Documentation / Tutorials
Tutorials
Hands-on walkthroughs for common simulation patterns.
Tutorial 1: Marketplace Dynamics
Simulate a simple marketplace where agents buy, review, and recommend products. Observe how word-of-mouth dynamics emerge from individual purchasing decisions.
Step 1: Create the scenario
Go to Simulations > New Scenario and configure:
- 10 Influencers (high extraversion, frequent posting)
- 15 Followers (high agreeableness, follow trends)
- 5 Analysts (low extraversion, detailed reviews)
Set the platform type to Marketplace and add 5 seed products with varying quality ratings.
Step 2: Run the simulation
Set tick count to 200 and start the run. Watch the action feed for emerging patterns:
- Influencers discover products early and post reviews
- Followers amplify positive reviews through likes and reposts
- Analysts provide detailed assessments that shift opinion over time
Step 3: Analyse results
In the Results tab, examine:
- Which products gained the most purchases and why
- Whether Influencer endorsements correlated with product quality
- How Analyst reviews shifted buying patterns in later ticks
Try varying the Influencer-to-Analyst ratio. A scenario with more Analysts typically shows slower but more accurate market convergence.
Tutorial 2: Information Spread
Measure how Contrarian agents affect the propagation of news through a social network. This tutorial explores the tension between viral spread and critical scrutiny.
Step 1: Set up two runs
Create two scenarios with identical agent counts but different compositions:
Run A: No Contrarians
- 20 Influencers
- 25 Followers
- 5 Connectors
Run B: With Contrarians
- 15 Influencers
- 20 Followers
- 5 Connectors
- 10 Contrarians
Step 2: Seed a news event
Use seed_content to inject a single high-engagement post at tick 0 in both scenarios. Set 300 ticks and run both.
Step 3: Compare propagation
Use the Compare view to overlay the two runs. Key metrics to compare:
- Reach speed -- How many ticks until 80% of agents have seen the news
- Engagement depth -- Reply chains vs. simple likes
- Sentiment drift -- Whether Contrarian pushback shifts overall sentiment
- Network clustering -- Do opinion clusters form around the Contrarians?
Typical finding: Contrarians slow spread by 15-25% but increase reply depth by 2-3x, creating richer discussion threads.
Tutorial 3: Community Health
Monitor engagement patterns over a long simulation to understand how community health evolves. This tutorial introduces Troll agents and the report/moderation mechanic.
Step 1: Design the community
Configure 40 agents:
- 8 Influencers -- content creators
- 15 Followers -- the audience
- 5 Analysts -- quality contributors
- 7 Connectors -- bridge builders
- 5 Trolls -- disruptive agents
Step 2: Run for 500 ticks
This is a longer run designed to show slow dynamics. Key things to watch during the live run:
- When do the first report actions appear?
- Do Trolls cluster in certain threads or spread out?
- Does overall posting frequency decline as Troll activity increases?
Step 3: Measure community health
In the Results tab, focus on these health indicators:
| Metric | What it tells you |
|---|---|
| engagement_rate | Ratio of interactions to posts -- declining means agents are disengaging |
| report_ratio | Reports per 100 actions -- high values indicate community stress |
| new_follows | Follow actions per tick -- declining means agents are withdrawing |
| sentiment_mean | Average post sentiment -- tracks overall community mood |
Experiment: Run the same scenario with 0, 3, and 5 Trolls. Plot engagement_rate over time. Most communities show a tipping point around 10-15% Troll density where engagement drops sharply.