Some years ago in Tokyo, I called a taxi using my smartphone.
My location was simple: Fujimi, Chiyoda-ku, Tokyo.
After a while, the driver called me.
“I cannot find you.”
It turned out he was searching for me in another Fujimi, in Kanagawa Prefecture.
This surprised me. After all, the smartphone had GPS.
How could such a mistake occur?
The answer was simple.
Someone assumed the system “already knew” which Fujimi was intended.
The boundary condition—Tokyo versus Kanagawa—was never confirmed.
And so the system worked perfectly… in the wrong prefecture.
This small incident illustrates something deeper about how humans make decisions.
Behavioral economics, particularly Prospect Theory, shows that human beings are extremely sensitive to immediate costs and relative gains.
Small preventive actions often feel unnecessary:
- confirming assumptions
- clarifying boundaries
- documenting responsibilities
These steps appear to be extra work.
In everyday language, they simply feel like a hassle.
Because of this, people often skip them.
In the short term, skipping them feels efficient.
But Prospect Theory also shows that humans systematically underestimate low-probability but high-impact risks.
In other words:
We avoid small costs today, even when doing so creates the possibility of very large costs tomorrow.
This pattern appears everywhere.
In organizations, conflicts rarely occur at the center of operations.
The center typically shares:
- the same culture
- the same objectives
- the same leadership
But at the boundaries—between departments, between systems, between jurisdictions—assumptions begin to diverge.
Different groups interpret the same concepts differently.
Responsibilities become unclear.
Small misunderstandings accumulate.
Eventually, conflicts emerge.
In other words:
Risk concentrates at the boundary.
Business history offers many examples.
One well-known case is Target Canada.
Its expansion into Canada failed largely because of supply-chain boundary problems:
- inconsistent product master data
- incompatible inventory systems
- unclear operational responsibilities
What initially looked like a manageable operational issue eventually resulted in losses exceeding $2 billion.
The effort required to clarify these boundaries earlier would likely have been far smaller.
At the level of nations, the stakes can become even larger.
Ambiguous borders between states have historically led to escalating conflicts.
In some cases, unresolved boundaries have contributed to long-term military competition and massive defense expenditures.
Again, the lesson is the same:
When boundaries remain unclear, costs grow exponentially.
Today, we are entering a similar moment in AI governance.
Much of the current discussion focuses on:
- model capability
- data governance
- regulatory compliance
These are important.
But beneath these topics lies a quieter issue:
boundary definition.
For example:
- Where does human responsibility end and AI autonomy begin?
- Which jurisdiction governs cross-border AI services?
- Who is accountable when decisions emerge from complex human-AI interactions?
These are not merely technical questions.
They are boundary questions.
And if history is any guide, leaving them ambiguous can become very expensive.
Human organizations naturally optimize for average efficiency.
But major failures rarely occur in the middle of well-defined systems.
They occur at the edges, where assumptions diverge and responsibilities blur.
Recognizing this pattern may help us approach AI governance more realistically.
Before building increasingly sophisticated systems, we may need to spend a little more effort clarifying the boundaries within which they operate—even if doing so occasionally feels a bit of a hassle.
The small taxi incident in Tokyo offers a modest reminder.
Technology worked exactly as designed.
The error occurred because a simple boundary assumption remained unspoken.
So perhaps the first practical rule of governance—whether for organizations, digital systems, or artificial intelligence—can be stated quite simply:
Always confirm the prefecture before calling the taxi.
- Logga in för att kommentera