Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.assesskit.com/llms.txt

Use this file to discover all available pages before exploring further.

AI Assistant workflow showing natural language to experiment creation

​
AI Assistant

The AI Assistant helps you build experiments faster using natural language descriptions. Describe what you want in plain English, and the AI creates components, modifies settings, and provides research-backed methodology guidance.

​
Overview

The AI Assistant can:
  • Create complete experiments from descriptions - “Create a Stroop task” → full experiment structure
  • Modify existing components - “Change all fixations to 1000ms” → bulk updates
  • Add new components - “Add instructions explaining the task” → inserts component with content
  • Answer methodology questions - “How many trials for a Stroop task?” → research guidance
  • Suggest improvements - Reviews your experiment and recommends enhancements
Benefits:
  • Save time building experiments
  • Learn experiment design best practices
  • Get research-backed recommendations
  • Iterate quickly on experiment designs
  • Build complex structures with simple requests

​
What the AI Can Do

​
Create Complete Experiments

Describe the experiment type, and AI generates the full structure. Example requests:
  • “Create a Stroop task with 160 trials”
  • “Build a 5-question mood questionnaire”
  • “Design a visual search experiment”
  • “Make a 2-alternative forced choice task”
What AI generates:
  • Instructions - Welcome message and task explanation
  • Component structure - Appropriate sequence (fixation → stimulus → response → feedback)
  • Timeline configuration - Research-backed timing defaults
  • Variables - Trial-by-trial parameters (for multi-trial tasks)
  • Frames - Loops and randomization where appropriate
Example output for “Create a Stroop task”:
  1. Consent form component
  2. Task instructions component
  3. Practice trials (5 trials with feedback)
  4. Begin main task transition
  5. Fixation component (500ms)
  6. Stimulus component (word display, 2000ms)
  7. Response component (F/J keys)
  8. Feedback component (1000ms, errors only)
  9. Loop frame (160 trials)
  10. Timeline variables (160 rows: word, color, congruency, correctKey)
  11. Debrief component

​
Modify Existing Components

Change properties across one or many components. Single component modifications:
  • “Change the fixation duration to 1000ms”
  • “Make the button blue”
  • “Add a border to the image”
Bulk modifications:
  • “Change all fixations to 1000ms”
  • “Make all buttons the same size”
  • “Update all text to 18px font”
Targeted modifications:
  • “For the Stroop stimulus component, change word duration to 1500ms”
  • “In the consent form, change button text to ‘I Agree’”
Advanced modifications:
  • “Add a minimum view time of 3 seconds to the consent form”
  • “Set response timeout to 2000ms for all response components”

​
Add New Components

Insert components into existing experiments. Adding specific components:
  • “Add instructions explaining the task”
  • “Add a break screen every 50 trials”
  • “Add a practice trial before the main task”
  • “Add a thank you message at the end”
AI determines:
  • Appropriate component type for request
  • Content for the component
  • Where to insert in sequence
  • How to connect to existing flow
Example: “Add instructions explaining the Stroop task”
  • AI creates Instruction component
  • Writes explanation of Stroop task
  • Inserts before practice trials
  • Adds “Continue” button

​
Answer Methodology Questions

Get research-backed advice on experimental design. Common questions:
  • “How many trials do I need for a Stroop task?”
  • “What’s the standard fixation duration?”
  • “Should I use blocked or randomized design?”
  • “What’s a good inter-trial interval?”
  • “How long should stimulus presentation be?”
  • “Do I need practice trials?”
AI provides:
  • Research-backed answers - Based on psychology literature
  • Specific recommendations - Concrete values, not just theory
  • Rationale - Why these values are standard
  • Alternatives - When to deviate from defaults
Example response to “How many trials for Stroop?”: “Standard Stroop tasks use 120-200 trials. I recommend 160 trials:
  • 80 congruent (word matches color)
  • 80 incongruent (word doesn’t match color)
  • Divided into 4 blocks of 40 trials each
  • Provides sufficient power for within-subject design
  • Allows counterbalancing across conditions”

​
Suggest Improvements

AI reviews your experiment and recommends enhancements. Request: “Review my experiment and suggest improvements” AI checks:
  • Missing instructions or consent
  • Unclear task explanations
  • Inappropriate timing
  • Missing practice trials
  • Poor trial organization
  • Accessibility issues
  • Missing data collection
Suggestions might include:
  • “Add practice trials with feedback before main task”
  • “Instructions don’t explain response key mappings - I can add this”
  • “Stimulus duration (500ms) is quite brief - standard is 2000ms for Stroop”
  • “Consider adding breaks every 50 trials (currently no breaks in 200 trial experiment)“

​
Opening the AI Chat

​
Where to Find AI Assistant

In Task Editor:
  1. Click AI Assistant button in top toolbar
  2. Or press keyboard shortcut (if configured)
  3. AI chat panel opens on right side
Standalone AI mode:
  • Some views may have dedicated AI button
  • Click to open focused AI interface

