People data and analytics is now one of the hottest, most in-demand skills for HR professionals and ranks as one of the most important human capital trends for organizations today. Companies are looking at data to inform decisions and answer questions such as: Who will leave and when? How do our capabilities compare to competitors? What skills and talent do we need? This kind of data provide insights and can help decision makers by answering these questions and more.
But this surge in data use has come with questions about the ability of data-based processes to introduce or perpetuate bias and its broader impact on fairness in the workplace.
As humans, we have nearly 200 types of cognitive bias that affect our decision-making. We see these when we notice others who tend to only seek out information that confirms what they already believe (confirmation bias) or when we hear someone say they “knew it all along” after an event has occurred (hindsight bias).
Not all biases are bad. Many help us make decisions more quickly and can save time and brainpower in a world with short deadlines and too much information. The problem is that, at times, our data can mirror harmful human biases and encode the unconscious errors of the people who design, collect, and use the data. The issue of fairness arises when biased data affects judgment and introduces error into our HR data, processes, and technology.
But the story is not all doom and gloom. Data need not be a sponge soaking up human biases. Data may also lead to fairer decision-making by reducing the subjective (and possibly biased) interpretations of human decisions. Analytics-based technology might even be able to identify and remove implicit human biases. Can data and technology eliminate human bias? No, not entirely. But it is possible to use it to minimize bias and increase fairness.
Here are seven ways leaders can help their company ensure a culture of fairer data-informed HR decisions:
You’ll probably never build a single metric or universal definition of fairness that applies in all cases — in fact, there are more than 45 definitions for “fairness” listed in the Google Developers Glossary. So start by identifying your organization’s highest-priority aspects of fairness. Then create several compatible definitions based on metrics and standards that work in different use cases and circumstances.
For example, if inclusion is a top priority, you can start by defining what inclusion means and looks like in your organization. Then define ways you’ll measure inclusion — or its absence — so you can track and evaluate progress. This may be a good time to ask your analytics colleagues to help you with measurement techniques and interpretation of the results.
Explain why fairness is essential to your bottom line. Present your ideas in business terms and gain the support of trusted stakeholders known for making evidence-based business decisions. For example, draw a clear line between fairness and diversity. Then share how a recent McKinsey study spanning 15 countries and 1,000+ companies found that those in the top quartile of gender, ethnic, and cultural diversity outperformed those at the bottom by 25 to 36%. There’s nothing like robust metrics to support your case.
Technology can be a powerful tool for identifying and correcting bias in people processes. It may never be able to ensure that all data-informed decisions are fair and that all bias is addressed, but don’t let the lack of perfection hold you back from leveraging it. New technology emerges daily that supports fairness in data-informed HR decision-making. So, keep up to date on new product developments and build a constantly evolving portfolio of software, tools, and procedures focused on increasing fairness.
Fairness audits are objective, systematic methods that look at an organization’s people data policies, practices, and procedures. Audit goals vary: One goal could be to look for potential issues that make an organization liable to legal action. Another could be to identify ways to improve fairness-related practices.
A fairness audit could be done through software tools, the help of an external expert, or even through an internal review. What’s most important is that they are completed on a regular basis by someone who understands your organization’s definitions of fairness and that any issues or opportunities are acted upon.
When you do discover a problem, look for the underlying assumptions or processes that need to be changed. Biases are often unconscious, and unfairness is generally the symptom of more than a single decision. This makes the underlying cause of a problem difficult to detect.
Consider using an approach loved by toddlers around the world: Ask “why?” at least five times. Here’s an example of how the technique might work for you. Problem: Diversity in our organization is 45%, but only 5% of executives match our definition of diverse. Why? Because we hire most executives from outside the company. Why? Because our internal talent doesn’t have enough leadership experience in our industry. Why? Because they haven’t had enough leadership development opportunities aligned to our business objectives. Why? Because we don’t have the resources to provide industry-specific leadership development opportunities. Why? Because we didn’t prioritize it when setting goals and budgets… oh… Bingo! Maybe the problem isn’t 5% executive diversity, maybe that is just a symptom.
There are usually multiple underlying causes and simply identifying potential causes won’t solve the problem. Action is still needed. However, until you identify and address the underlying issues you won’t see change.
A diverse group will be better able to identify and solve issues of unfair bias in systems. So be sure to include individuals with a diversity of thought and experience at each stage of the process to lower the likelihood of unintended biases creeping in unnoticed.
Solicit input from across the organization by bringing together experts from multiple departments such as HR, data, technology, and legal to identify and promote opportunities for new or improved people-related data systems. Collaborate with vendors or in-house data professionals to develop and improve operational practices and ethical standards that make the use of people-related data systems fairer.
In many cases, there isn’t — and shouldn’t be — a substitute for human judgment in HR decision-making. Fairness in recruiting, employee evaluation, and other HR decision making requires human participation.
Consider the interview process: It is completely possible these days to remove the human and allow data from video interviews to make decisions. When asked about such an approach, 67% of respondents in a Pew Research study felt this would be unacceptable. The reasons they gave? It would be flawed or biased and that “humans should evaluate humans.” Among those who believe it is acceptable, they still believe this approach would have flaws, bias, and fairness implications. There was also an acknowledgement that it shouldn’t be the only data point used in the process. Rather than let the data take over, explore how humans and data work best together. Some promising HR decision-making methods combine machines and humans to reduce bias. This class of techniques includes “human-in-the-loop” decision making in which data is used to provide options or recommendations that humans then verify or choose from. Such techniques can help to remind us that HR data represents real people with real lives, not just numbers to be analyzed without further thought.
Fairness is becoming an important part of business success — achieved by doing right by people and doing right in the world. Our new, analytics-based environment provides opportunities to define fairness for the organization, build a business case for it, and collaborate with groups within and outside the organization to create and maintain it. The challenge is to be aware of the opportunities and seize them when they appear.
This content was originally published here.