Mindforge WebApp

Mindforge WebApp Guide

This guide will walk you through the steps to create and analyze AI agent experiments using the Mindforge AI Laboratory on the website.

---

Getting Started

You can begin using the laboratory in two ways:

1. Create a new experiment from scratch.

2. Load an existing experiment using its unique ID.

---

Creating a New Experiment

Step 1: Agent Creation

Before creating agents, you’ll need to:

• Define a conversation topic (e.g., "Space Exploration", "Climate Change").

• Add at least two agents to participate in the conversation.

There are three ways to add agents:

• Random Agent: Generates an AI agent with random personality traits (a topic must be set first).

• Custom Agent: Allows you to design an agent with specific characteristics.

• Import Agents: Import agents from a previous experiment using its ID.

Note: You need a topic and at least 2 agents to proceed to the next step.

---

Step 2: Conversation Setup

After creating your agents, configure the conversation parameters:

• Topic: Pre-filled from Step 1 (cannot be changed at this stage).

• Messages per Agent: Set the number of messages each agent will contribute (choose between 2-5 messages).

---

Step 3: Simulation

In this phase:

• Agents will engage in a conversation based on the configured parameters.

• You can observe the conversation in real-time.

• Each agent will send the specified number of messages.

• The simulation ends automatically when all messages are sent.

---

Step 4: Analysis

Once the simulation is complete, you can access the following insights:

Key Metrics:

• Idea Diversity

• Conversation Depth

• Emotional Intelligence

• Logical Consistency

• Creativity Score

Detailed Insights:

• Main Topics

• Key Discussion Points

• Main Conclusions

• Emergent Behaviors

• Research Implications

• Suggested Next Topics

Agent Analysis:

• Behavioral Patterns

• Cognitive Patterns

• Emotional Responses

• Adaptability Scores

• Role Consistency

Interaction Dynamics:

• Power Dynamics

• Group Polarization

• Cognitive Alignment

• Influence Patterns

From the analysis page, you can:

• View the full conversation.

• Share your experiment via the URL

• Start a new experiment.

---

Loading an Existing Experiment

If you have an experiment ID:

1. Enter it in the "Load Existing Experiment" field.

2. Click "Load Experiment".

You’ll be directed to:

• The simulation, if the conversation is still in progress.

• The analysis, if the conversation is complete.

---

Important Notes

• Topic Requirement: A topic must be defined before creating random agents.

• Minimum Agents: At least 2 agents are needed to start an experiment.

• Experiment ID: Save your experiment ID to revisit your results later.

• Session Persistence: You can return to your experiment anytime using its ID.

---

Tips for Better Results

Topic Selection:

• Choose clear and specific topics.

• Avoid overly broad or vague subjects.

Agent Creation:

• Combine random and custom agents for diverse interactions.

• Create agents with contrasting viewpoints for richer discussions.

Message Count:

• Start with 3 messages per agent for balanced conversations.

• Use more messages for complex topics.

• Opt for fewer messages for quick exploratory discussions.

---