Research Methodology

Understanding our research approach and design

Quick Facts
  • Student respondents
  • Multiple universities
  • ~3-5 min survey
  • Ongoing data collection
Jump To
Sample Controlled Variables Measures Analysis Plan Limitations

Research Question

"How do university students perceive the use of AI in cancer diagnosis, particularly regarding trust, fairness, and the role of human oversight?"

This research aims to understand student attitudes toward AI-assisted medical diagnosis, with a focus on ethical considerations and policy preferences.

Sample

Our target population is undergraduate and graduate students at Canadian universities.

Inclusion Criteria
  • Currently enrolled at a participating university
  • 18 years of age or older
  • Able to complete the survey in English
Participating Universities
  • Various Canadian universities
  • International students welcome
  • Students from all disciplines
Sample size: Data collection ongoing. Results will be reported in aggregate.

Controlled Variables

These variables allow us to compare attitudes across different demographic groups:

Variable Type Categories
Illness Experience Independent Personal / Family / None
Field of Study Independent Health / CS / Business / etc.
AI Familiarity Moderator 1-4 Likert Scale
Ancestry Independent Multi-select
University Control 5+ Universities
Why These Variables?

The illness experience variable tests whether personal experience with serious illness affects attitudes toward AI. The ancestry variable allows us to examine whether concerns about AI bias vary across different communities - directly relevant to equity considerations in the case.

Measures

Question Types
  • Single Choice: University, field of study, year
  • Multi-Select: Ancestry (analyzed separately per option)
  • Likert Scales: Comfort, importance (1-5)
  • Open Response: Thematic coding planned
Key Constructs
Trust in AI

Measured via comfort level and willingness to use

Ethical Concerns

Bias, transparency, privacy (select up to 3)

Safeguard Preferences

Doctor oversight, bias testing, regulation

Policy Support

Disclosure, consent, oversight requirements

Analysis Plan

Descriptive Statistics
  • Frequency distributions for all categorical variables
  • Mean and standard deviation for Likert scales
Comparative Analysis
  • Chi-square tests for categorical group comparisons
  • t-tests for comparing means across binary groups
  • Cross-tabulations for multi-variable analysis
Key Comparisons Planned
Illness Experience

Do those with experience show different trust patterns?

Field of Study

Do STEM vs. non-STEM students differ?

Ancestry

Do bias concerns vary across communities?

AI Familiarity

Does knowledge affect trust?

Limitations

Every research study has limitations. We acknowledge the following:

Sample Limitations
  • Convenience sample: Respondents are self-selected, not randomly sampled
  • University-only: Results may not generalize to the general population
  • Sample size: Current N=127 limits statistical power for subgroup analyses
Response Bias
  • Students with strong opinions may be more likely to respond
  • Tech-aware students may be overrepresented
  • Attitudes may change as AI technology evolves
Data Quality
  • Self-reported data (no verification of responses)
  • Multi-ancestry responses handled by counting each ancestry separately
  • Open-ended responses require thematic coding (planned)

Ethics & Privacy

  • Anonymity: No personal identifying information collected
  • Consent: All participants must consent before starting
  • Age verification: Participants must confirm they are 18+
  • Data security: Responses stored locally; reported only in aggregate
  • Voluntary: Participants can stop at any time
View Survey Questions