When I wrote “Messy Data, Meet Computational Thinking” for the HEUG Connected Campus community last September, I was mostly trying to make sense of a very unglamorous afternoon that comprised of merging a CSV, a delimited text file, and a duplicate-riddled Excel sheet into something Tableau could actually read. What surprised me wasn’t that AI (Microsoft Analyst) cleaned it up. It was that watching the tool work felt like watching computational thinking come to life: decomposition, pattern recognition, abstraction, and algorithmic thinking, all on display in one real problem.
Recently I came across Conrad Wolfram’s work on computational thinking (Getting Smart Staff, 2020), and it reframed that afternoon for me. Wolfram, a strategy director at Wolfram Research and author of The Math(s) Fix, has spent years arguing that we’ve been teaching the wrong thing. Schools drill hand-calculation, but computers do calculation faster and better than any of us ever will. What humans actually need is the ability to use computing tools to take on harder and harder problems. And crucially, he insists that computational thinking is now required in every field and in everyday life (Getting Smart Staff, 2020).
Here’s what clicked for me. My article leaned on the classic four pillars of computational thinking (decomposition, pattern recognition, abstraction, and algorithmic thinking). Wolfram describes a four-step process instead:
1) define the question,
2) abstract it into a computable form,
3) compute the answer, and
4) interpret the result
Essentially, different vocabularies but the same underlying skill. And my messy-data task maps almost perfectly onto Wolfram’s loop:
Define: What unified dataset do I actually need to answer questions in Tableau?
Abstract: Reframe three inconsistent files into one clean schema: IDs as text, leading zeroes preserved, a single convention for “missing.”
Compute: Let Analyst do the merging, deduplication, and fuzzy matching. This is the “calculation” no analyst should be doing by hand.
Interpret: Review the ambiguous duplicates, approve the merges, sanity-check the integrity. This is the human judgement.
Wolfram’s whole point is that step three (the mechanical part) is exactly what we should stop doing ourselves. He calls our era a “computational knowledge economy,” where the value isn’t in what you know but in what you can compute from what you know (Getting Smart Staff, 2020). MS Analyst didn’t replace my thinking; it took over the calculating so my thinking could move up a level.
Why this matters for higher ed analysts
If you work in institutional research, enrollment analytics, or student success, you already live in messy data. And AI tools are about to absorb an enormous share of the wrangling. That’s the opportunity and also the trap.
The opportunity: our value is now more toward framing the right questions and interpreting results in context. Wolfram makes a point I keep returning to which is that setting up a problem is usually more important than solving it. A tool can merge three files; it can’t tell you whether a missing advisor field means a data-entry error or an unadvised student who might soon be leaving your institution. That judgment is the job.
The trap: if we let the tool do the defining and interpreting too we hollow out the very skill that makes us useful. Computational thinking is what keeps a human meaningfully on top in Wolfram’s “hybrid human-machine world,” instead of just clicking approve. This is also a nod to another article from back in August 2025 that argues for the importance to not fall into cognitive surrender and avoid being an AI idiot.
The new literacy
Wolfram compares computational literacy to reading and writing: once considered too hard for most people, now assumed to be universal, and a serious equity issue wherever access is uneven (Getting Smart Staff, 2020). That framing is relevant to higher ed. The analysts who thrive in the AI age won’t be the ones with the fanciest tools. The way I see it, they’ll be the ones who can define, abstract, compute, and interpret, and who know which of those steps to keep for themselves.
So to my fellow techies and non-techies alike, and to anyone hoping to become an analyst: the tool is not the skill. Learn to frame the question and interpret the answer, and let the machine do the long division. Wolfram would approve.
And yes, Microsoft Analyst is still a hero in my workflow. Who else is experimenting? Show and tell. 😄
Reference
Getting Smart Staff. (2020, August 5). Conrad Wolfram on computational thinking. Getting Smart. https://www.gettingsmart.com/podcast/conrad-wolfram-on-computational-thinking/