Skip to main content

Network Reciprocity

GitHub code
Network reciprocity is cooperation sustained by population structure. When interactions are restricted to a local neighborhood rather than the whole population, cooperators can form clusters that shield each other from exploitation by defectors — the network does the work that memory or reputation does in reciprocity models.

How It Is Implemented Here​

The positive routing kernel is uniform over local grid neighbors with no additional bias. The mechanism relies entirely on the spatial grid structure: the von Neumann or Moore neighborhood restricts who receives cooperative benefit and who competes for reproduction. Cooperator clusters that form through local reproduction maintain higher average fitness within the cluster than the surrounding population, allowing the cluster to expand.

There is no memory, reputation, or lineage bias — the only structural ingredient is the local neighborhood.

Key Parameters​

ParameterDefaultRole
positive_kernel_mode"uniform"Uniform routing over local neighbors
initial_identity_count1Number of distinct initial lineage identities
B_plus_scale1.0Scales cooperative benefit produced per unit trait
C_scale0.2Private cost per unit trait
Display 1: Key parameters controlling spatial routing and payoff scaling in the network reciprocity model.

Python Module Layout​

moran_models/nowak_mechanisms/network_reciprocity/
  __init__.py
  network_reciprocity_model.py
  network_reciprocity_pygame_ui.py
  config/
    network_reciprocity_config.py

Usage​

./.conda/bin/python -m moran_models.nowak_mechanisms.network_reciprocity.network_reciprocity_model

Live viewer:

./.conda/bin/python -m moran_models.nowak_mechanisms.network_reciprocity.network_reciprocity_pygame_ui

See Also​

  • Spatial Altruism — a concrete historical model, Mitteldorf & Wilson (2000), that is mechanically the closest real-world relative of this page: a lattice, purely local interaction, no memory or reputation. That page's own "See Also" section goes further and argues the paper's authors actually frame their result as a Group Selection argument (patches of altruists sustaining higher density than patches of non-altruists), not a network-reciprocity one — worth reading if the mechanical similarity here seems too tidy.

References​