​
Chat Interface Overview

The AI chat interface provides: Message area:
  • Your messages - What you type
  • AI responses - AI answers and actions
  • Action items - What AI did (components created, properties modified)
  • Reasoning - Why AI made certain choices (optional display)
Input box:
  • Type your request or question
  • Multi-line support (Shift+Enter for new line)
  • Character count (shows remaining characters)
Controls:
  • Send button - Submit message
  • Clear chat - Start fresh conversation
  • Mode toggle - Switch between Action and Plan mode
  • Attach context - Reference specific components (if supported)
History:
  • Previous messages persist in session
  • AI remembers earlier context in conversation
  • Can reference previous requests

​
AI Modes

The AI Assistant has two modes for different workflows.

​
Action Mode

What it does:
  • AI makes changes immediately
  • No review step
  • Changes applied as soon as AI responds
When to use:
  • Quick modifications you’re confident about
  • Simple additions (adding component)
  • Exploratory changes you can easily undo
  • When you trust the AI’s judgment
Example: 
You: "Change fixation to 1000ms"
AI: "Done. Updated fixation duration to 1000ms."
[Fixation component updated immediately]
Benefit: Fast iteration, no interruption to workflow.

​
Plan Mode

What it does:
  • AI shows plan before executing
  • You review and approve changes
  • Changes only applied after your approval
When to use:
  • Major changes to experiment structure
  • Bulk modifications across many components
  • When learning what AI can do
  • When you want to verify before applying
  • Complex requests where AI might misunderstand
Example: 
You: "Create a Stroop task"
AI: "I'll create a Stroop task with the following structure:
     1. Consent form
     2. Instructions
     3. Practice trials (5)
     ...
     Do you want me to proceed?"
[You review plan]
You: "Yes" or "No, modify..." or "Cancel"
[AI executes if approved]
Benefit: Safety, learning, verification before changes.

​
Switching Modes

Toggle between modes:
  • Click mode selector in AI chat header
  • Or specify in message: “In plan mode, create a Stroop task”
Default mode:
  • Usually Action mode for speed
  • Plan mode can be set as default in settings

​
Creating Experiments with AI

​
From Scratch

Start with empty experiment, describe what you want. Common experiment types AI knows: Cognitive tasks:
  • “Create a Stroop task”
  • “Build an N-Back task”
  • “Make a flanker task”
  • “Design a visual search experiment”
  • “Create a lexical decision task”
  • “Build a go/no-go task”
Questionnaires:
  • “Create a mood questionnaire with 5 questions”
  • “Build a demographics survey”
  • “Make a personality assessment”
  • “Design a satisfaction survey”
Social psychology:
  • “Create an implicit association test (IAT)”
  • “Build a trust game”
  • “Make a social judgment task”
What AI generates automatically:
  • Instructions explaining the task
  • Appropriate component types
  • Research-backed timing defaults
  • Trial structure and randomization
  • Data collection configuration
  • Variables for multi-trial tasks
Customizing AI output: After AI creates initial structure, refine:
  • “Make the instructions more detailed”
  • “Change stimulus duration to 1500ms”
  • “Add a practice block before main trials”
  • “Increase to 200 trials instead of 160”

​
Modifying Experiments

Refine existing experiments with AI. Timing changes:
  • “Change stimulus duration to 1500ms”
  • “Make all fixations 1000ms”
  • “Set response timeout to 3000ms”
Content changes:
  • “Update instructions to be clearer”
  • “Change button text to ‘Continue’”
  • “Add example to instructions”
Structure changes:
  • “Add a break every 50 trials”
  • “Insert practice trials before main task”
  • “Add debrief at the end”
Visual changes:
  • “Make all text 20px”
  • “Change button color to blue”
  • “Center all elements”

​
AI Prompt Best Practices

Get better results with well-crafted prompts.

​
Be Specific

Vague:
  • “Add fixation”
  • “Change duration”
  • “Make it better”
Specific:
  • “Add a 500ms fixation cross before each stimulus”
  • “Change stimulus duration from 2000ms to 1500ms”
  • “Add practice trials with feedback before the main task”
Why: AI can execute specific requests accurately. Vague requests require guessing your intent.

​
Use Psychology Task Names

AI trained on standard psychology paradigms. Good (AI knows these):
  • “Create a Stroop task”
  • “Build a 2-alternative forced choice”
  • “Make an N-Back task”
  • “Design a semantic priming experiment”
Less helpful:
  • “Make something where people press buttons for colors”
  • “Create a memory thing”
Benefit: AI knows standard parameters, timing, and structure for named tasks.

