AI Note-Taking and Long-Term Memory

ISEF Category: Behavioral and Social Sciences

Subcategory: Cognitive Psychology  ·  Difficulty: Intermediate  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

A note can look great and still fail you a week later. That is the real test of memory. If an AI summary acts like a perfect cheat sheet, does it help you remember, or does it make you feel fluent without lasting learning? You can test that with one clean comparison and a delayed free-recall task.

What Is It?

Generative-AI-assisted note-taking means you let a language model turn a reading, lecture, or transcript into a summary. Cornell notes mean you write cues, details, and a short summary yourself. Think of it like two ways to pack for a trip, one is a fast prepacked suitcase, the other is packing each item by hand. The big question is which method helps your brain keep the details after a delay.

This project focuses on long-term retention, which means what you remember later, not what feels easy right after reading. Free recall is a blank-page memory test where you write down everything you can remember without hints. A within-subject test means each student uses both note-taking methods, so you compare the same person against themselves instead of comparing two different groups.

Why This Is a Good Topic

This is a strong science fair topic because you can measure it with clear data, simple materials, and a real cognitive question. It connects to a real problem, how students should take notes when AI can summarize almost anything for them. You can learn how to control variables, score recall fairly, and use basic statistics to compare memory after a delay.

Research Questions

• How does generative-AI-assisted note-taking affect one-week free recall compared with handwritten Cornell notes?
• What is the effect of note-taking method on recall of main ideas versus specific details?
• Does the difference between AI summaries and Cornell notes change with passage difficulty?
• To what extent does note length predict delayed recall within each note-taking method?
• Which method leads to fewer incorrect ideas in delayed free recall?
• How does prior familiarity with the topic change the effect of note-taking method on retention?

Basic Materials

• Two comparable reading passages at similar difficulty level.
• Paper notebook or printed Cornell note template.
• Pen or pencil.
• Laptop or tablet with access to a free LLM chat tool.
• Stopwatch or digital timer.
• Printer or PDF reader for consistent passage display.
• Spreadsheet for recording recall scores.
• Randomization sheet for assigning passage order.

Advanced Materials

• IRB-approved survey or task platform.
• University participant pool or access to a larger sample.
• Qualtrics or similar survey tool for counterbalanced task delivery.
• PsychoPy or another cognitive testing platform.
• Secure data storage approved by the institution.
• R or JASP for within-subject analysis and effect sizes.
• Optional audio or screen recording setup for protocol checks.
• Access to a faculty reviewer for study design feedback.

Software & Tools

• Google Forms: Collects participant responses and free-recall text in a structured format.
• Google Sheets: Organizes scores, note lengths, and counterbalancing assignments.
• JASP: Runs paired tests, effect sizes, and simple within-subject comparisons without coding.
• R: Handles more advanced statistics and custom plots for repeated-measures data.
• Qualtrics: Builds a cleaner counterbalanced survey if you have access through school or a university.

Experiment Steps

1. Define the exact comparison, including what counts as an AI summary, what counts as Cornell notes, and what you will score in recall.
2. Select two reading passages that match in length and difficulty so the note-taking method, not the text, drives the result.
3. Plan a within-subject counterbalance so each participant uses both methods, but on different passages and in different orders.
4. Build a scoring rubric that separates correct facts, main ideas, and added errors so your data reflect memory quality, not just word count.
5. Pilot the task with a small sample to check whether the passages, prompts, and scoring rules are fair and easy to follow.
6. Pre-plan your analysis, including paired comparisons, effect sizes, and any subgroup checks for prior familiarity or note length.

Common Pitfalls

• Letting the AI summary add facts that were never in the passage, which inflates recall unless you score only verified ideas.
• Using passages that differ in difficulty, which makes the note-taking method look stronger or weaker for the wrong reason.
• Comparing polished AI summaries with messy handwritten notes without tracking note length, which turns quantity into a hidden variable.
• Testing recall too soon after note-taking, which measures short-term memory instead of one-week retention.
• Scoring only total words recalled, which misses whether the student remembered the right ideas and accurate details.

What Makes This Competitive

A competitive version would use tight counterbalancing, a clear scoring rubric, and a deeper analysis than simple recall totals. You could separate facts, ideas, and errors, then test whether AI changes accuracy, confidence, or source memory. A stronger project might also compare topic types or passage complexity, which can reveal when AI helps and when it hurts. That turns a simple classroom comparison into a sharper cognitive study.

Project Variations

• Compare AI summaries, Cornell notes, and self-written summaries on the same passages.
• Test whether AI note-taking helps more with dense science passages than with narrative passages.
• Measure both delayed recall and source memory to see whether AI changes accuracy, confidence, or both.

Learn More

• MIT OpenCourseWare: Search for cognition or learning lectures that explain how memory works.