AI and Unconscious Bias

Every month we dive into what we are all so passionate about here at FMP: human capital management. This month we are so excited to continue to learn about the fascinating topic of Artificial Intelligence (AI) and unconscious bias from one of our partners, Deb Cohen. We recently asked Deb a few questions about the topic based on one of her articles, Unconscious Bias: An AI Perspective. 

Before we jump in, could you explain what unconscious bias in the workplace looks like?

It is widely known that we all have unconscious biases. While raising awareness of unconscious bias is an important first step, awareness alone is not sufficient to stop bias that occurs in the workplace or anywhere else. Unconscious bias is when a person unknowingly allows attitudes, feelings, stereotypes, or beliefs to impact their judgment about people. It is unconscious because it isn’t done deliberately; it is an involuntary reaction based on deep-seated thoughts. Everyone is prone to jumping to conclusions, misjudging people, and favoring some more than others whether they realize it or not. Explicit bias, on the other hand, refers to attitudes and beliefs people have on a conscious level.

Unconscious bias in the workplace should be mitigated for a variety of reasons. Subjecting candidates and workers to unconscious bias exposes the organization to ethical, moral, and legal hazards. Besides being the right thing to do, organizations can achieve better outcomes if unconscious bias is reduced. Having a more inclusive workplace provides access to a broader population of workers. These workers can bring new and different ideas, which fosters innovation and increased productivity. Diversity and Inclusion (D&I) programs seek to address conscious and unconscious bias, so employers and employees are treated fairly. The objective of D&I programs is to eliminate or reduce the effects of overt discrimination and unconscious bias.

Unconscious bias can creep into the workplace by many means, such as job descriptions, interview questions, the perceptions employees have of co-workers, and employment decisions managers make. For example, “clear,” “confident” and “outspoken” may be qualities used to positively describe men but are often considered negative qualities when attributed to women. Further examples are selecting teams that look like the leader or assuming working mothers are less career oriented than others. Although many biases are small and seemingly inconsequential, they can add up and affect an organization’s decision-making processes and culture, particularly with respect to talent acquisition and development.

Why should Human Resource (HR) professionals use Artificial Intelligence (AI) to prevent unconscious bias?

Artificial Intelligence (AI) and automation will dramatically change how work gets done in the future. This is often viewed optimistically, as routine tasks can be automated by a machine, freeing up humans for more creative and strategic endeavors. Though what will remain the same as organizations tackle these issues? One thing that will not change is unconscious bias. The unfortunate reality is that unconscious bias can also sneak into AI applications and algorithms. Critics warn that AI may not only perpetuate unconscious bias but exacerbate it. In some cases, it already has.

A large metropolitan city decided to use an AI app for residents to report rough surfaces while driving. The app’s intent was to pinpoint where potholes existed to prioritize resources for repairs. But the residents using the app were predominantly younger and more affluent than the city’s population as a whole. Rather than getting a picture of the entire city, the app pointed the city to the more affluent areas, perpetuating the challenge of how limited resources may not be evenly distributed across populations. The importance of D&I programs will not diminish due to the arrival of AI. Organizations should continue take proactive steps to address their unconscious bias or D&I challenges.

What happens when AI replaces people-based talent acquisition processes?

It is already happening. For example, chatbots (programs designed to simulate conversations with humans over the internet) are interacting with job candidates, approving requests for paid time off, and improving wording in job listings. If a candidate screening chatbot is programmed to use certain words or phrases that are not familiar within a subgroup, then that system is inadvertently creating a biased candidate pool. Even organizations with well-regarded and effective talent management practices must focus on the underlying issue of unconscious bias. Like it or not, AI will be changing how organizations attract, select and manage talent.

Interesting! Can you share an example of AI being leveraged in HR currently?

The benefits of AI-based talent acquisition are potentially vast. Hilton Hotels was able to drop the time-to-invite for call center job candidates by 88% by using AI in their pre-hire assessments. Cost savings and efficiency measures such as this are quite impressive and exciting. But is widespread AI application in talent acquisition possible, and are there drawbacks? Because D&I and unconscious bias have become so important in organizations, the lure of AI and automation is significant. However, critics remain concerned that AI applications intended to improve talent processes could perpetuate or even exacerbate biases in hiring and promotion rather than alleviate them.

What can we expect from AI in the future and how can it help to prevent unconscious bias?

AI systems must be programmed and trained by humans. AI trained with bias will become biased. D&I programs must analysis AI-based talent programs for biased decision making and outcomes, just as they would for a traditionally human-based program. AI may ultimately help reduce unconscious bias, but it is not a shortcut to removing bias. The unconscious bias code has not been cracked yet with decades of research and focus. AI is not expected to do this quickly, perfectly, or without evaluation and refinement. In the future, AI systems may be making judgments instead of people. If those systems can be designed to be bias free, then candidate selection will take less time, be more objective and potentially be more diverse. But the “if” must be emphasized.

What final thoughts do you want to share about this topic?

Remember that everyone has biases, intended or not. Realize that bias is not necessarily eliminated because an AI system made the decision instead of a person. Organizations should:

• Continue to review outcomes for talent management, especially after AI-tools are adopted.
• Set granular and specific goals for D&I programs
• Evaluate how systems, both people and AI-based, perform with respect to attraction, selection, promotion, training, performance evaluation, and so forth.
• Promote awareness of unconscious bias and do so with visible advocacy from the top.
• Invest in training and support-related training initiatives. Awareness is not enough.
• Conduct regular audits of culture and outcomes; unconscious bias can shape an organization’s culture and skew decision making tools such as performance reviews

Unconscious bias adds risk and cost to the talent acquisition process. AI in and of itself does not alleviate that risk. These concerns and cautions, however, should not thwart attempts to use AI for good. There is a distinct upside to using AI, and its implementation is gaining speed, not slowing. The use of AI in the talent acquisition and management process may help eliminate or ease unconscious bias. So, go forth with AI solutions relating to talent but do so carefully and monitor the outcomes. Create accountability for both the technology application and with the outcomes from the use of the technology.

To learn more about AI and Unconscious Bias, as well as several other important Human Capital Management topics, visit www.DebCohenLLC.com.

Deb Cohen, Ph.D., is an author, trainer and management consultant with more than 25 years of experience advising, speaking and guiding nonprofit, academic and for-profit entities.  Her latest book, Developing Management Proficiency: A Self-directed Learning Approach (2020), focuses on how managers can develop their management competence through self-directed efforts. Deb is a Distinguished Principal Research Fellow with The Conference Board and is an Adjunct Professor at George Washington University teaching MBA courses in Strategic HR. A subject matter expert in HR, management and organizational behavior, Cohen’s expertise lies in creating and executing new initiatives that support and develop organization strategy.