​
Provide Context

Help AI understand what you want to modify. Without context:
  • “Change duration to 1000ms”
    • AI thinks: Which component? Which duration property?
With context:
  • “For the fixation component, change duration to 1000ms”
    • AI knows exactly what to modify
For complex experiments:
  • “In the Stroop task, for the stimulus display component, change word duration to 1500ms”

​
Iterate and Refine

Don’t expect perfection on first try - refine AI output. Initial request: “Create a Stroop task” Refinements: “Make instructions more detailed” “Add example in instructions showing a congruent and incongruent trial” “Change response keys from F/J to 1/2” “Add block intermissions every 40 trials” Benefit: AI builds on previous context, understanding your intent better with each iteration.

​
Common AI Workflows

​
Workflow 1: Starting New Task

Goal: Create experiment from scratch Steps:
  1. Describe experiment
    • “Create a [task type] with [parameters]”
    • Example: “Create a Stroop task with 160 trials”
  2. Review AI’s creation
    • Check component structure
    • Verify timing
    • Review instructions
  3. Refine specific details
    • “Make instructions clearer”
    • “Change timing to [value]”
    • “Add [missing element]”
  4. Add trial variables (if multi-trial task)
    • AI may generate automatically
    • Or: “Create variables for the Stroop trials”
  5. Preview and test
    • Test with AI’s creation
    • Note any issues
    • Ask AI to fix: “The fixation is too short, increase to 1000ms”

​
Workflow 2: Modifying Existing

Goal: Improve or change current experiment Steps:
  1. Select or describe what to change
    • “Change the fixation component…”
    • “For all stimuli…”
    • “In the consent form…”
  2. Provide new parameters
    • Specific values
    • Clear direction
    • Reference previous values if helpful
  3. Review changes
    • Check that modification matches intent
    • Verify nothing else changed unexpectedly
  4. Further adjustments if needed
    • Iterate until correct
    • Build on AI’s changes

​
Workflow 3: Getting Help

Goal: Learn best practices or solve design problem Steps:
  1. Ask methodology question
    • “How many trials do I need for [task]?”
    • “What’s the standard [parameter] for [paradigm]?”
    • “Should I use [option A] or [option B]?”
  2. Receive research-backed advice
    • AI provides recommendation
    • Explanation of rationale
    • References to standards (if applicable)
  3. Apply recommendations
    • “Set it to the standard you mentioned”
    • Or manually apply advice
  4. Verify in preview
    • Test recommended parameters
    • Adjust if needed for your specific case

​
AI Features

​
Component Mentions

Reference specific components in your requests. How to mention:
  • Use component name: “For the ‘Stroop Stimulus’ component…”
  • Use component type: “For all fixation components…”
  • Use position: “For the third component in the timeline…”
Why useful:
  • Precise targeting in complex experiments
  • Avoid ambiguity when multiple similar components exist
  • Bulk operations on specific subset
Example: “For all components in Block 1, change duration to 2000ms”

​
Detailed Instructions

Provide rich context for complex requests. When to give details:
  • Creating custom experiment (not standard paradigm)
  • Specific methodological requirements
  • Unusual constraints or requirements
  • Target population considerations
Example detailed request: “Create a visual search task for children (ages 6-8):
  • Colorful, friendly stimuli (animals)
  • Longer than usual display time (no timeout)
  • Encouraging feedback after each trial
  • Simplified instructions with examples
  • Large buttons for response collection
  • Frequent breaks (every 15 trials)”
AI adapts to details:
  • Child-appropriate language
  • Larger font sizes
  • More colorful, engaging design
  • Adjusted timing for younger population

​
Methodology Guidance

AI trained on psychology research can advise on design decisions. Questions AI can answer: Trial counts:
  • “How many trials for sufficient statistical power?”
  • “Is 50 trials enough for a within-subject design?”
Timing parameters:
  • “What’s the standard fixation duration?”
  • “How long should stimulus be displayed?”
  • “What’s a good inter-trial interval?”
Design choices:
  • “Should I use blocked or randomized design?”
  • “Do I need practice trials?”
  • “How many conditions can participants handle?”
Population considerations:
  • “How should I adjust timing for older adults?”
  • “What’s appropriate for children?”
  • “Considerations for online vs. lab testing?”
AI provides:
  • Typical values from literature
  • Rationale for recommendations
  • When to deviate from standards
  • Trade-offs between options

​
Tips for Better AI Results

​
Use Clear, Direct Language

Good:
  • “Add a 500ms fixation cross before each trial”
  • “Change response keys from F/J to left/right arrows”
  • “Create 160 Stroop trials with 50% congruent and 50% incongruent”
Less effective:
  • “Maybe add something before trials”
  • “Change the keys”
  • “Make a lot of Stroop trials”
Why: Clear language → clear intent → accurate execution

​
Reference Psychology Literature When Relevant

AI knows standard paradigms and can match literature. Example: “Create a Stroop task following the parameters from MacLeod (1991)” Benefit: AI aligns with specific methodology, replicates established protocols.

