In 2018, Karen Britton joined LMI to support the consultancy in its efforts to help government solve its most complex challenges. As vice president of communications and marketing and chief digital officer, she’s lately focused on data, especially the uses of artificial intelligence and machine learning in shaping the workforce of the future.
For government and the GovCon community alike, workforce issues are top of mind. Here, Britton sheds light on the ways in which data — when used properly — can accelerate time to insight and drive improved decision-making around recruiting, workforce development, long-term HR planning and other critical strategic areas.
How can AI and ML work in support of “people analytics”?
People analytics really is about using big data and analytics to make sure that you are making decisions that are good for the employees as well as good for the organization. You find the patterns and inconsistencies or gaps in talent data, and you utilize big data analytics to predict what kind of outcomes will there be if you are on (or off) the trajectory of your goals.
I’m finding that human resources leaders are working more closely with data scientists to use analytics as part of their toolbox. If people understand how to use it well, it can be enlightening for organizational leaders.
What exactly can the data tell us?
People analytics starts with defining the talent management business question, leveraging the sources of data available to you and determining how you will use that data to discover new insights.
A couple of preliminary key questions organization leaders should ask themselves are: What affects the retention of underrepresented groups in our organization? How can our Diversity, Equity, Inclusion and Accessibility programs exert a positive influence on retention? Talent data, payroll, attrition, etc. are in various data sources in human information systems and can be used to answer key questions.
As an example, companies have been known to promote white males based on potential, whereas women and minorities are based on experience, which reflects the prove it again bias. Analysts can review promotion data, engagement data by ethnicity and so on. Ultimately, the entire employee lifecycle can be reviewed for patterns.
What’s the benefit of this approach?
People analytics allows organization leaders to measure results. When you’re looking at diversity and inclusion, you’re looking at the human capital lifecycle. One can use analytics to evaluate the talent pipeline, to ensure that diverse populations are being hired and advancing through the organization.
With data visualization tools such as Tableau and Python, leaders can understand the patterns and relationships across the data by role, segment and so on. This gives leaders the opportunity to answer the “what if” questions and provides leaders with a firsthand understanding of tracking representation of pipeline by gender and race to meet desired demographics in their organization.
But AI/ML models can also be poorly designed or misused. What’s the risk here?
While there are pockets of excellence in AI and ML, we’ve also seen examples of advanced analytics models sentencing people of color more harshly, erroneously accusing low-income and immigrant families of fraud and awarding lower grades to students from less privileged neighborhoods.
Fair treatment is core to the mission of public agencies but can be hard to preserve if decisions are based on algorithms built upon biased data sets. Bias might creep in because the data includes biased human decisions or reflects historical or social inequities. Or it can stem from flawed data sampling — the under or overrepresentation of certain groups of people, for example.
Lack of transparency can also exacerbate the problem. AI and ML techniques can make it hard to track how the underlying data drives model outputs, making it difficult to spot bias or unfairness.
How can organizations address bias in people analytics?
Data must be structured, digitized and cleansed. For example, there should be a check whether data sources are disconnected within an organization. If you are looking at payroll, compensation, employee ID, etc., the information is likely in different sources. How do you connect them? How do you work to improve your understanding of your employees work experience?
Many organizations take six months to a year just to integrate these data sources, to create a unified data source. There are a lot of tools that organization leaders can use, such as TrustSphere, Polinode and Volumetrics. By tracking underrepresented groups, leaders are holding themselves accountable to their employees and goals.
What other best practices can help mitigate the risk of bias?
A senior leader should be made responsible for risk management. In some organizations, this falls to the chief risk officer, the chief information officer, the chief data officer or someone who has responsibility for governance and oversight of technology.
Analytical practices and standards should be codified, widely communicated and adhered to. These could include clarity about the specific problem a model seeks to address, a rigorous peer review process and an empirical review of outcomes to detect any unintended bias.
Bias training throughout agencies and companies is also essential, especially for team members who are developing models.
Enterprise risk management governance processes and models should also be regularly reviewed. Create and maintain an inventory of models in use across the agency. Develop and maintain the standard workflow for models to ensure the widespread adoption of best practices in data science and bias awareness/reduction. This might include incorporating algorithm review panels and even engaging with academic and industry partners in this space.
Finally, advanced analytics can be more effective when embraced agencywide, rather than just by a handful of advocates. It helps to have a centralized effort to systematically identify and prioritize the highest-impact use cases; to support them with adequate funding; and to communicate progress, lessons learned, and professional standards across the entire agency, in order to improve effectiveness.