Question and hypothesis
Choose one course-linked mechanism and state a directional, falsifiable prediction.
Design · Build · Test · Explain
From course idea to behavioral experiment
Design a focused question, implement a browser-based task, collect anonymous data responsibly, analyze the result, and explain what it means.
New to building browser experiments?
Complete the Vibe Coding Experiment first. It is an individual practice assignment that rehearses prompting, debugging, data handling, visualization, and GitHub Pages deployment before the larger final project.
Workflow
Six connected stages
Each stage should reduce uncertainty for the next one. Do not start coding until the hypothesis, variables, and trial structure are explicit.
Design specification
Define the independent variable, dependent variable, conditions, trial sequence, exclusion rules, sample, and planned analysis.
Prototype and pilot
Build the smallest complete version, then test instructions, timing, edge cases, mobile behavior, and saved fields.
Ethical data collection
Recruit voluntarily, explain the general purpose, collect no direct identifiers, permit withdrawal, and restrict access to raw data.
Analysis and interpretation
Validate the dataset, apply the planned test, report effect sizes and uncertainty, and distinguish evidence from speculation.
Report and presentation
Connect the hypothesis, method, result, limitation, and neuroscientific implication into one coherent argument.
Before coding
Write the experiment on paper first
Scientific specification
- Research question and explicit hypothesis
- Manipulated and measured variables
- Within- or between-subjects structure
- Operational definitions
- Expected pattern and statistical test
- Connection to a neural or computational mechanism
Technical specification
- Screen-by-screen flow
- Practice and main-trial structure
- Stimulus timing and valid responses
- Fields saved on every trial
- Completion, failure, and withdrawal behavior
- Browser and mobile requirements
Data responsibility
Privacy by design
Raw participant data must not be public. Use random participant codes, never collect names or email addresses, limit database access to the research group and teaching team, and share only deidentified summaries. A public task URL does not imply a public data URL.
Collect less
Save only fields required to test the hypothesis. Avoid free-text fields and device fingerprints.
Control access
Do not use publicly readable spreadsheets or databases for participant-level records.
Explain clearly
Tell participants what is collected, why, whether it is stored, and how to stop.
AI-assisted development
Use AI as a coding collaborator
You remain responsible for the design, scientific reasoning, privacy, testing, and every behavior of the finished task.
A productive prompt includes
- Goal and participant flow
- Exact number and order of trials
- Timing and response rules
- Required data schema
- Privacy constraints
- Acceptance tests
A productive review checks
- Does the code match the design?
- Can a participant double-submit?
- What happens offline or on failure?
- Are reaction times measured correctly?
- Can private data be read publicly?
- Can every group member explain it?
Deliverable quality
What a complete project communicates
Web task
Clear instructions, practice, reliable timing, anonymous records, tested edge cases, and an accessible interface.
Scientific report
Motivation, explicit hypothesis, reproducible methods, inferential statistics, effect sizes, limitations, neuroscience, and contribution attribution.
Presentation
A concise question, hypothesis, task demonstration, readable figure, answer, limitation, and prepared Q&A.
Use Canvas for exact requirements. Current deadlines, weights, minimum reference counts, submission links, group assignments, and presentation schedules are intentionally not duplicated here.