Preference and effect are not the same
I like pie. My preferred way to use data is with a pie chart.
No one says that. At least, as a data expert, I haven’t heard much about it. If you said something like that in a meeting, you’d get a polite laugh reserved for people who might be joking. Because we all understand, instinctively or by experience, that liking something has nothing to do with whether it helps us understand something or make something better. More precisely, the message or story you share will determine the most effective way to convey it to your audience. When you get something you like, you feel more satisfied and motivated. However, that is different from studying.
A pie chart of 14 nearly identical slices tells us nothing. You can also see that pie charts that compare values over time don’t give you anything (and can be misleading). And no matter how personal your pastry preferences may be, it doesn’t change what the human eye can and cannot judge. We are not good at comparing angles. We are good at comparing lengths. Not to mention the raging 3D chart!!
That’s why boring bar graphs keep winning. There are no votes for your preference. As data professionals, we choose visualizations based on data. Whether you want to see more pie is entirely up to you.
But I prefer pie charts.
of course. It may also be fun to read quarterly financial information as a limerick. The question has never been about what you enjoy. The question is, what helps you confirm data insights?
learning pie chart
Which brings me to a never-ending story about learning and learning styles. We’ve been doing this with a straight face for decades. When someone declares, “I’m a visual learner,” instead of a polite laugh, you’re greeted with a redesigned, graphics-heavy course. We survey people about their preferred learning style. We categorize them into visual, auditory, and kinesthetic buckets. Build your content accordingly. We call this learner-centered design, and we feel that’s a good thing.
When researchers went looking for evidence that matching instruction and learning styles improves learning, they found no weak evidence. They found essentially nothing, and the few well-designed studies contradicted this idea (Pashler, McDaniel, Rohrer, & Bjork, 2008). However, a systematic review found that approximately 89% of educators still believe it is important to match instruction to learning styles, and most report that they actually do it (Newton and Salvi, 2020). And now AI is being trained on all of those myths. No wonder AI agents are just as confused as humans.
confusion
We are confusing two different questions. “What do people like?” is the real question. It’s worth asking. Preferences influence motivation, and motivation is important. “What actually works?” is another question. It has different answers. And when the two conflict (as they often do), “what works” must win. Because the learner’s goals are never fulfilled. It was about making something better.
If you want to see everything in a pie chart, you don’t need any more pie charts. They need to show someone a bar graph and say, “Look how quickly you found the answer.” If you call yourself a visual learner, there’s no need to translate your safety training into infographics. They need search practice, spacing, feedback, worked examples, and down-to-earth evidence with real evidence, no matter what sensory channels you claim as a brand.
Measurement is the key
My preference is data. But that’s data about comfort, not effectiveness. By designing accordingly, you can optimize for how learning feels, not whether it happens. These two measures diverge more often than we would like. Experiences that are smooth and enjoyable generally have lower retention rates than those that require effort and are slightly unpleasant. If you’ve ever seen glowing course ratings next to flat performance numbers, you’ve seen a divergence occur.
So the next time someone asks for a training version of a pie chart (shorter, prettier, more tailored to their style), take their request seriously as a motivational signal. Then ask better questions. Instead of “What do you like?”
“What does it take to make measurable progress in 90 days?” And measure it. No one answered “increase the pie” to this question.
Wait, AI makes things even worse.
There is one thing that has quietly protected us from our own bad theories for decades. That’s the cost. The cost of resources, time, and effort to create content that matches your “learning style.” It cost money to build three versions of the course (visual, auditory, and kinesthetic) for just three of the 75 different styles. Budgets led to tradeoffs, tradeoffs led to questions, and somewhere in the process someone always asked, “Wait, do we really need this?” Friction was our accidental quality control.
That friction is gone. Get ready for your personalized pie chart course.
AI can now generate a version of your course for each learner. All 3000 instead of 3 styles. Podcast version for “auditory learners”. Infographic for “visual learners”. Simulation for those who checked “Practice”. Each one is created in minutes, each polished and personalized to the point where humans can’t keep up. The dashboard lights up. Learners will report that they like it. Sales reps can play interesting AI stories about product knowledge twice as fast. Win-win!
Did I mention the measurement issue?
From now on, that’s all that matters. If the underlying theory on which AI is trained is wrong, then we are not actually producing personalized learning. We industrialized the pie chart. It speaks in 3D with rainbow-colored shading.
AI does not validate design assumptions. It amplifies them. You give it a myth, and it executes that myth perfectly and on a large scale, with proof-like confidence. In the past, bad ideas would slowly fail in one course until someone might notice. Entire enterprises can now fail spectacularly, wrapped up in the word “adaptive.”
Hidden within the first trap is the second trap of measuring the wrong thing. When an AI is trained to satisfy learners by giving them what they want instead of what they need, satisfaction scores skyrocket. Smooth, engaging, and fun. It feels good to learn. The problem is that that conflict makes the practice stick. An unusual format that attracts attention. Wrong answers must sit before being revealed. Satisfaction-oriented optimization loops are happy to optimize learning immediately after learning.
None of this makes AI a bad thing. The same machine that can generate infinite pies is the same machine that can finally do things we couldn’t do before. Adapting to what the learner knows rather than what he or she likes, producing retrieval practice at the edge of someone’s ability, spacing it out over several weeks, and intentionally changing the format to break the comfort of the style rather than match it.
We humans are responsible for how we implement AI for learning. We humans have a responsibility to select and influence technology vendors’ “AI solution” approaches. As humans, we have a responsibility to say “no” to courses we don’t need, and “yes” to solutions that may take us outside our traditional expertise. You don’t need to write a white paper to convert learning style believers. Measures need to be put in place to show the impact on work. It’s not just a memory or recall after a program, but an actual, lasting change in behavior that occurs under realistic circumstances.
A “feed of content” is as scary as a 3D pie chart for all your data. The problem isn’t AI. The problem is humans. We have an opportunity to stop hiding behind the lack of resources and technology to do the right thing. No more excuses. But if the system doesn’t think and focus on the behavioral changes that drive impact, this “feed of content” can literally become just a PDF you upload and a pie chart magically created from it.
No pie was harmed during this article.
References: Pashler, H., M. McDaniel, D. Rohrer, and R. Bjork. 2008. “Learning Styles: Concepts and Evidence.” Psychology in the Public Interest 9 (3): 105–19. Newton, Chancellor, and A. Salvi. 2020. “How commonly believed are neuromyths about learning styles? Do they matter? A practical systematic review.” Frontiers in Education, 5, 602451.
Share with
