Are you a visual learner?
Lots of people describe themselves as visual learners, and lots of teachers were trained to sort students into visual, auditory and kinesthetic. The idea that you learn best when lessons match your preferred style is one of the most popular beliefs in education. It also has very little evidence behind it.
In 2008, the psychologists Harold Pashler, Mark McDaniel, Doug Rohrer and Robert Bjork reviewed the research on learning styles. They looked for studies that tested the idea properly, by teaching people in their preferred style and in another style and comparing results. They found almost no studies designed that way, and the few that were didn't support the idea.
Your preferences are real. You may enjoy diagrams more than text, or find a video easier to sit through than a chapter. That matters for motivation. But matching the format to a "style" isn't what makes people learn.
What does help, for everyone
The better-supported idea is called dual coding. Pairing words with a relevant picture gives you two ways into the same idea. The psychologist Allan Paivio proposed it in the early 1970s, and later experiments on memory for words and pictures have broadly backed it.
Richard Mayer's research on multimedia learning builds on this, and a few of his findings are directly useful when you study with ChatGPT:
- Words and pictures beat words alone. A labeled diagram of the heart plus an explanation works better than the explanation by itself.
- Narration plus visuals beats on-screen text plus visuals. When you're looking at a diagram, hearing the explanation lets your eyes stay on the diagram. Reading a paragraph next to it splits your attention.
- Less is more. Decorative images, background music and interesting asides that aren't the point all make learning worse, not better.
- Point at what matters. Highlighting, arrows and circling the key part of a visual help people know where to look.
None of this depends on your "style". It's about how working memory handles information, and it applies to visual and verbal people alike.
What kind of visual fits what kind of idea
Different ideas need different pictures. Ask ChatGPT for the one that matches what you're learning:
- A process like digestion, a chemical reaction or how a bill becomes law. Ask for a numbered flowchart.
- A structure like a cell, an engine or a sentence. Ask for a labeled diagram.
- Change over time like a war, the history of an idea or a business cycle. Ask for a timeline.
- Comparisons like mitosis vs meiosis or two economic theories. Ask for a table with the same rows for each.
- Relationships between many concepts. Ask for a concept map that says how each pair is linked.
- Math of almost any kind. Ask for a worked example with small numbers, laid out one step per line.
How to get visuals from ChatGPT
Text diagrams and tables
ChatGPT is reliable at tables, timelines and step-by-step layouts in text. They're quick to read and easy to copy into your notes.
Make a comparison table of mitosis and meiosis. Use the same six rows for both: purpose, number of divisions, daughter cells, chromosome number, genetic variation, where it happens.
Flowcharts and concept maps
You can ask ChatGPT for a flowchart as Mermaid code and paste it into any Mermaid viewer to see it rendered. Or ask it to describe the boxes and arrows, and draw them by hand. Drawing it yourself is a form of active recall, so it's not wasted time.
Make a flowchart of the steps of cellular respiration as Mermaid code. Keep each box under six words.
Generated images
ChatGPT can generate images, which helps when you want a picture of something rather than a schematic. Be careful with labeled scientific diagrams. Image generators can still misspell labels or put a part in the wrong place, and that's easy to miss when you're new to a topic. For anything you'll memorize, check it against your textbook.
Worked examples
For math and physics, the most useful visual is often a calculation laid out one step at a time.
Show me how to find the derivative of x² sin(x), one step per line, with a short note on each line saying which rule you used.
Narrated visuals fit the research
Look back at Mayer's findings. The ideal is a simple visual, a spoken explanation, attention pointed at the key part, and nothing extra. That's a good description of a short narrated lesson, and it's what Lesson Video makes inside ChatGPT.
You ask ChatGPT for a lesson about what you've been discussing. It writes a few scenes and a script, and the lesson plays in the chat with a voice, simple animation and highlights on the part that matters. Captions follow the narration, which helps when your sound is off or English isn't your first language. Each lesson ends with a question, so you find out whether the picture actually helped.
To keep a lesson focused on the visuals, ask ChatGPT to keep the on-screen text short. Mayer's research found that long text repeating the narration on a busy screen tends to hurt more than help.
Make it stick
A good visual makes something easier to understand. It doesn't make it stick by itself. After you've looked at a diagram or watched a lesson:
- Redraw it from memory. Then compare it with the original. Whatever you left out is what you need to review.
- Explain it out loud. Walk through the diagram as if teaching someone. See the Feynman technique with ChatGPT.
- Get quizzed on it. Ask ChatGPT to quiz you on the process, without the diagram in front of you. Active recall with ChatGPT has the prompts.
Prompts to copy
- "What kind of diagram would explain this best? Then make it."
- "Turn this explanation into a numbered flowchart, one short step per box."
- "Make a timeline of these events with one line on why each one mattered."
- "Give me a table comparing these two ideas, same rows for both."
- "Make this a short lesson I can watch." This one works with Lesson Video added to ChatGPT.