This week I had a chance to assist a conference with The Economist Editor Daniel Franklin. Titled as “The World in 2016”, the speech was all about expectations for this year. Mr. Franklin exposed their views of our world for the next 12 month through 12 different groups of issues. And one of the new trends were the emergence of the so called superforecasters.
I immediately realized that this may be something interesting. I love forecasting, maybe because of my love for Economics and Politics, but I never thought that there are some terminology for people that are doing this (in my case I just do it for fun, but hey, who knows…). So I immediately googled this new term.
Superforecaster is the person who excel at predicting future. Generally they are amateurs and not professional analysts. Initially these guys were identified through a contest led by IARPA (Intelligence Advanced Research Projects Activity) and were challenged against real world CIA analysts with access to sensitive information. Superforecasters supposed to use only public data. Using Brier score, a proper score function that measures the accuracy of probabilistic predictions, amateur superforectasters scored 30% higher than CIA analysts. Isn’t this amazing?
Superforectasters are not genius. They score high at IQ test (above 80% of other respondents) but far from being at top positions. They are good predicting elections, conflicts, political situations and so on. To forecast they use mathematical models and build probabilistic scenarios with different grades of possibility for each event. Basically this is part of what I learned at decision making course while doing my MBA. And I should say that there’s a lot of fun in doing predictions. You start with feeling, then you take data, build and correlate it and if you get a trend you try to give it a percentage of probability that the trend will continue in the future. When you forecast a far future, you maybe even make only qualitative approach, but if your try to predict near future you cannot do it without consistent data.
What can we predict just now with available data? For example the fact that current generation in Spain most likely will have to pay for pension twice: now for current pensioners and afterwards for their own pension when they retire. It’s easy to predict because there’s a clear data and trend is consistent. For the time when I’ll retire, there will be less than 2 people in working age to pay my pension. If I imagine the world of full employment, it’s still not enough. So these forecast are easy. There are more difficult forecast, specially related to situations when paradigm changes. Such change may be an emergence of Artificial Intelligence (AI). In this case we may speak about the dawn of Singularity, a threshold of events beyond which point it’s impossible to predict anything. If we are thinking about ultra-short term predictions, the emergence of big data is playing crucial role in all the industries. Forecasting the demand, flow of vehicles and people, climate changes, virus epidemics and so on… The possibilities that are opening now with more and more sources of information and processing algorithms is just huge. I think we live in the amazing moment when forecasting becomes ubiquitous. And superforecasters are these who naturally have the ability to digest the data and propose proper algorithm to understand it.
In 2015, Philip E. Tetlock and Dan Gardner, prominent forecasters, released a book Superforecasting: The Art and Science of Prediction. I hope I’ll have a chance to read it soon because as per critics it seems quite interesting. According to The Wall Street Journal, Superforecasting is
“The most important book on decision making since Daniel Kahneman’s “Thinking, Fast and Slow”
The Harvard Business Review paired it to the book The Power of Mathematical Thinking by Jordan Ellenberg.
For these curious about what The Economist predicted for 2016, here you have a short film.
Below you also can find an interesting article from Financial Times about history of forecasting.