DBA Thesis Code — Startup Coopetition (Threshold Hawk-Dove Game)

Agent-based model (NetLogo) and analysis scripts (Python) for a DBA thesis on the emergence of innovation ecosystems. A population of startups chooses between cooperating ("dove") and competing ("hawk") in groups of size N, following an N-person Hawk-Dove game with an engagement threshold (HDG-T), after Chen et al. (2017) [full reference to be added].

Contents

File	Purpose
hawk-dove-model.nlogox	NetLogo 7.0.4 model: analytical computation of interior equilibria (reproduction of Figure 5 in Chen et al.) and imitation-dynamics simulation
analysis.ipynb	Analysis of BehaviorSpace outputs (P1 to P5): regime classification, tables and figures used in the manuscript
resultats/	BehaviorSpace CSV inputs (P1.csv … P5.csv) and figure outputs (p1.png … p5.png)

Requirements

• NetLogo 7.0.4
• Python 3.12 with pandas, matplotlib and Jupyter

pip install pandas matplotlib jupyter

The model

• Fixed population of Z = 100 agents, each a "dove" (cooperator) or a "hawk" (competitor).
• Payoffs are computed for groups of N agents drawn from the population (hypergeometric sampling). If the share of doves reaches the threshold T, doves exclude hawks and share the resource R (cost c_d per excluded hawk); otherwise hawks fight over R (cost c_h per rival).
• At each tick, every agent compares its payoff with that of a randomly drawn agent and imitates it with a Fermi probability (β = 1). Updates are synchronous.
• The simulation stops at fixation (all doves or all hawks) or after 2000 ticks. There is no mutation.

Reference parameters: Z = 100, N = 5, c_h = c_d = 0.8, T = 0.4, R = 1, x₀ = 0.5.

Reproducing the results

1. Open hawk-dove-model.nlogox in NetLogo.
2. Run the BehaviorSpace experiments below (50 replications each, 2000-tick horizon) and export the output in table format to resultats/.
3. Run manuscript_version.ipynb.

Analysis	BehaviorSpace experiment	Varied parameter(s)	File
P1	P1 (variation de c-h et c-d)	c_h, c_d ∈ [0, 1], step 0.05	P1.csv
P2	P2 (variation de R)	R ∈ [0.1, 2], step 0.1	P2.csv
P3	P3 (variation de T)	T ∈ [0.1, 1], step 0.1	P3.csv
P4	P4 (variation de init)	x₀ ∈ [0.05, 0.95], step 0.05	P4.csv
P5	P5 (variation de N)	N ∈ [3, 15]	P5.csv

No random seed is set, so results vary slightly between runs (sampling noise with 50 replications).

Regime definitions

Each run is classified as follows (threshold = 500 ticks, i.e. ¼ of the horizon):

• Coexistence: cooperators and competitors coexist for at least 500 ticks;
• Cooperation: fixation on cooperation before 500 ticks;
• Competition: fixation on competition before 500 ticks.

For P1, each (c_h, c_d) cell takes the dominant regime across its 50 replications.

Author

Sébastien — DBA, IAE de Nice / Inria Startup Studio.