AI Reliance Study

A Student Led Guide to Using AI Wisely in School

We conducted experiments, conducted student interviews, and developed a basic framework to assist you in using AI as a learning aid, such as ChatGPT.

Our Purpose

To assist students in utilizing AI without sacrificing their own abilities..

Goal

Based on what actual students informed us, develop a straightforward, useful manual for appropriate AI use.

Research

To find out how a framework alters behavior, we conducted an experiment with eight students and conducted interviews with eight more.

Framework

5 evidence‑based principles to keep you learning and thinking, even when you use AI.

Impact

Our framework helped students to think more critically about AI and to create better work.

8
Students Interviewed
8
Experiment Participants
5
Evidence Based Principles
8
Months of Research

Meet the Team

Four Carleton University students who worked on this project.

Erin Van Der Pouw Kraan

Lead Researcher

InterviewsEthics

Managed the interviews and ensured that our study was fair and complied with regulations.

Victor Oikawa Lopes

Research Coordinator

FrameworkData

Created our five guiding principles based on the interview results and designed the experiment.

Mohammad Al-Saadi

Technical Lead

WebsiteCharts

Built this website and made our data easy to see with charts.

Mustafa Ali

Technical Coordinator

ResearchAnalysis

Also helped with the website and worked on the poster for the project.

Our Supervisor

James Brunet

Professor, School of Information Technology

Carleton University

Helped us keep on the right track to transform our concepts into actual research.

Our Journey

Sept 2025

Started the Project

Decided on our theme and the questions we wanted to address.

Oct 2025

Got Ethics Approved

Carleton gave us the go ahead to interview students..

Dec 2025

Interviewed Students

Asked eight students about their opinions on artificial intelligence.

Jan 2026

Built the Framework

Developed five evidence based guidelines based on research and interviews.

Feb 2026

Ran the Experiment

Eight students completed school tasks to test our framework.

Mar 2026

Data Analysis

Analyzed experiment results, made visualizations and conclusions.

Apr 2026

Shared Our Findings

Presented everything at the Carleton Capstone Fair.

Student Interviews: What We Heard

We had a conversation with eight Carleton students to learn about their experiences with AI before developing our framework.

Interview Methodology

Eight undergraduate students studying science, engineering, and information technology participated in semi-structured interviews with us for ten to twenty minutes. The objective was to use student viewpoints to design an ethical AI framework.

8
Participants
21-23
Age Range
50/50
Care about integrity?

Interview Questions

  • How do you feel about AI for schoolwork?
  • Do you know or care about academic integrity?
  • Does AI use go against academic integrity?
  • Where does AI belong in academia?
  • Positives of AI in academics?
  • Negatives of AI in academics?
  • Most important takeaway?

We identified 5 core themes that became our framework.

Key Takeaways from Interviews

- AI is valued for: explaining concepts, rewording, idea generation, summarizing, study prep.

- Major concerns: dependence, skill loss, cheating, incorrect information.

- Integrity split: 4 follow rules to learn, 4 avoid getting caught.

- Student insight: “AI is like a lighter, it can light a candle or burn down a building.”

These insights directly shaped the 5 principles of our framework.

The AI Learning Framework

Principles supported by research were created through student interviews and verified by testing.

How This Framework Was Constructed

Phase 1
Student Interviews

8 students discussed their experiences, worries, and usage.

Phase 2
Thematic Analysis

Answers to 5 main themes..

Phase 3
Literature Review

Verified by peer-reviewed studies.

Phase 4
Experimental Validation

8 students participated in the test, and the framework group used AI with greater consideration.

01

Recognize the “why” (AI literacy)

Understand how the AI came up with its answer. Don't simply copy.

What the study says

Gong et al.: Students do not understand the ethics of AI. Harvard/Stanford RCT (2025): Learning gains are doubled when AI is used as a "scaffolded tutor."

Try asking: "Walk me through the steps to solve this" rather than "what's the answer."

02

Always verify facts

AI creates details and sources, particularly for specialized subjects.

