The Thorn in the Interaction
KJ法という思考の整理
本章では、物語本文と設問を通して、問題解決の初期段階に焦点を当てます。
桜子と西野がまず行ったのは、KJ法による整理でした。
KJ法は結論を出すための手法ではありません。
観察された事実や気づきを言語化し、関連性を可視化するための方法です。
湿度、納入時期、取扱い、ロット内分散。
それぞれの断片を並べても、すぐに答えは現れません。
しかし、整理された情報の中から「違和感」が浮かび上がることがあります。
本章では、
•KJ法がどのように仮説形成の土台となったのか
•どの観察が次の分析へとつながったのか
に注目しながら読み進めてください。
ここで行われているのは断定ではなく、
思考の構造化です。
After reorganizing the post-it notes using the KJ method, Sakurako stepped back from the whiteboard.
Dozens of notes—each distilled from careful observation on the factory floor—were clustered into tentative themes.
Humidity fluctuation.
Delivery timing.
Handling variation.
Within-lot variance.
Yet the clusters did not converge.
“They don’t line up,” Nishino murmured. “The t-test says storage alone isn’t significant. And yet… something is clearly unstable.”
Sakurako traced her finger across the cluster labeled:
‘Variance within the same lot.’
“If storage were the dominant cause,” she said slowly, “we would expect consistent degradation across all lots under high humidity.”
She paused.
“But this note… this suggests variability even within identical material. That shouldn’t happen if humidity is acting alone.”
Nishino frowned.
“So… you’re saying humidity might not be acting alone?”
“I’m not sure,” she replied. “But what if the issue isn’t ‘good versus bad storage’? What if certain materials respond differently depending on when they were delivered?”
There was a silence—not of certainty, but of alignment.
“An interaction effect?” Nishino asked carefully.
“Possibly. Let’s not assume. Let’s test it.”
⸻
Nishino turned to the PC.
First, he categorized humidity into operational bands.
# Categorize humidity into levels
test_data$humidity_level <- cut(test_data$humidity,
breaks=c(0,50,70,100),
labels=c("Low","Medium","High"))
# Interaction plot
interaction.plot(test_data$humidity_level,
test_data$delivery_period,
test_data$defect_rate,
type="b",
col=c("blue","red"),
pch=c(1,19),
main="Humidity × Delivery Period Interaction")
The Enter key broke the quiet hum of the fluorescent lights.
The graph appeared.
Two lines.
Not parallel.
They crossed.
For most delivery periods, defect rates rose gradually with humidity.
But for one specific delivery period, defect rates spiked sharply once humidity entered the “High” band.
It was not impossible.
But it was structurally inconsistent with stable material behavior.
Nishino leaned closer.
“This isn’t a general storage issue. Most lots tolerate humidity within range. But this period… something shifts.”
A sharp bend formed in the graph—a thorn protruding from order.
Sakurako did not feel triumph.
She felt unease.
“If this is real,” she said quietly, “then the issue isn’t discipline. It’s specificity.”
She glanced back at the neatly organized KJ clusters.
They had not given an answer.
They had revealed a question.
“Get me the shipping slips from that delivery period.”
Reading Comprehension Questions
The Thorn in the Interaction
⸻
Question 1
What was the primary role of the KJ method in this chapter?
(A) To statistically prove the root cause of the defects
(B) To visually confirm the interaction effect
(C) To organize observations and clarify emerging questions
(D) To replace the need for further data analysis
⸻
Question 2
Why did Sakurako begin to suspect an interaction effect?
(A) Because humidity alone showed a strong statistical significance
(B) Because defect rates increased uniformly across all delivery periods
(C) Because variance appeared even within the same lot under certain conditions
(D) Because Yamashita instructed her to examine delivery timing
⸻
Question 3
What does the crossing of lines in the interaction plot most strongly suggest?
(A) There is no relationship between humidity and defect rate
(B) The effect of humidity depends on delivery period
(C) Storage conditions have no measurable impact
(D) The data was incorrectly categorized
⸻
Question 4
How does Sakurako respond emotionally to the discovery of the interaction?
(A) She celebrates the confirmation of her hypothesis
(B) She immediately accuses Yamashita of misconduct
(C) She experiences unease rather than triumph
(D) She abandons the investigation
⸻
Question 5
What does the phrase “the issue isn’t discipline. It’s specificity.” most likely mean?
(A) The problem is caused by workers ignoring procedures
(B) The defect arises only under certain combinations of conditions
(C) Environmental management should be eliminated
(D) Delivery timing is irrelevant
効果の否定と交互作用の発見
前章では、単一要因（主効果）としての保管条件が否定されました。
しかし本章で浮かび上がったのは、交互作用という視点です。
交互作用とは、
「ある要因の効果が、別の要因の水準によって変わる」
という現象です。
湿度だけでは説明できない。
納入時期だけでも説明できない。
しかし、その組み合わせによって挙動が変化する。
これは現場で頻繁に見られる構造です。
問題は、単純ではない。
単一要因の改善だけでは解決しない場合があります。
桜子が感じたのは確信ではありません。
交差する線が示した、構造の歪みでした。
もしこの章に興味を持ったなら、
二元配置分散分析や交互作用効果について調べてみるのも良いでしょう。
主効果だけでは見えない世界が、そこにあります。
