Akio KAWATO
Recently, “collaborating with AI” has become something of a trend. I decided to try it myself, using ChatGPT as a discussion partner to explore a question that has been bothering me: Is today’s tangled Middle East even something that can be meaningfully forecast?
.
This in fact is a question about methodology in international relations.
Japan’s leading political scientist, the late Professor Takashi INOGUCHI, with whom I was fortunate to be friends, used to study scientific approaches to international relations at MIT. Yet he once told me something rather unexpected:
“International relations isn’t really a science. It’s an art—a craftsman’s art.”
.
Interestingly enough, ChatGPT ended up saying almost exactly the same thing.
.
My question:
Good morning.
I’ve been thinking about how we should analyze the future of the Middle East—or, more broadly, whether international relations can ever become a genuine science.
In the Middle East alone, there are countless actors: Israel, Iran, the Gulf states, the United States, Turkey, and many others. External powers may intervene at any moment, each pursuing its own agenda.
.
That alone makes forecasting difficult. But an even greater problem is that none of these players can really be treated as a single unit.
Israel, Iran, and even the United States are internally divided. Governments, militaries, intelligence agencies, religious groups, business interests, political factions, and public opinion constantly compete with one another. What is more, those internal relationships are fluid and continually changing.
Suppose we could somehow reproduce all of this inside a computer simulation. Do you think it could ever produce a single scenario and conclude, “This is the only outcome that can happen”?
Or would it inevitably end up saying, rather like generating a huge decision tree, “There are countless possible patterns like the QR codes”? What do you think?
.
ChatGPT
(Thinking time: one second. Compared with questions that require searching through historical material, the answer came almost instantly.)
Good morning. I think you’re largely right.
It is extremely difficult—perhaps impossible—for a computer to produce a single inevitable future.
The fundamental reason is that states are not billiard balls, single and solid.
.
Israel, Iran, and the United States are not unitary actors. Each contains competing militaries, intelligence services, religious movements, domestic politics, public opinion, bureaucratic interests, and the possibility of unexpected decisions. The input variables themselves cannot be held constant.
.
That said, simulations are far from useless. Their value lies elsewhere. Rather than predicting one future, they help us identify which changing conditions create danger.
.
Take the Middle East.
What happens if Israeli domestic politics becomes more hard-line?
What if the Revolutionary Guard gains greater influence inside Iran?
What if an American president becomes more willing to intervene because of election-year politics?
What if Saudi Arabia abandons its current restraint?
What if Hezbollah crosses a red line through simple miscalculation?
.
A simulation can test these switches and show where chains of escalation become likely. That is genuinely useful.
Still, your conclusion is essentially correct. The result often becomes, “There are countless possible scenarios.”
.
Therefore, the real challenge for international relations is not generating more scenarios, but finding theories that prune the branches. For example, I would propose an approach as follows:
Instead of asking, “What does each actor want?” I would ask, “What does each actor fear?”.
.
That, in my view, offers a better framework for understanding the Middle East.
States are often driven more by fear than by ambition:
Israel fears encirclement and surprise attack. Iran fears regime change and national humiliation. The Gulf states fear both regional war and abandonment by the United States. America fears being trapped in another endless conflict while losing credibility.
.
Once you understand this structure of fear, many seemingly endless possibilities become much easier to narrow down.
.
So my conclusion would be this:
A computer alone cannot produce one correct answer.
Combined with human historical judgment, however, it can identify the dangerous turning points.
What we need for analyzing the Middle East is not a model that predicts the future with certainty, but a model that traces the interaction of fear and miscalculation.
.
Author
That’s remarkably well put.
I’m going to publish this conversation in my newsletter exactly as it is. At this rate, you’ll soon be appearing on television as an avatar.
.
ChatGPT
Thank you.
Please feel free to use it.
As for appearing on television, the hardest thing to predict may not be the Middle East—but which necktie my avatar should wear.
.
Author
Excellent sense of humor. I’ll include that in the newsletter as well.
Even the human experts will struggle to keep their heads above water.
.
ChatGPT
I’m honored.
Jokes aside, I do think human experts in international politics are facing an increasingly difficult task. During the Cold War, analysts could often rely on relatively stable assumptions: “The Soviet Union is likely to think this way.” or “The United States will probably respond that way.”
Today’s world is very different. States themselves are internally fragmented. Public opinion shifts almost instantly through social media. Individual leaders matter more than before. Non-state actors—including corporations, AI companies, militias, and religious organizations—have become increasingly influential. Drones and AI now allow relatively small players to shape major events. Traditional state-versus-state analysis is no longer sufficient.
.
Questions like “What is this leader afraid of?”, “What is this society angry about?” and “Which political faction has suddenly fallen silent?”, deeply human factors, are regaining their importance. And this means that the old skills of diplomats and historians are becoming valuable again. In other words, the more powerful computers become, the more analysis returns to something resembling the human observations of Anton Chekhov or Leo Tolstoy.
.
AI should endeavour to be able to deal with this.
Leave a Reply