“You don’t need to predict the future. Just choose a future — a good future, a useful future — and make the kind of prediction that will alter human emotions and reactions in such a way that the future you predicted will be brought about. Better to make a good future than predict a bad one.”
Isaac Asimov, Prelude to Foundation
If you like hard science fiction with stories that evolve over thousands of years and detailed characters, then you should read Asimov[i]. In particular, the Foundation series. How is this related to Predictive Analytics? The Foundation series is based on the concept of Psychohistory, which is the study of the future, based on maths, at the level of an entire population[ii].
Isaac Asimov: Professor of biochemistry and prolific author
Hari Seldon, the main character at the beginning of the series beginning, is a mathematician specializing on Psychohistory. He can predict the future but only at a large scale. He will use his knowledge to prepare Humanity to recover from an unavoidable war which will destroy nearly everything (I will let you read the series to know more about that). The first question is: was Asimov a precursor in Predictive Analytics?
There is a major difference between Psychohistory and Predictive Analytics: the target scale. Whereas Psychohistory focuses on predicting behaviour at a population level, Predictive Analytics is applied at an individual level. Obviously, objectives are not the same. Predictive Analytics is driven by applications such as marketing and fraud detection. Psychohistory is used to study the evolution of an entire civilization.
Foundation and Empire, the second book of the Foundation Series
The next question is: which of the two disciplines is the toughest? According to Pedros Domingo[iii], Psychohistory is harder than Predictive Analytics. In his book The Master Algorithm[iv], he writes:
“In Isaac Asimov Foundation, the scientist Hari Seldon manages to mathematically predict the future of humanity and thereby save it from decadence. […] According to Seldon people are like molecules in a gas, and the law of large numbers ensures that even if individuals are unpredictable, whole societies aren’t. Relation learning reveals why this is not the case. If people were independent each making decisions in isolation, societies would indeed be predictable, because all those random decisions would add up to a fairly constant average. But when people interact, larger assemblies can be less predictable than smaller ones, not more.”
This is certainly the reason why Predictive Analytics is currently used, while Psychohistory remains science fiction as of today. Asimov proposed two main axioms for Psychohistory, and their relations to Predictive Analytics are worth discussing:
Axiom 1 – “The population whose behaviour was modeled should be sufficiently large”
Axiom 2 – “The population should remain in ignorance of the results of the application of psychohistorical analyses”
On the first axiom, Daniel Zeng, writes in a recent editorial of IEEE Intelligent Systems[v] that it is becoming “irrelevant” in a world of Big Data. The second axiom is also valid for specific Predictive Analytics applications. In fraud detection, for example, if the fraudster knows the model used to discover him, he will change his behaviour accordingly.
In conclusion, although Psychohistory is fictional, it shares common aspects with Predictive Analytics. Psychohistory, although still science fiction, is definitely a key research area and actors from Wall Street and Governments would pay a fortune to predict the future of an entire country, over a long period of time. Wouldn’t you?
This article was originally published in the Swiss Analytics Magazine (issue 5)
[iii] Professor and author of The Master Algorithm (http://tinyurl.com/zxqqfow)
[iv] Reviewed on dataminingblog.com (http://www.dataminingblog.com/data-science-book-review-the-master-algorithm)