What’s next: The future of data literacy education

[#1]

We have come a long way in four episodes. We started with what makes JMP Student Edition genuinely useful in a teaching context before AI enters the picture. We looked at how AI can change the instructor’s role, explored what JMP Assistant and LearnBot actually do in practice, and then spent some honest time examining the risks, in particular, where AI gets in the way and how to build guardrails into course design.

Now I want to step back and ask the bigger question: where is all of this going?

I don’t have a crystal ball. Nobody does. But I have more than 15 years of working with university instructors across Europe, watching how they adapt – slowly at first, then suddenly – to new tools and new pressures. And what I am seeing right now feels different from previous transitions. Not because AI is magic, but because it is forcing a conversation that statistics educators are now having in earnest: what are we actually trying to teach?

The curriculum question that AI forces us to answer

For decades, introductory statistics courses were organized around procedures. Teach the t-test. Teach ANOVA. Teach regression. It made sense. Students need to understand what these methods do, when they apply, and what assumptions they rest on. What is currently under pressure is not that knowledge itself, but the idea that covering procedures is sufficient. AI can navigate the procedural landscape; what it cannot do is reason about whether the right procedure was chosen or what the result actually means in context.

Now AI can navigate the procedural landscape too. Not perfectly, not reliably, but well enough to make a student feel like the question is answered. AI is forcing us to answer a question we have delayed for years: what do we actually want students to learn?

I think the answer is: everything that matters most. Statistical thinking. The ability to translate a real-world question into a data problem. Judgment about whether a result is trustworthy. The discipline to ask “wait, does this actually make sense?” and to mean it.

There is an important parallel here. The problems where human statistical judgment tends to matter most – novel data, rare situations, genuinely complex analytical questions with no established template – are often where AI tools are least reliable. LLMs are trained on abundance, but they can struggle with rarity and novelty. The exploratory, creative, and innovative dimensions of data analysis remain fundamentally human territory. That is not a consolation prize. It is the argument for why statistical thinking still needs to be taught and taught well.

The profession is responding, slowly but seriously

The statistics education community has not ignored this. The GAISE College Report [13], the most widely cited framework for introductory statistics education, is currently undergoing a new revision [14], explicitly driven by the arrival of AI, which is significant. It means the field is treating it as a genuine curriculum challenge, not just a classroom management problem.

The live draft is worth reading, not because it has all the answers, but because the questions it is wrestling with are exactly the right ones. What should students be able to do with data that AI cannot do for them? What does it mean to understand a statistical result in 2026?

I have been following this revision with genuine interest. If the outcome is a framework that puts statistical reasoning, judgment, and interpretive skill at the center, rather than procedural coverage, it will give instructors the professional backing to redesign their courses in ways that AI cannot easily undermine.

What remains unresolved

I want to be honest about the issues for which we do not yet have good answers. Three questions are worth sitting with:

Assessment. If AI can complete most traditional statistics assignments, what does meaningful assessment look like? Oral examination? Portfolio-based evaluation? Live data problems with no advance preparation? These are not new ideas, but implementing them at scale in large undergraduate courses is genuinely hard. Nobody has a clean answer.

Neil Selwyn [10][11] raises a sharper version of this: as we redesign assessment to accommodate AI, are we at risk of making education machine readable, optimized for what AI can evaluate rather than what humans actually need to learn? That concern deserves more attention than it currently gets.

Equity. Access to AI tools is uneven. Students at well-resourced institutions with good digital infrastructure and strong instructor support will have very different experiences from those who do not. If AI fluency becomes a core graduate competency, we need to be honest about who is being prepared and who is not.

What good looks like. We are still early enough in this transition that there are very few well-documented examples of AI integration in statistics education that unambiguously worked, where students learned more, reasoned better, and left with stronger analytical judgment than they would have without AI. I believe those examples exist. I would like to see more of them documented and shared.

“The best analysts, engineers, and scientists will always stay one step ahead of the tools they use.”

 

Where JMP fits in this picture

I have been deliberate throughout this series about not turning it into a product pitch. JMP Student Edition is a tool, a genuinely good one for teaching, in my view, but a tool nonetheless. The pedagogical arguments in this series stand or fall on their own merits, not because JMP makes them.

That said, I do think the design philosophy behind JMP’s approach to AI – tools that augment analytical reasoning rather than replace it, that keep the visual and exploratory workflow at the center, that make the thinking visible rather than hiding it – is the right direction. Whether other tools will converge on similar principles or whether the market will push toward more automation and less transparency remains to be seen.

What I am confident about: instructors who understand what they want students to learn and who design their courses around that understanding will navigate this transition well, regardless of which tools they use. The pedagogy comes first. The tools follow.

An open ending

I started this series by going back to the early 1990s, to a time when AI was a research topic, not a browser tab, and when statistical reasoning had to be developed through genuine struggle with data and method. I am not nostalgic for that time. The tools we have now are genuinely better, and students today have access to analytical capabilities that would have astonished my doctoral supervisor.

But Advait Sarkar’s warning [7] stays with me: the risk is not that AI is too powerful, but that we stop thinking before we have had a chance to develop the judgment to evaluate what AI produces. Ethan Mollick [9] is right that AI will be a co-teacher, a coach, a collaborator. And the structural concerns that Selwyn [10][11] raises about who benefits from AI in education are questions every instructor should be pondering.

The people who stand out in the future will not be those who can run an analysis, but those who can look at the output and say: “I’m not convinced and here is why.” That is the skill worth teaching. And it is exactly the skill that statistics education, done well, has always been about.

If you have been using JMP Student Edition, LearnBot, or Assistant in your teaching, what’s worked? What has surprised you? What question are you still thinking about? I would love to hear from you directly at [email protected].

Software versions used in this series

JMP Student Edition 19.1.3 • Data Analysis Director 4.0 • LearnBot 1.1.1 • JMP Assistant 2.3

References

[1] JMP Academic Program, jmp.com/academic

[2] JMP Student Edition, jmp.com/student

[3] JMP Community

[4] Data Analysis Director (JMP Marketplace), marketplace.jmp.com

[5] LearnBot (JMP Marketplace), marketplace.jmp.com

[6] JMP Assistant (JMP Marketplace), marketplace.jmp.com

[7] Advait Sarkar: “How to stop AI from killing your critical thinking”, TED Talk, 2025

[8] Amy Bruckman: “Reinventing Teaching After The Introduction Of AI”, TEDx Talk, 2026

[9] Ethan Mollick: Co-Intelligence: Living and Working with AI, Portfolio/Penguin, 2024

[10] Neil Selwyn: Should Robots Replace Teachers?, Polity Press, 2019

[11] Neil Selwyn: On the limits of artificial intelligence in education, 2024

[12] Russ Wolfinger: The Top Seven Ways Scientists and Engineers Should Prepare for the AI-Driven Era, JMP white paper

[13] GAISE College Report (2016)

[14] GAISE College Report revision (in progress, 2025)

[15] Lost in JMP? Slimming Down for a Better Fit, Kraft and Rijpkema, JMP Discovery Summit Europe 2026

About the Author

Volker Kraft is Principal Academic Ambassador (EMEA) at JMP Statistical Discovery (a SAS company), part of the JMP Global Academic Team. Since 2011, he has supported universities across Europe in integrating JMP into teaching and research. His background is in speech technology research, a field where statistical rigor and engineering pragmatism had to coexist long before AI became a household word. He can be reached via the JMP Community or on LinkedIn.

Episode 5 of 5  •  End of series