Google Effect Memory in LLM Chatbots

ISEF Category: Behavioral and Social Sciences

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

The Hook

If you know you can look something up later, your brain may stop saving the fact itself. That pattern is called the Google effect. Your project can test whether an LLM chatbot creates the same memory shift, because it can act like a second brain. With a jsPsych task, you can compare recall, confidence, and response time in a clean experiment.

What Is It?

The Google effect is a memory shift. When people expect easy access to information, they often remember where to find it instead of remembering the content itself. Psychologists call that cognitive offloading, which means handing part of the memory job to a tool.

A chatbot makes this question more interesting than a search box alone. An LLM can answer in full sentences, which may make the answer feel easier to recover later. Your study can check whether that changes free recall, recognition, and confidence after learning the same material.

Why This Is a Good Topic

This is a strong science fair topic because you can test it with a clear comparison, and the outcome is easy to measure. You can ask whether chatbot access changes memory, confidence, or response time, then compare those results across conditions. The topic also connects to a real problem students know well, how much your phone, search engine, or chatbot changes the way you remember.

Research Questions

• How does knowing a chatbot is available later affect immediate recall of facts? ?
• What is the effect of chatbot access on delayed recall after a short delay? ?
• Does chatbot access change confidence ratings even when recall stays the same? ?
• To what extent does prior use of chatbots affect how much students offload memory? ?
• Which item types, such as facts or short explanations, show the biggest memory drop when a chatbot is available? ?
• How does chatbot access compare with search engine access in changing memory performance? ?

Basic Materials

• Laptop or desktop computer with a modern browser.
• Stable internet connection.
• Free jsPsych experiment template or browser-based task page.
• Consent and survey form tool like Google Forms or Qualtrics.
• Spreadsheet software like Google Sheets or Excel.

Advanced Materials

• University lab computer with secure browser logging and file storage.
• Participant recruitment system with approved consent workflow.
• Quiet testing room with controlled lighting and seating.
• Optional eye-tracking system for attention measures.
• Optional EEG system for measuring mental load during information offloading.

Software & Tools

• jsPsych: Runs browser-based memory tasks and records trial-level responses.
• R: Cleans data, builds plots, and fits statistical models.
• RStudio: Gives you a free workspace for writing and running R code.
• JASP: Lets you run common statistical tests with a simple interface.
• Google Sheets: Helps you sort raw data, code responses, and track participants.

Experiment Steps

1. Define the memory outcome you will score, such as free recall, recognition, or delayed recall.
2. Choose one external memory condition to compare, such as chatbot access, search access, or no-help.
3. Build a trial flow that shows the same facts, prompts, and test order to every participant.
4. Plan controls for item difficulty, wording, and prior chatbot use so the groups stay comparable.
5. Decide how you will analyze recall, confidence, and response time, then check whether the effect holds across item types.

Common Pitfalls

• Letting one condition see easier facts, which makes the memory drop look bigger than it really is.
• Mixing the learning screen and the test screen, which lets students copy wording instead of recalling the idea.
• Using only trivia or only textbook terms, which hides whether the effect depends on item type.
• Ignoring response time, which can make fast guessing look like confident memory.
• Running too few participants, which leaves you with noisy results that shift from one class period to the next.

What Makes This Competitive

A stronger version of this project separates simple access from real memory change, then compares chatbot, search engine, and no-help conditions. You can also measure confidence, response time, and item type, not just right or wrong answers. If you add a careful analysis plan, such as mixed-effects models, you can show whether the effect holds across different people and different questions. That moves the project from a classroom demo to a serious cognitive psychology study.

Project Variations

• Compare chatbot access with search engine access and no external help to see whether the effect is specific to LLMs.
• Test factual items versus short explanations to see whether deeper material changes how much students offload memory.
• Add a delayed recall round to see whether chatbot access affects short-term and long-term memory differently.

Learn More

• PubMed: Search review articles on cognitive offloading, transactive memory, and the Google effect.
• NIH Research Matters: Find plain-language summaries on memory, attention, and digital cognition.
• APA Dictionary of Psychology: Look up terms like recall, recognition, working memory, and cognitive offloading.
• MIT OpenCourseWare: Read lecture notes on memory, learning, and experimental design.
• Noba Project: Browse free psychology chapters on memory systems and forgetting.
• Journal of Experimental Psychology: Learning, Memory, and Cognition: Search abstracts and open articles on memory offloading and external memory aids.