Skip to main content

Altruism

Altruism is a form of social behavior where an individual incurs a cost to benefit another. It appears in biological systems (e.g., kin selection, reciprocal altruism) and in human societies through norms, institutions, and moral frameworks.

  • Biological roots: kin selection (Hamilton’s rule), reciprocal altruism (Trivers), group-beneficial norms.
  • Mechanisms in multi-agent systems: reward shaping for prosocial outcomes, shared value functions, communication and reputation.
  • Trade-offs: vulnerability to free-riding, need for enforcement or reputation systems.

Future pages will include examples, simple models, and experiments connecting altruistic behavior to cooperation and coordination in multi-agent environments.

Reference implementation​

This page summarizes and links to a working, vectorized Python implementation of an evolutionary altruism model from the SocialBehavior repository:

Model at a glance​

  • Patch-based grid world: each cell is empty, selfish, or altruist
  • Dynamics: cost of altruism, benefit to neighbors, fitness, and generational update via a genetic lottery
  • Performance: fully vectorized NumPy implementation
  • Visualization: Pygame UI for live interaction, Matplotlib for population plots

Quick start​

Run the command-line simulation:

python altruism_model.py --steps 200 --width 101 --height 101 --seed 42

Launch the interactive UI:

python altruism_pygame_ui.py

Controls:

  • SPACE: Start/Stop
  • R: Reset
  • S: Step
  • P: Plot population history

Import as a module:

from altruism_model import AltruismModel, Params
model = AltruismModel(Params(width=51, height=51))

Requirements​

  • Python 3.8+
  • numpy
  • pygame (UI)
  • matplotlib (plots)

Install dependencies:

pip install numpy pygame matplotlib

System dependency for Pygame visualization (example):

conda install -y -c conda-forge gcc=14.2.0

Model parameters (examples)​

  • altruistic_probability: initial chance a patch is altruist
  • selfish_probability: initial chance a patch is selfish
  • benefit_from_altruism: benefit received from altruists
  • cost_of_altruism: cost paid by altruists
  • disease, harshness: optional environment effects

Conceptual summary​

The model simulates a genetic lottery in plus-shaped neighborhoods. Each patch contributes a weighted “seed” to determine which trait (altruist vs. selfish) occupies a cell next generation. Under non-harsh conditions, selfishness tends to dominate; under harsher conditions, altruism can persist or prevail.

Learn more / Source files​

References in the upstream README include the original NetLogo model and BEAGLE/EACH curriculum materials.

Snippet: Parameters and model setup​

Below is a minimal excerpt illustrating how parameters and the model class are defined in the SocialBehavior implementation. For the full, up-to-date source, see the repository links above.

from dataclasses import dataclass
import numpy as np

BLACK, GREEN, PINK = 0, 1, 2  # empty, selfish, altruist

@dataclass
class Params:
	width: int = 51
	height: int = 51
	torus: bool = True
	altruistic_probability: float = 0.33
	selfish_probability: float = 0.33
	benefit_from_altruism: float = 0.5
	cost_of_altruism: float = 0.2
	disease: float = 0.0
	harshness: float = 0.0
	seed: int | None = None

class AltruismModel:
	def __init__(self, p: Params):
		self.p = p
		if self.p.seed is not None:
			np.random.seed(self.p.seed)
		# initialize state arrays and set up the world...

Attribution: Code adapted from SocialBehavior/altruism (altruism_model.py). Please refer to the repository for the authoritative version.

Try locally​