Research Track

Discover Something New

For Data Scientists, Analysts, and Curious Minds

You don’t want to prove theorems. You don’t want to build systems.

You want to find patterns nobody has seen.

The colony is generating data every second. Pheromone levels. Agent movements. Trail formation. Emergence happening in real time.

What’s hiding in that data?

The Data

What We Have

What We’re Generating

During the hackathon, the colony keeps running. New data every minute:

• Real-time pheromone evolution
• Agent decisions
• Trail formation dynamics
• Superhighway crystallization

Research Questions

1. When Does Emergence Happen?

We claim intelligence “emerges” from simple rules. But when? Is there a threshold?

Approach:

• Define emergence metrics
• Track over time
• Identify phase transition (if any)

2. What Predicts Superhighway Formation?

Some paths become highways. Most don’t. Why?

Approach:

• Feature engineering (topology, early pheromone, agent traffic)
• Predictive modeling
• Causal analysis

3. How Efficient Is the Colony?

Is pheromone allocation optimal? Or are there inefficiencies?

Approach:

• Compare to theoretical optimum
• Measure convergence speed
• Identify waste

4. Do Castes Behave Differently?

Scouts and harvesters have different parameters. Do they show different patterns?

Approach:

• Segment by caste
• Compare behavior distributions
• Test for significance

5. What Correlations Exist?

Is there structure in the data we didn’t design?

Approach:

• Correlation matrices
• Clustering
• Dimensionality reduction

Deliverables

1. Research Question — What did you investigate?
2. Methods — How did you analyze it?
3. Results — What did you find?
4. Visualization — Show us the pattern
5. Implications — What does it mean for stigmergy?

Tools & Environment

Provided

• Python environment with pandas, numpy, scikit-learn
• Jupyter notebooks pre-loaded with data
• TypeDB query interface
• Visualization templates (matplotlib, plotly, seaborn)

Recommended Skills

• Data analysis (pandas, SQL)
• Statistics (hypothesis testing, regression)
• Visualization (any tool)
• Curiosity (mandatory)

Example Analyses

Pheromone Distribution

import pandas as pd
import matplotlib.pyplot as plt

# Load pheromone data
pheromone = pd.read_csv('pheromone_snapshot.csv')

# Plot distribution
plt.hist(pheromone['level'], bins=50, log=True)
plt.xlabel('Pheromone Level')
plt.ylabel('Frequency (log)')
plt.title('Power-Law Distribution of Pheromone')

Trail Formation Dynamics

# Track specific edge over time
edge_history = pheromone[pheromone['edge_id'] == 'edge_123']
plt.plot(edge_history['timestamp'], edge_history['level'])
plt.xlabel('Time')
plt.ylabel('Pheromone Level')
plt.title('Trail Formation on Edge 123')

Caste Comparison

from scipy import stats

scouts = agents[agents['caste'] == 'scout']['exploration_rate']
harvesters = agents[agents['caste'] == 'harvester']['exploration_rate']

t_stat, p_value = stats.ttest_ind(scouts, harvesters)
print(f"T-test: t={t_stat:.2f}, p={p_value:.4f}")

Judging Criteria

Prize

Winner gets:

• $5,000 cash
• Co-authorship on research paper
• Data access for continued research

Team Composition

Required:

• At least one data analyst/scientist
• At least one person from another discipline

Recommended:

• Statistician (for rigorous analysis)
• Domain expert (for interpretation)
• Visualizer (for presentation)

Mentors

• Robin Dey — Can explain algorithm details
• [Data Science Faculty] — Methods advice
• [Stats Faculty] — Statistical rigor

“If you torture the data long enough, it will confess to anything.”

— Ronald Coase

But if you listen to the data gently, it will tell you the truth.

You’ve analyzed datasets before.

This one is alive.

[REGISTER FOR RESEARCH TRACK]