Structured workflow
The workflow starts with structured experiment data. Phases, routines, components, timing, variables, canvas layout, and data fields stay readable before the task is compiled.
- ExperimentIR draft
- Editable routines
- Canvas placement
- Export field table
Human-editable after generation
The builder separates experiment design from code output. Reviewers can adjust the task in visual and form-based controls, then compile the same structure into browser-runnable output.
- Instruction screens
- Practice blocks
- Formal trials
- Loops and branches
- Result or feedback routines
Quality gate
The audit gate is designed to catch common launch failures before a participant sees the task: blank screens, missing buttons, invalid branches, weak export fields, low trial counts, and failed runtime smoke tests.
- Timeline scan
- Runtime smoke
- Data contract
- Publishing readiness
Best-fit tasks
A good first use case is a small experiment that needs clean response mapping and clear rows in the exported data: Stroop, Simon, Flanker, Go/No-Go, or a short survey followed by a timed task.
- Response key mapping
- Correctness fields
- Reaction time fields
- Condition labels
- Session identifiers
Public product routes
The public builder preview shows the experiment workflow directly. The assessment homepage covers public tests, while the experiment overview documents editable structure, browser delivery, and review checks.
- Public assessments: https://www.dopsycho.com/
- Experiment overview: https://www.dopsycho.com/blog/ai-psychology-experiments
- Builder flagship: https://www.dopsycho.com/solutions/jspsych-experiment-generator
FAQ
What does the jsPsych experiment generator create?
The generator drafts structured experiment data, then a compiler creates a browser-runnable jsPsych-style task from that structure.
What can be edited after generation?
Editors can review and change phases, routines, loops, branches, randomization, stimuli, timing, canvas placement, keys, variables, and export fields.
What does the audit layer check?
The audit layer checks empty screens, missing response paths, weak data fields, broken flow references, runtime smoke results, and publishing readiness.
Which experiments fit this workflow?
Good first targets include Stroop, Simon, Flanker, Go/No-Go, reaction-time sorting, attention checks, and short survey-plus-task designs.