Oct. 12, 2026

“I like math because there’s a single correct answer.”

I’ve heard this a million times, and it resonates with me too. You can know and confirm that the answer is correct, and there’s no subjective interpretation involved.

However, this principle does not always hold in applied mathematics contexts, like statistics and HR analytics.

Granted, there will only be one correct result of a particular calculation, regardless of whether it’s framed in terms of abstract mathematics or a concrete business situation.

But how you get that result, as well as what you do with it (the judgments you make about the calculated values, the decisions you base on those values…) remains highly subjective.

1. How do you decide what analytic question to ask in the first place? What’s a good vs. bad question?
2. What’s the right analysis for your specific business question?
3. How complex should the analysis be?
4. How do you decide whether the assumptions of your statistical procedure are met?
5. How do you know if you’ve accurately quantified the things you’re interested in?
6. Should you take a Frequentist or Bayesian approach? (Or a blend?)
7. Should you emphasize p-values or confidence intervals? (Or both?)
8. Which data should be included vs. omitted?
9. What’s the best way to handle missing data?
10. What’s the most effective way to visualize the findings?

We could go on and on like this.

All these questions are matters of judgment, which must be based on an underlying philosophy of data analytics. (Otherwise, you’re probably just following a flowchart of semi-arbitrary decision points.)

---

As a sidebar, if you ask a chatbot to answer any one of these questions, it will usually respond with an overarching review of the subject matter, including a survey of both good and bad* advice.

(*The reason I call some of its advice “bad” is because that’s a subjective judgment I can make based on my years of investigation into statistical analysis.)

This is also why you can’t simply have the AI run all your analyses for you…

AI programs excel in situations with clear, set rules and procedures. As such, they’re fantastic at writing the code to execute a particular statistical analysis.

But they can’t make any actual judgment calls for you, at least not without relying on arbitrary decision rules. They have no years of experience, nor context to ground judgments in.

The notion that an AI can make any judgments at all is, like any form of AI anthropomorphizing, a human hallucination.