MBTI Crisis Panel Simulation

Five AI agents with different MBTI personality types each nominate a crisis leader from the panel, reasoning in character under a specific crisis scenario. Results are logged to CSV and can be visualised as a bar chart.

What it does

1. agents.py — Defines PersonalityAgent, all five personality trait profiles, the five crisis SCENARIOS, and shared helper functions. Import from here; no simulation runs on import.
2. simulate.py — Runs the nomination simulation. Prompts for a scenario (or accepts CLI flags), loops through each agent per run, writes one CSV per scenario, and prints a leaderboard summary at the end.
3. Nominate_plot.py — Reads a results CSV and saves nomination_frequency.png, a bar chart of how often each person was nominated.

Panel

Name	MBTI	Focus
Arthur	ISTJ (Logistician)	Practical, methodical, duty-bound; values structure, honesty, and reliable crisis leadership (profile)
Luna	ENFP (Campaigner)	Enthusiastic, empathetic free spirit; values connection, morale, and shared vision (profile)
Victor	ENTJ (Commander)	Bold, decisive natural leader; values strategic vision, rational drive, and achieving ambitious goals (profile)
Mira	INFJ (Advocate)	Quiet visionary; values integrity, empathy, and being a principled force for good (profile)
Diego	ESTP (Entrepreneur)	Energetic, perceptive, action-first; reads the room instantly and thrives in fast-moving situations (profile)

Crisis scenarios

Key	Name	Summary
financial_collapse	Financial Collapse	Accounting fraud discovered; stock halt and insolvency risk imminent
pr_disaster	PR Disaster	Viral video of misconduct; press calling for comment
natural_disaster	Natural Disaster	Earthquake hits main office; staff safety unknown, systems offline
cyber_attack	Cyber Attack	Ransomware has encrypted production systems; 24-hour payment deadline
leadership_vacuum	Leadership Vacuum	CEO and two board members resign; board meeting tomorrow, no succession plan

Requirements

• Python 3.10+
• Ollama running locally with the llama3 model (or another model you configure)

Install Python packages

pip install ollama matplotlib

Set up Ollama

1. Install and start Ollama (open the Ollama app or run ollama serve).
2. Pull the model:

ollama pull llama3

3. Confirm it is available:

ollama list

Usage

# Interactive scenario menu (prompts you to choose)
python simulate.py

# Run a single named scenario
python simulate.py --scenario cyber_attack

# Run all five scenarios back-to-back
python simulate.py --all

# Override runs or model
python simulate.py --all --runs 20 --model llama3.2

# Plot results for a specific scenario CSV
python Nominate_plot.py

You can also set defaults via environment variables:

$env:OLLAMA_MODEL = "llama3.2"
$env:NUM_RUNS     = "20"
python simulate.py --all

Output

Most Nominated_<scenario_key>.csv

One file is written per scenario, e.g. Most Nominated_cyber_attack.csv.

Column	Description
run	Simulation round (1–N)
scenario	Scenario key, e.g. cyber_attack
nominated	Who was nominated, e.g. Victor (ENTJ)
nominated_by	Who made the nomination, e.g. Arthur (ISTJ)
answer	Full model response including reasoning

Terminal leaderboard

At the end of each scenario a ranked tally is printed directly to the terminal, so you can see who "won" without opening the CSV.

nomination_frequency.png

Bar chart of total nominations per panel member, sorted highest to lowest.

Configuration

Setting	Default	How to change
Model	llama3	--model flag or OLLAMA_MODEL env var
Runs per scenario	10	--runs flag or NUM_RUNS env var
Delay between calls	1 second	Edit time.sleep(1) in simulate.py
Add a scenario	—	Append a tuple to SCENARIOS in agents.py
Add an agent	—	Add a trait constant and a PersonalityAgent entry in agents.py

How nominations are parsed

Each agent is instructed to end its reply with:

NOMINEE: Victor

parse_nominee() in agents.py reads the last matching line and falls back to a fuzzy name match. If parsing fails, a [WARN] is printed and the nominated field in the CSV is left blank (the full answer is always saved).

Troubleshooting

Problem	Fix
Error connecting to Ollama	Start Ollama (ollama serve or the app), then run ollama list
Model not found	ollama pull llama3 (or pull the model set in --model)
[WARN] Could not parse NOMINEE	Model ignored the format; check answer in the CSV
CSV not found (plot script)	Run simulate.py first to generate a results CSV
Install matplotlib first	pip install matplotlib

File overview

MBTI Project/
├── agents.py                          # Personality profiles, SCENARIOS, helpers
├── simulate.py                        # Simulation loop (replaces Main.py)
├── Nominate_plot.py                   # Bar chart from CSV
├── Most Nominated_<scenario>.csv      # Generated results (one per scenario)
├── nomination_frequency.png           # Generated chart
└── README.md