Keeping it Ethical
“Just because you can measure everything doesn’t mean you should.”
– W. Edward Deming
When you conduct people analytics work, there are countless ethical questions to consider. Could your analyses lead to unfair or discriminatory outcomes? Are you being transparent about how data is used? Are you considering the long-term impact of your work on employees and the organization’s culture? These are the kinds of ethical questions you should be asking yourself regularly. Ethical people analytics requires critical thinking, empathy, and a commitment to doing what’s right, even when it’s difficult. It’s not just about following rules; it’s about developing a strong ethical framework and using it to guide every aspect of your people analytics work.
When I interview someone for a people analytics job, I like to ask this question: What’s the difference between business analytics and people analytics?” It can be a confusing question since a typical definition of business analytics might not be different from the definition I gave for people analytics at the beginning of this book: “the application of data and insights to improve business outcomes through better decision-making regarding people, work, and business objectives.” It is correct to say that people analytics is specific to the information about, for, or relating to the people and their work or work environment in some way, while business analytics may look at other aspects like analyzing equipment, finances, or market trends. But my favorite answers are the ones that dive deeper into what it means to focus on people and the ethical responsibility that comes with our work. To me, that is the key difference. People analytics and business analytics use much of the same processes, tools, and technology approaches, but in people analytics we’re not just analyzing numbers to make business decisions; we’re dealing with people’s lives. The decisions we inform affect someone’s career, wealth, well-being, family, and whole future. It’s not just about doing what’s best for the company financially; it’s about doing what’s right for the people involved.
But knowing what’s “right” in people analytics can be tricky. There’s no simple list of “good” and “bad” things. There’s obviously terrible stuff like privacy breaches, creepy monitoring, and things that’ll get you in legal trouble. Then there are clear-cut things that need to happen involving people data, like ensuring people are employed, paid, and cared for. But people analytics is filled with so many gray areas around what is “okay” to measure or analyze and there are myriad nuances about how to do so appropriately. The stakes are higher and the uncertainty is greater in people analytics than in many other analytics-focused career areas.
Knowing what to do (or not do) can be confusing but ethical decision-making can be learned and there are skills that can be strengthened in this area. There are also experts to learn from and existing best practices to follow. For example, the International Organization for Standardization’s “Human Resource Management: Safe Handling of Data” standard provides best practices for handling people data appropriately at every stage of the data life cycle from creation through usage to storage or deletion. (Full disclosure: I’m a U.S. representative for ISO and served as the global lead responsible for creating this standard. This was a completely voluntary role; I receive no compensation from ISO. I’m simply passionate about ensuring people data is handled responsibly.) You can also work to learn about and build the following people analytics ethical skills. Being ethical in people analytics includes considering fairness, transparency, and the impact of people analytics practices on individuals and groups. It requires accountability, ensuring data privacy, and the identification and avoidance of potential biases. It’s also about maintaining data integrity, respecting individual rights, and ensuring that data is used in a way that promotes positive outcomes.
People Analytics Ethical Skills
Data privacy: Data privacy in people analytics involves understanding and implementing practices to protect employee data from unauthorized access, use, or disclosure. It requires knowledge of relevant regulations, ethical considerations, and best practices for data handling, storage, and anonymization. Privacy isn’t just about ticking boxes on a compliance checklist; it’s about respecting people. Protect the data like it’s your most precious possession. Because it is.
Consent: You can’t just collect data without asking. Consent means obtaining explicit and informed permission from people before collecting or using their data. Employees need to understand what they’re consenting to so explain it clearly and simply and keep in mind that consent isn’t forever. People should be able to withdraw their consent at any time. You will want to understand the principles of informed consent, communicate clearly about data usage, and provide individuals with the ability to easily withdraw their consent. And, remember that consent applies to only the use that you have permission for, you can’t collect data “just in case” and you can only use it for its specified purpose (at least until you go get consent for a new legitimate use).
Legitimate Use Cases: Sometimes when you are passionate about a thing, you just dive right in. I know I feel that way about doing people analytics sometimes, especially if I learn about a new technique or discover a data set I didn’t know about. But it’s important to always stop and ask: Why am I collecting or analyzing this data, is there a legitimate and approved use case for doing so? Saying something has a “legitimate use case” means there is a real and valid reason for using it, not a hypothetical or potentially exploitative one. It means that all of your data uses are aligned with business needs, have already considered the impact on employees, and the rationale for it is documented. Having pre-determined what the legitimate uses for people data are will ensure that data is used responsibly and ethically. “Because we can” is not a valid reason. To justify use, your people analytics work should tie to business needs, help the organization achieve its goals, and should not harm the individuals involved. Always document what you are doing, why you are doing it, and the impacts it may have. This will help you stay on track and demonstrate accountability.
Ethical Data Collection, Analysis, and Reporting/Sharing: Ethical practices need to be in place anytime people data is handled – from its collection to analysis to sharing and use.
-
Collection and Storage: Ethical data collection starts with asking the right questions. Should you collect this data? Why are you collecting it? Is it necessary for a legitimate use case? Have you obtained informed consent? Are you being transparent about your data collection practices? Are you minimizing the data you collect to only what is necessary? These are crucial questions to consider before you even start gathering data. Think about the potential impact on employees. Are you collecting sensitive information that could be misused? Are you being mindful of power dynamics in the workplace when requesting data? Make sure you also have safe systems and processes for storing this data once you’ve collected it so that it stays private and secure.
