人間関係の偏りには“避けられないノイズ”があるのかも ── 262法則をPythonで検証してみた

import numpy as np
import matplotlib.pyplot as plt

# =====================================
# Parameters　（パラメータ設定）
# =====================================
np.random.seed(0)     # <-- 乱数シードを固定（再現性の確保）
N = 10                # 人数
trials = 1000         # モンテカルロ試行回数

p_like = 0.2
p_dislike = 0.2
n_like = int(N * p_like)
n_dislike = int(N * p_dislike)

max_scores = []
min_scores = []

# すべての試行でのスコアを保存
all_scores = []

for _ in range(trials):
    like  = np.zeros(N)
    dislike = np.zeros(N)

    for voter in range(N):
        others = [i for i in range(N) if i != voter]

        liked   = np.random.choice(others, size=n_like, replace=False)
        remaining = [i for i in others if i not in liked]
        disliked = np.random.choice(remaining, size=n_dislike, replace=False)

        like[liked] += 1
        dislike[disliked] += 1

    score = like - dislike  # net evaluation = like − dislike（好意−嫌悪）

    max_scores.append(score.max())
    min_scores.append(score.min())
    all_scores.append(score.copy())

print(f"=== {N}人の集団を{trials}回シミュレーションした場合の人間関係スコア分布の要約 ===")
print(f"最も好かれた人の平均スコア: {np.mean(max_scores):.2f}")
print(f"最も嫌われた人の平均スコア: {np.mean(min_scores):.2f}")
print(f"好かれ・嫌われの幅（平均）: {np.mean(np.array(max_scores) - np.array(min_scores)):.2f}")

import numpy as np
import networkx as nx
import matplotlib.pyplot as plt

# =====================================
# Graph visualization for ONE example trial
# （1試行分のネットワーク可視化）
# =====================================
N = 10                # 人数
np.random.seed(0)     # <-- 乱数シードを固定（再現性の確保）
like  = np.zeros(N)
dislike = np.zeros(N)
edges = []

p_like = 0.2
p_dislike = 0.2
n_like = int(N * p_like)
n_dislike = int(N * p_dislike)

for voter in range(N):
    others = [i for i in range(N) if i != voter]
    liked   = np.random.choice(others, size=n_like, replace=False)
    remaining = [i for i in others if i not in liked]
    disliked = np.random.choice(remaining, size=n_dislike, replace=False)

    like[liked] += 1
    dislike[disliked] += 1

    # 好意 → 赤
    for l in liked:
        edges.append((voter, l, "red"))

    # 嫌悪 → 青
    for d in disliked:
        edges.append((voter, d, "blue"))

score = like - dislike

G = nx.DiGraph()

for i in range(N):
    G.add_node(
        i,
        label=f"score\n{int(score[i])}"
    )

for (a,b,c) in edges:
    G.add_edge(a,b,color=c)

pos = nx.spring_layout(G, seed=0)
plt.figure(figsize=(12,10))

# ノード（人）を描画
nx.draw_networkx_nodes(G, pos, node_size=600, node_color="lightgray")

# エッジ（関係）を矢印つきで描画
for (u,v,c) in edges:
    nx.draw_networkx_edges(
        G,
        pos,
        edgelist=[(u,v)],
        edge_color=c,
        arrows=True,
        arrowstyle='-|>',
        arrowsize=14,
        width=1.2,
        alpha=0.7,
        connectionstyle="arc3,rad=0.2"
    )

# 各ノードに score を表示
node_labels = nx.get_node_attributes(G,'label')
nx.draw_networkx_labels(G,pos,labels=node_labels,font_size=7)

plt.title("Graph representation of preferences\n(red = like, blue = dislike)")
plt.axis('off')
plt.show()