Failing with DEI? People Analytics can help.

Here is a term I would prefer didn’t exist: tokenism.

Research by the Harvard Business Review found that when organizations focus solely on diversity and fail to address inclusion, they may inadvertently create a"token" environment where underrepresented groups are hired but not fully integrated into the organization's culture and decision-making processes.

When evaluating DEI in the workplace, there are two points that we can all (hopefully) agree with:

1. Furthering DEI is good for society.  It’s also good for business.  McKinsey & Company found that companies in the top quartile for gender diversity on executive teams were 25% more likely to have above-average profitability than companies in the bottom quartile. Additionally, companies in the top quartile for racial and ethnic diversity were 36% more likely to have above-average profitability than companies in the bottom quartile.  

Numerous studies link diversity to innovation, creativity, and improved decision-making.

2. Many companies are failing, and some are failing quite badly at implementing DEI.  Simply increasing diversity metrics does not necessarily lead to increased diversity and inclusion in the workplace. Instead, organizations need to take a comprehensive approach that includes not only hiring diverse candidates but also addressing biases and inequities in the workplace culture and practices.

So, if we know the problem, and we know the solution: what’s not working?

Change is hard. Stagnation is dangerous.

There are numerous reasons why many companies have failed to reach their DEI goals. Two factors stand out:

First, there is resistance to change and systemic barriers.  There is a perception of a zero-sum game that if one group is promoted, it comes at the expense of the other. In other words, there are people who resist efforts towards DEI because they fear that it will lead to discrimination against them.  

We also need to be frank:  biases continue to exist in the workplace, whether we admit it or not. This is reflected in studies of the hiring and promotion of minorities.

Second, there is a lack of accountability. Research by the Harvard Business Review found that while 78% of organizations surveyed had a diversity and inclusion strategy in place, only 32% had metrics in place to measure success.

Where does AI come in?

At its most elementary level, if configured correctly, AI can be applied to a dataset to answer a question without bias. Whereas people are biased, if we correctly remove all bias-related data, decisions can be made based on objective data. While this is not an easy task, as AI algorithms are sometimes trained on data that is biased to begin with, there are available solutions today to overcome these problems.

Let’s start with Resistance and Systemic Bias, using an example where the organization needs information about emerging talent to fill open positions with internal candidates.

There is a risk of relying exclusively on management (and peer) feedback because this can be subject to bias.  A Machine Learning algorithm can be trained to find hidden talent without any consideration of the person’s ethnicity, gender, etc.  Of course, we are not advocating that management input be ignored in these decisions but that they not be relied upon exclusively.

An absence of accountability is another limitation to the advancement of DEI.  Other than the basic information on the numbers of different population groups, management often does not have the insights to fully understand DEI.

From a technical perspective, there is another aspect of AI that is less complicated but can impact the organization significantly nonetheless. HR dashboards can be a tool in the adoption of DEI providing the right information is presented so that actionable decisions can be taken.

In the chart below, we can see the bias in compensation levels based on population groups.                    

This simplified depiction of wage data can be presented across the organization creating visibility about a critical topic.

Does this type of information address inclusion and tokenism?

Not immediately and not without real efforts to tackle bias.  But shedding light on inequalities and acknowledging their existence is the first step to change.

‍