had to scrap its AI system for screening résumés several years ago because it penalized women applicants? It’s a perfect — albeit unfortunate — example of the power of training data. At the time, the majority of Amazon’s employees were men, so the algorithm powering the AI system, trained on the company’s own data, was associating successful applications with male-oriented words.
In doing so, well-qualified women candidates were simply overlooked by the model. The lesson: If the data used to train your AI models is biased, then the deployed AI system will also be biased. And it’ll continue to reinforce that bias indefinitely.
Both outsourced AI systems and company cultures require a closer look
In Amazon’s case, the AI system for screening résumés was built in-house and trained with data from the company’s own job applicants. But most companies don’t have the resources to build internal AI systems for their HR departments. So HR teams are increasingly outsourcing that work to providers like Workday or Google Cloud. Unfortunately, too often, they’re outsourcing their due diligence as well.
It’s more important than ever that HR teams acknowledge the enormous responsibility that comes with outsourcing any AI implementation. Don’t just blindly accept and implement your AI provider’s models. You and your teams need to review the systems repeatedly to ensure they aren’t biased. You need to constantly be asking:
- What data sources (or combination of data sources) are being used to train the models?
- What specific factors does the model use to make its decisions?
- Are the results being produced satisfactory, or is something askew? Does the system need to be temporarily shut down and reevaluated?
It’s so essential to carefully review training data, particularly within outsourced AI systems. But it’s not the only requirement for mitigating bias—biased data originates from biased work environments.
So your HR teams have a duty to also evaluate any issues of bias or unfairness within your organization. For example, do men hold more power than women in the company? What questionable conduct has long been considered acceptable? Are employees from underrepresented groups provided every opportunity to succeed?
The diversity, equity, and inclusiveness of your company’s culture are absolutely relevant when incorporating AI, because it drives how AI systems and results will be deployed. Remember, AI doesn’t know it’s biased. That’s up to us to figure out.
Three best practices for leveraging AI fairly and without bias
Ultimately, HR teams need to be able to understand what their AI systems can do and what they can’t. Now, your HR teams don’t have to be technology experts or understand the algorithms powering AI models.
But they do need to know what kinds of biases are reflected in training data, how biases are built into company cultures, and how AI systems perpetuate those biases.
Below are three tactical best practices that can help your HR teams leverage AI technology in a fair and unbiased manner.
- Regularly audit the AI system. Whether your systems are built in-house or outsourced to a provider, routinely review the data being collected to train the models and the results being produced. Is the dataset large and varied enough? Does it include information about protected groups, including race and gender? Don’t hesitate to shut down the system to shift course if its results are unsatisfactory.
- Understand the data supply chain. When relying on outsourced, off-the-shelf AI systems, recognize that the training data may reflect the vendor’s own biases or the biases of third-party datasets. Keep an eye out.
- Use AI to augment, not replace. The capabilities of AI are advancing rapidly, but the reality is that AI still needs to be managed. Because of the risks involved, HR teams should leverage AI to augment their role, not replace it. Humans still need to make any final hiring and HR decisions.
With the help of AI, HR teams can uncover corporate inequities
Your HR teams are in a unique position to leverage AI technology in a fair and unbiased manner because they’re already well versed in systemic issues of bias and inequity.
Recognize the responsibility AI systems require and consistently work to understand how they’re being trained and producing results.
When done correctly, AI will help your HR teams uncover bias rather than perpetuate it, improve the efficiency and efficacy of HR duties—and advance the careers of deserving applicants and valued employees.
Image Credit: rodnae productions; pexels; thank you!