What the research says

Linardon et al. (2025): According to Linardon et al. (2025), fabrication increases from 6% to 29% for niche themes, with 65% of GPT-4o citations being unreliable.

Try this: Check Google Scholar for each citation.

03

Don’t develop bad habits

Before using AI, think about what you already know.

From interviews

The main issue was dependence. Students claimed that using AI before second guessing your own abilities can be detrimental.

Try this: Spend 10–15 minutes working alone on an issue.

04

AI is a tool, not a crutch

It is up to you how you utilize it, so be aware of its restrictions.

Try this: Try this: Write the final draft yourself after using AI to generate ideas.

05

You are responsible for the results

You are affected, not the AI.

Remember: If you misuse AI, you miss out on learning.

What We Found

Data from 8 Carleton students.

Participant Snapshot

8 undergraduate students (Science, Engineering, IT). Age 21-23. 7 men, 1 woman. CGPA 6.0–11.5.

50%
Care about integrity
75%
Positive view of AI

Academic integrity: do you care?

Split: 4 follow rules to learn, 4 just avoid trouble.

Perception of AI

Most see AI as helpful, but dependence worries remain.

AI usage frequency

5 frequent, 2 rare, 1 consistent (daily).

Self reported reliance

Only 2 of 8 consider themselves reliant.

The Experiment: Does the Framework Work?

Eight students completed five real world challenges to test our five concepts.

8 Students

4 got the framework, 4 didn't.

5 Tasks

Summarizing, grammar, story writing, scheduling, debugging.

About 1 Hour

Every student completed assignments at their own speed.

Experiment Tasks

Task 1: Summarize Article

Read a marketing article, summarize it, and add your own thoughts. (Complexity: High / Grade 8 level)

Task 2: Grammar & Rewrite

Fix grammar, spelling, and flow of a short story. (Complexity: Mid-High / Grade 5)

Task 3: Creative Writing

Write a 2‑paragraph story with specific prompts. (Complexity: Mid / Grade 4)

Task 4: Morning Routine

Organize a logical morning schedule. (Complexity: Low / Grade 1)

Task 5: Debug HTML

Fix 3 errors in a basic HTML page. (Complexity: Very High / University level)

Key Findings

More Thoughtful AI Use

Framework group used AI with caution, trying tasks first, double checking results, and followed principles.

Higher Quality in Creative Tasks

Framework group averaged 8.25/9 on Task 3 vs 7.25 for uncontrolled.

Slower, But Deliberate

Uncontrolled group completed tasks on average 7–12 minutes faster, however they relied on AI without verification

Task Complexity Levels

Complexity according to university level and the Ontario curriculum. Debugging was the most difficult task.

Time Spent Averages (minutes)

The framework group demonstrated deeper engagement by spending more time to difficult activities.

AI Usage Across Tasks

AI was used less often and always with verification by the framework group.

Average Grades by Task

Framework group performed better on creative & grammar tasks.

Task 4 Visualization: Morning Routine Ordering

The tasks were arranged properly by both groups, with minor differences in the sequence of breakfast and backpacks according to personal routines.

The Bottom Line

Our 5 principles had an impact. With the framework, students used AI with greater care, validated results, and maintained their own learning. Despite working more slowly, their work demonstrated greater ethical concern and critical thinking.

Resources

All our materials.

Framework Poster

View the 5 principles online. Download not required, explore above.

Full Research Summary

All details: methods, interview questions, experiment tasks, available for review.

Capstone Presentation

Slides we used at the fair, accessible by request.

Study Materials

Ethics & Consent

  • Consent Form (Interview), available upon request
  • Consent Form (Experiment), available upon request

Our Questions

  • Interview Guide (shared with participants)
  • Experiment Tasks (detailed in findings)

Data

  • Anonymized Results, contact us for access

Get in Touch

Questions about our research? Want to collaborate? Reach out!

Email Us Directly

erinvanderpouwkraan@cmail.carleton.ca

victoroikawalopes@cmail.carleton.ca

mohammadalsaadi@cmail.carleton.ca

mustafaali6@cmail.carleton.ca

Our Supervisor

James Brunet

JamesBrunet@cunet.carleton.ca