​
Specify Exact Numbers

For durations, counts, sizes - be precise. Specify:
  • Durations in milliseconds: “500ms” not “half a second”
  • Trial counts: “160 trials” not “a lot of trials”
  • Font sizes: “18px” not “bigger”
  • Percentages: “50% congruent” not “half congruent”
Why: No ambiguity, AI sets exact values.

​
Preview AI Changes Before Finalizing

Always test AI-generated or modified components. Process:
  1. AI makes changes
  2. Preview experiment
  3. Note any issues
  4. Tell AI what to fix
  5. Re-preview
  6. Iterate until correct
Catch errors early:
  • Timing not quite right
  • Instructions unclear
  • Response keys confusing
  • Structure not as intended

​
Iterate - Refine AI’s Work Step by Step

Think of AI as collaborative partner, not one-shot solution. Iteration pattern: 
You: Create a Stroop task
AI: [Creates structure]

You: Make instructions more detailed
AI: [Updates instructions]

You: Add example showing congruent trial
AI: [Adds example]

You: Change timing to 1500ms
AI: [Updates timing]
Benefit: Incremental refinement leads to exactly what you want.

​
What AI Cannot Do

​
Cannot Replace Methodological Expertise

AI provides suggestions based on common practices, but:
  • Can’t design novel paradigms - AI follows established patterns
  • Can’t determine your specific research needs - You know your hypotheses best
  • Can’t guarantee validity - You must verify AI’s suggestions match your goals
  • Can’t handle highly specialized tasks - Unusual paradigms may confuse AI
You still need:
  • Domain expertise in your research area
  • Understanding of your specific hypotheses
  • Knowledge of your participant population
  • Ability to evaluate AI’s suggestions critically

​
May Suggest Configurations That Need Verification

AI’s suggestions are starting points, not final authority. Always check:
  • Do suggested trial counts provide sufficient power for your design?
  • Are timing values appropriate for your specific stimuli?
  • Do instructions match your experimental goals?
  • Is randomization scheme correct for your hypotheses?
Verify against:
  • Your research plan
  • Literature in your specific area
  • Pilot data
  • Expert consultation (advisor, collaborators)

​
Limited to Built-In Component Types

AI can only create components that exist in the system. Can create:
  • All standard psychology task components
  • Questionnaires and surveys
  • Instruction and consent screens
  • Standard cognitive paradigms
Cannot create:
  • Custom component types not in library
  • Specialized neuroscience equipment integration
  • External software connections (unless built-in)
Workaround: Combine existing components creatively or request new component types be added to platform.

​
Cannot Access External Data Without Your Input

AI doesn’t browse the web or access your local files automatically. Can’t automatically:
  • Fetch your stimulus images from your computer
  • Access your Excel file of trial parameters
  • Download stimuli from online databases
  • Connect to your external experiment software
You must:
  • Upload stimuli to media library first
  • Import variables from CSV manually
  • Provide external data through interface

​
Troubleshooting

​
If AI Doesn’t Understand

Symptom: AI responds with confusion or asks for clarification Solutions:
  1. Rephrase more specifically - Add details, remove ambiguity
  2. Break into smaller requests - Instead of one complex request, multiple simple requests
  3. Use standard terminology - Psychology terms AI knows
  4. Provide examples - “Like a Stroop task, but with colors and shapes instead of words”
Example: 
You: "Make it better"
AI: "I'm not sure what to improve. Could you specify what aspect?"

You: "Add practice trials with feedback before the main task"
AI: "Done. Added 5 practice trials with feedback before the main experiment."

​
If Changes Aren’t What You Wanted

Symptom: AI modified wrong thing or made unexpected changes Solutions:
  1. Undo changes - Use undo button or keyboard shortcut
  2. Describe the difference - “That changed the wrong component - I wanted the stimulus component, not the fixation”
  3. Be more specific in retry - Add component names, positions, or other identifiers
  4. Use Plan mode - Review before applying next time

​
If AI Makes Errors

Symptom: AI creates invalid configuration or illogical structure Solutions:
  1. Describe the issue - “The response keys you set are backwards - F should be for red, not blue”
  2. Ask for correction - “Fix the response key mappings”
  3. Provide correct values - “Set correct response to F when color is red, and J when color is blue”
  4. Verify in preview - Always test AI changes
  5. Report persistent issues - If AI consistently makes same error, report as bug

​
Next Steps

Now that you understand the AI Assistant:

Task Editor

Create experiments with AI assistance

Variables View

Use AI to generate trial variables

Preview

Test AI-created experiments

Sharing

Share your AI-built experiment
The AI Assistant accelerates experiment building by understanding natural language requests and applying research-backed best practices. Use it to create experiments faster, learn experimental design, and iterate quickly on your research ideas.