Can Bruce Bueno de Mesquita predict the future?

So here’s one I first took on in Quora, and it’s a fun topic –

Can Bruce Bueno de Mesquita predict the future?

NYU Professor BBdM claims to predict international policy events with >90% accuracy in his book “The predictioneers game”. It seems grotesque, now we have a paper where he descibes his method. Can you falsify his statement?  http://irworkshop.sites.yale.edu…BBdM Ted Talk
Ted Galpin, Astrophysicist cheerleader turned professional strategist

I’ve been following Bruce Bueno de Mesquita on and off for roughly a decade.

In theory – Yes, sometimes, with limits he can predict the future (And his marketing is probably exaggerating, but he’s not really doing anything that other mathematicians are not doing, he just sells it really well).

He’s been doing this stuff since the 80’s. But I guess it took a TED talk and a Book for people to really take notice.

Here’s some context:

  • He famously was one of the earlier academics to successfully apply game theory to political science; and everybody laughed at him until he proved to be right (He predicted an “unlikely” candidate would win an Iranian election in the 1980’s; and 2 years later proved to be right, reportedly surprising the political science community). That when the Poly Sci Quantitative vs Qualitative culture wars began.
  • He consequently started selling his services to the government.
  • His models are supposed to be rational choice driven game theory models that are heavily data dependent.
  • The famous CIA quote is that a good CIA analyst can tell you a what political party should win the next election with about 80 percent accuracy.  Bruce Bueno deMesquita – BBdM could then interview the CIA analyst, populate the variables, do the math, and give you a NAME that was right 90% of the time. Technically BBdM adds precision to existing accuracy.

Good analysis give you accuracy. BBdM adds precision by using good math on top of good behavioral science. If the underlying data points you in the wrong direction, the math probably won’t change that.

  • We know from fundamentals of math that every model he builds would be custom based on available data; using a underlying algorithm that he’s been tweaking and using since the 1980’s.
  • We know from mathematical modeling 101 that the ability to predict the future is limited by available data and known patterns. That has plagued engineers and scientists for centuries.
  • The fundamental assumption in BBdM’s models is they are built on a version of rational choice theory.
  • Rational Choice theory is often applied incorrectly. It states that people usually do what is in their best interests.
    • Rational Choice theory actually means people usually do what they believe is in their best interests (from their own point of view). Important pragmatic semantic.
      • For example – you know people will order Pizza because they like Pizza, but really they should order a salad because it is much healthier.
      • Rational choice is relative to understanding the subject. Hence why it has been misused and easily critiqued in the past.
  • BBdM’s best work tops out at 90% for a few reasons:
    • 1 – Incomplete information leads to incomplete mathematical precision and accuracy. I.E. You are vulnerable to Black Swans (what you don’t know).
    • 2 – It requires you to accurately read people’s minds – this is where the knowledge of an analyst comes in. If your analysts don’t understand the psychology of the individuals, then they may guess wrong on the rational choice, and the math can’t fix that.
    • 3 – It really should be lower than 90%, I’m guessing they only pick battles they know they can win to cheat the metric – a common business practice.
  • Because the model’s are based on rational choice, BBdM can only predict decision making of people or groups of people. And only if you can identify the influencing players of the game, model the interactions of the players, the decisions they will make, and the resultant net result when you add up all the decisions based on rational choice. Basically a giant decision tree matrix.
  • Technically speaking, a good strategist with good data using the right math and the right science should be able to consistently predict the future. This is what good management consultants try to emulate.
  • If you have any doubt, I know from my business, every Billion Dollar piece of technology (think jets, mines, power plants, bridges, refineries, Sub stations, sky scrapers) is predicatively modeled on paper and in computers to make sure it is possible, will work, and to estimate cost and economics long before they spend the billions of dollars to build it (often in the form of an engineering feasibility study).
  • Any situation where you have enough science and measurement,  the appropriate math can be used to effectively predict the future. That’s what operations research has been doing for years. That’s what Gantt logic and Agile Velocity try to do with projects.

