Overview

Project Planets is a sophisticated JavaScript-based web application designed for conducting research on decision-making, risk assessment, and learning patterns. It presents an interactive space-themed environment where participants engage with different planets to earn rewards while managing the risk of pirate attacks.

Built using the jsPsych framework and deployed with Docker, the application provides a fully configurable platform for psychological research with detailed data collection capabilities. The experiment guides participants through multiple phases with varying reward-punishment contingencies, systematically revealing how people adapt their behavior based on experience.

This project showcases my abilities in both frontend development and research technology, combining clean, maintainable code with scientifically rigorous experimental design.

Challenge

Developing a behavioral experiment platform presented several unique challenges:

• Creating a complex interactive environment that maintains scientific control while being engaging for participants
• Implementing precise timing mechanisms for stimulus presentation and response measurement
• Designing a system to handle probabilistic rewards and outcomes with carefully calibrated parameters
• Developing custom plugins to extend the core jsPsych functionality for specialized experiment requirements
• Building a secure, scalable deployment solution that can handle concurrent participants and protect research data
• Ensuring consistent behavior across different browsers and devices
• Creating a flexible architecture that allows researchers to modify parameters without needing to understand the codebase

Approach

My approach combined software engineering best practices with an understanding of behavioral research requirements:

1. Built a modular architecture using JavaScript and jsPsych as the foundation
2. Developed custom jsPsych plugins for the specialized functionality required by the experiment, including:
   • planet-response.js: A core plugin handling the main interactions with planets and ships
   • jspsych-inference-check.js: A plugin for collecting participants' inferences about relationships between stimuli
   • jspsych-valence-check.js: A plugin for collecting emotional responses to game elements
   • jspsych-html-slider-triangle.js: A custom plugin for ternary choice data collection
3. Implemented a comprehensive Docker-based deployment system allowing for:
   • Easy installation on any cloud provider
   • Secure SSL/TLS encryption for all participant data
   • Integrated data storage and experiment management through JATOS
   • Simple file management through a dedicated file browser interface
4. Created thorough documentation for both researchers and developers to ensure sustainability of the platform
5. Designed a flexible configuration system allowing researchers to modify experimental parameters without code changes

Code Samples

Custom Planet Response Plugin

// Example from the custom planet-response-command.js plugin
let planet_ship = {
  type: "planet-response-command",
  show_ship: true,
  prompt: planet_labels,
  stimulus: planet_list,
  stimulus_select: stim_selector_highlight,
  ship_stimulus: ship_list,
  reset_planet_wait: reset_planet_wait_const,
  shield_charging_time: shield_charging_time_const,
  ship_attack_time: ship_attack_time_const,
  ship_attack_effect: get_indexed_constant_array('ship_attack_effect'),
  show_whether_shield_blocked_attack_or_bonus:
    show_whether_shield_blocked_attack_or_bonus,
  block_duration: block_duration,
  probability_trade: get_indexed_constant_array('probability_trade'),
  probability_ship: get_indexed_constant_array('probability_ship'),
  probability_shield: get_indexed_constant_array('probability_shield'),
  trade_outcomes: get_indexed_constant_array('trade_outcome_set'),
  ship_outcomes: get_indexed_constant_array('ship_emergence_set'),
  shield_outcomes: get_indexed_constant_array('shield_available_set'),
  attack_images: get_indexed_constant_array('attack_img'),
  image_paths_to_preload: get_indexed_constant_array('attack_img_path'),
  attack_text_colours: get_indexed_constant_array('attack_text_colour'),
  attack_blocked_images: get_indexed_constant_array('attack_blocked_img'),
  win_100_text: win_100_text,
  data: {
    phase: "phase2",
    block_type: "planet_ship",
  },
  on_start: function(trial) {
    trial.data.points = points;
    trial.data.block_number = block_number;
    trial.data.trial_number = trial_number;
  },
  on_finish: function(data) {
    points = data.points_total;
    trial_number = data.trial_number;
    trial_number++;
    // script for continuous response block
    if (continuousResp) {
      jsPsych.endCurrentTimeline();
      block_number = data.block_number;
      block_number++;
      console.log("Block " + block_number);
    } else {
      if (trial_number >= nTrialspBlk) {
        trial_number = 0;
        block_number = data.block_number;
        block_number++;
        console.log("Block " + block_number);
      }
    }
  },
};

Docker Deployment Configuration

version: "3.8"

name: researchproject
services:
  proxy:
    image: jc21/nginx-proxy-manager:2.10.4
    restart: unless-stopped
    ports:
      - 80:80
      - 443:443
      - 81:81
    depends_on:
      jatos:
        condition: service_healthy
    volumes:
      - ./data:/data
      - ./letsencrypt:/etc/letsecrypt

  jatos:
    image: jatos/jatos:3.9.3
    restart: unless-stopped
    ports:
      - 9000:9000
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:9000/ping"]
      interval: 2s
      timeout: 2s
      retries: 1000
      start_period: 10s
    volumes:
      - ./jatos-logs:/opt/jatos/logs
      - ./jatos-data:/opt/jatos_data
      - ./jatos.conf:/opt/jatos/conf/jatos.conf:ro

  filebrowser:
    image: filebrowser/filebrowser:v2.31.1
    restart: unless-stopped
    ports:
      - 8080:8080
    depends_on:
      jatos:
        condition: service_healthy
    volumes:
      - ./jatos-data/study_assets_root:/srv
      - ./filebrowser/database.db:/database.db
      - ./filebrowser/.filebrowser.json:/.filebrowser.json
    environment:
      - PUID=1000
      - GUID=1000

Demo

The Project Planets experiment follows a structured flow designed to measure how people learn from experience and adjust their behavior:

Experiment Flow

1. Phase 1: Participants interact with planets to earn rewards, learning the baseline reward probabilities for each planet
2. Assessment: Participants complete valence and inference checks to measure their current understanding
3. Phase 2: Ships are introduced that can attack and reduce points, creating a risk-reward tradeoff
4. Assessment: Further valence and inference checks to measure updated understanding
5. Phase 3: Participants are explicitly told the contingencies between planets and ships
6. Final Phase: Behavior is measured to see how explicit knowledge affects decision-making

Outcomes

Technical Achievements

• Successfully deployed as a scalable, containerized application usable by researchers worldwide
• Created reusable components that have been adopted in other psychological research projects
• Developed comprehensive documentation enabling other researchers to modify and extend the platform
• Built a system capable of running concurrent experiments with different parameters
• Implemented secure data collection and storage solutions compliant with research ethics requirements

Research Impact

Project Planets has been used to study how people learn from experience and make decisions under risk. The platform has enabled researchers to systematically investigate:

• How experience shapes risk perception and decision-making
• The relationship between explicit knowledge and behavior
• Individual differences in learning patterns and risk tolerance
• The effects of different reward schedules on behavior

Skills Demonstrated

• Advanced JavaScript development
• Custom plugin development for jsPsych
• Docker containerization and deployment
• UI/UX design for research applications
• Data collection and management systems
• Secure web application development
• Technical documentation for diverse audiences