AI Disclosure Labels and Teen Essay Persuasion Effects

ISEF Category: Behavioral and Social Sciences

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

The Hook

A tiny label can change how the same essay feels. If readers see "written by AI," they may trust the message less, even when the argument stays strong. That gives you a neat way to test persuasion, bias, and attitude change in one study. You can ask whether the label itself becomes part of the message.

What Is It?

Think of a persuasive essay like a pitch and the AI label like a sticker on the package. The words still say the same thing, but the sticker can change how you judge the writer before you even finish reading. In social psychology, that extra judgment can shape attitude change, which means how far someone's opinion moves after exposure to a message.

Your 2x2 experiment changes two things, argument quality and AI disclosure. Strong and weak essays let you see whether people react to the argument itself or to the label around it. By comparing pretest and posttest opinions, you can separate a simple liking score from actual persuasion.

Why This Is a Good Topic

This topic works well because you can test it with a web survey, a small sample, and clear before-and-after measures. It connects to a real question teens face now, whether AI labels change trust in school, media, and online writing. You can learn how to build a clean experiment, compare four conditions, and use simple statistics to look for an interaction effect.

Research Questions

• How does an AI disclosure label affect attitude change after reading a strong persuasive essay?
• How does an AI disclosure label affect attitude change after reading a weak persuasive essay?
• What is the effect of argument quality on attitude change when the essay is labeled as AI-written?
• To what extent does an AI disclosure label change perceived trust in the writer?
• Which combination of label status and argument quality produces the largest shift in opinion?
• Does prior use of chatbots change how strongly the AI label affects attitude change?

Basic Materials

• A laptop or Chromebook with internet access.
• Google Forms or Microsoft Forms for random assignment and response collection.
• Spreadsheet software such as Google Sheets or Excel for cleaning data and charts.
• A word processor for drafting matched essay versions.
• Printed or digital consent and assent forms for participant recruitment.
• A quiet place to run the survey without distractions.

Advanced Materials

• Qualtrics with randomization and attention checks.
• R or JASP for factorial analysis and effect sizes.
• PsyToolkit or jsPsych for browser-based experimental flow.
• An encrypted institutional drive for de-identified data.
• A validated attitude scale and source-credibility scale.

Software & Tools

• Google Forms: Randomizes participants, collects pretest and posttest answers, and exports a clean spreadsheet.
• JASP: Runs factorial tests, effect sizes, and plots without coding.
• R: Cleans responses, calculates attitude change scores, and makes publication-style figures.
• G*Power: Estimates how many participants you need for a 2x2 design.

Experiment Steps

1. Define the attitude you want to change, and write one scale that measures it before and after the essay.
2. Draft two essays on the same issue, then make one version clearly stronger than the other without changing the topic.
3. Add or remove the AI disclosure label while keeping every other detail the same across conditions.
4. Plan random assignment, attention checks, and exclusion rules before anyone sees the survey.
5. Decide how you will score attitude change and compare the four groups with a 2x2 analysis.
6. Pilot the survey with a few readers, then fix confusing wording before full data collection.

Common Pitfalls

• Changing both the topic and the label at once, which makes it impossible to tell what caused the attitude shift.
• Writing the AI-labeled essay in a less polished way, which turns label effects into writing-quality effects.
• Using a weak attitude scale, which misses small but real opinion changes.
• Skipping random assignment, which lets participant bias shape your results.
• Collecting too few responses in each cell, which makes the 2x2 comparison noisy and hard to trust.

What Makes This Competitive

A stronger version of this project does more than ask whether people like AI essays. It tests the interaction between disclosure and argument quality, uses matched essay pairs, and reports effect sizes with confidence intervals. You can also check whether the label works differently for students who use AI often, versus students who rarely do. That turns a simple survey into a sharper social psychology study.

Project Variations

• Compare a bare AI label with a short note that explains how the AI helped.
• Test the same design with school policy essays, health essays, or environmental essays.
• Measure whether trust in the writer, not just attitude change, shifts when readers see the label.

Learn More

• PubMed: Search review articles on persuasion, source credibility, and adolescent media literacy.
• NIH PubMed Central: Find full-text studies on disclosure, trust, and attitude change.
• APA Dictionary of Psychology: Check plain-language definitions for attitude, persuasion, and credibility.
• MIT OpenCourseWare: Look for free social psychology lectures on attitudes and persuasion.
• Journal of Experimental Social Psychology: Search for recent papers on message evaluation and disclosure effects.
• OSF: Find preregistration examples and open survey materials for experiments.