For example – In WWII German submarines where hard to sink because they went underwater when they saw planes, so it was hard to kill them with aircraft. Operations research analysts ran lots of what we now call data analytics (by hand in the 1940’s), and figured out that if you put bright lights on a bomber, it looks more like the bright blue sky, and an anti submarine bomber could get close enough to kill a submarine before it was spotted and the submarine did a crash dive. The math predicted right that time.

So given all that.

Can BBdM – Bruce Beuno de Mesquita Predict the future? Sometimes, if you ask the right kind of questions, and he can find the right input data for the math to work. He specializes in political science and the decision making of large groups. If pressed he may even be able to model and predict a stand alone complex – but that’s kind of obscure and I don’t think anyone is looking for those yet (outside of social moment and viral based marketing techniques).

But Hannah Fry could predict how riots happen in Python a few years ago (not when, but how). Operations research has made progress in future prediction for many, many decades. And Game theory is pretty old – that’s a big part of how Rand sold Mutually Assured Destruction theory and led them to purportedly recommend economic brinkmanship to end the cold war – knowing the expensive weapons being built were unlikely to be used in WWIII. And that worked, the cold war escalated to the point of bankrupting the Soviet Union with military spending. Technically we tricked them into building a military too large to economically sustain.

To answer your question – yes, you can use math to predict the future. That is sometimes, if you know the science, the math, and have the right data and analysis. That’s why engineers are correct 98% of the time, and guys like BBdM are right only 90% of the time, and only if they are careful and only pick questions they know they can answer.

Given more time – we are likely to see the rise of something analogous to the psychohistorians of Asimov’s Foundation. Actually if you read that link – We are not that far off from that today.

As of 2016? A good Strategist working with a Good mathematician, a Good Data Scientist, a Good Operations Researcher, and a Good Behavioral Scientist should be able to predict accurately, and even maybe precisely the things we understand and know how to measure, but not the things we don’t understand, or can’t measure. And each question would take them months to answer (analyze and model), and even then it is doubtful anyone can ever beat BBdM’s claimed 90% because everybody makes mistakes and has limits.

In reality hitting 90% accuracy is crazy hard, unless you get to select your problems. In my experience “complete” models are about 80% precise (you get within 20% of what you expected) and are accurate (pointed in the right place) about 80% of the time. But getting adequate information to make a complete model is rare.

I’m guessing that BBdM gets his success by having access to the best experts and data the US Government can provide. Most good teams should be able to replicate BBdM’s results given access to that level of resources. Not easy by any means, but theoretically doable. I’ve seen comparable results in different modeling techniques in business – if you constrain the scope of the model enough, you can enhance your accuracy.

The honest take away you can get from BBdM is if you understand what someone thinks or believes in their own best self interest from their own point of view, 9 out of 10 times you should be able to predict what they will do next. Any Game theorists out there able to confirm or deny that with empirical data?

The only claim BBdM has really made is he can mathematically model that rational choice up to scale of decisions made by large populations like how nations vote in electrons. Seeing as he’s been on a sole source contract to the US Government doing that for like 20 years or more – I’m guessing he’s been doing it, and his success rate is good enough to keep him on retainer.

Interesting and very skeptical article that goes into some details here: Bueno de Mesquita’s prediction of Iran’s future – it reminds me of 5 mathematicians having 25 opinions on one topic.

Hope that was enlightening and answered the question.

-Ted S Galpin

This entry was posted in Geopolitics, Math, Strategic Science. Bookmark the permalink.

2 Responses to Can Bruce Bueno de Mesquita predict the future?

  1. Dave says:

    Typo in the next to last paragraph.

    The only claim BBdM has really made is he can mathematically model that rational choice up to scale of decisions made by large populations like how nations vote in electrons

  2. Dave says:

    Seems to me his methods work when the result is predictable by statistics. As gamblers and bookies know, that is not always the case. If he has been advising the U.S. Government for 20 years, how come they have gotten so many statistically predictable thing wrong in that time?

Leave a Reply

Your email address will not be published. Required fields are marked *