-
Analysis and Interpretation: Once you have the data, ethical analysis comes into play. Are you using appropriate statistical methods? Are you aware of potential biases in your data and algorithms? Are you interpreting the data responsibly? Are you avoiding “data dredging” – searching for findings without a clear hypothesis? Be careful not to draw misleading conclusions or fall prey to the many common tempting biases like confirmation bias, availability bias, automation bias, or selection bias. And, are you protecting the privacy of individuals when you finalize your analyses? Even if you’ve anonymized or aggregated the data, it’s sometimes possible to re-identify individuals. Be mindful of the many analysis risks and take steps to mitigate them.
-
Sharing and Use: Finally, ethical reporting and sharing are essential. Are you communicating your findings accurately and transparently? Are you avoiding sensationalism or exaggeration? Are you being clear about the limitations of your data and analyses? Are you sharing your findings only with those who have a legitimate need to know? Are you protecting the privacy of individuals in your reports? Consider your audience. Are you presenting the data in a way that they can understand? Are you avoiding jargon or technical terms that might be confusing? And remember, data is powerful. It can be used to influence decisions, and those decisions can have real consequences for people. Therefore, it’s your responsibility to present your findings in a way that is fair, balanced, and objective.
People Analytics Research Ethics: In the academic world there are Institutional Review Boards. These are formal groups that review and monitor any research involving people to ensure things are done ethically. Unfortunately, there isn’t a formal structure like that available for all people analytics practitioners. That means you will be responsible for ensuring that your research is designed and conducted ethically. Ethics in people analytics research requires minimizing risks to participants, ensuring confidentiality, and maintaining objectivity. Avoid research designs that could harm or disadvantage certain groups of employees.
Mitigating Bias: Mitigating bias in people analytics is an ongoing process of identifying and addressing potential biases in data, algorithms, technology, and decision-making processes. It requires critical thinking, data analysis skills, knowledge of algorithmic bias, and an understanding of how human biases can influence outcomes. It’s about striving for fairness and equity in people analytics practices and ensuring that decisions are based on objective and unbiased information.Bias can creep into every stage of the people analytics process as mentioned above, but don’t forget to look for and address larger-scale biases that may be present in the tools, processes, systems, and ways of working present today. As many people analytics approaches are getting automated into algorithms, algorithmic bias is becoming a greater challenge. Algorithms are only as good as the data they’re trained on and only as well controlled as they are designed. Regularly audit all of your processes, systems, and approaches for bias. Even the tools we use can introduce bias. Be aware of the limitations of your technology and how it might be affecting your results. Finally, don’t forget human bias. Even with the best data and algorithms, human interpretations and decisions can still be biased. Implement checks and balances to ensure fairness and objectivity in decision-making. We all have biases, whether we realize it or not. Be aware of your own biases and how they might be influencing your work. Seek out diverse perspectives to challenge your assumptions.
Standing Your Ground: People Analytics Lessons Learned
I used to build predictive attrition models to forecast how many employees were likely to leave the company and identify the reasons why people leave. The models were quite good at predicting what might happen. One day, a leader asked me for a list showing the likelihood that each member of her team would quit. It would have been easy to generate probabilities, but we had already decided not to analyze any individual person. To deny the request I simply said, “I’m sorry, we don’t provide individual-level predictions; these are statistical trends meant to be used in aggregate for planning purposes.” Her pushback was immediate: “I know, but can’t you apply the model’s factors to each individual and calculate a probability?”
Yes, I could. Easily. I remember stammering and getting flustered. I thought about a colleague, let’s call her Sarah, who worked for this person. If Sarah was labeled “likely to leave,” would her boss treat her differently? What if Sarah wasn’t planning to leave, but this analysis caused her boss to treat her poorly and she did end up quitting? Even a “not likely to leave” label could be detrimental. The boss might prioritize retention efforts for those deemed at risk, overlooking Sarah for rewards she deserved. A single data point—one I was asked to create—could unfairly influence Sarah’s career.
With Sarah in mind, I refused, stating it was inappropriate and that I wouldn’t share the model’s details. This didn’t go over well. She became angry, the situation escalated, and I held firm. She invoked her seniority and threatened to go to my boss. I replied, with perhaps too much smugness, “Good. She’ll be proud to hear how I’m protecting our employees from requests like this,” and walked out.
Predictive models are powerful tools, but they have limitations. Used responsibly, they can highlight areas for improvement and inform better practices. However, relying on them for individual decisions is disastrous. We’ve seen countless examples of predictive algorithms for hiring, performance, and promotion with negative consequences for real people—like Sarah. It’s up to people analytics professionals to uphold ethical data use.
While I don’t recommend confrontations, I’m proud I stood my ground. Though we never became friends, I earned my boss’s respect (who shared the story), and our head of legal later thanked me personally, effectively gaining the legal team’s support.
In hindsight, this experience taught me valuable lessons. First, pushing back against questionable requests is always uncomfortable. Standing your ground can be unpleasant. I stammered, got flustered, didn’t articulate myself perfectly, and was shaken afterward. But I grew from it and protected people (even if they would never know about it.) So, I encourage you to stand your ground when necessary. It’s how you build trust and respect as an ethical leader.
Second, I did it wrong – don’t be like me. If there had been a set of people data principles already established, I could have pointed to those and simply said, “Yes, mathematically that is possible, but unfortunately, that’s against our policy.” I also tried to tackle a difficult situation alone. I didn’t leverage my community. I could have sought support from my boss (who was supportive) or other allies. I could have said, “We haven’t done it that way; let me check with my boss,” removing myself from the immediate pressure and returning with backup. Even if you’re unsure of support, this approach—deferring to a higher authority—can buy you time and potentially uncover allies.