I apologize up front for my abrasive style, but fortunately I have not substituted my emotional strength with a lifeless ’super crunching’ machine, despite being a software architect. This is not about atacking a fellow author at all. It is about a strong disagreement in opinion.Ian Ayres is a Yale Law School professor and claims that statistical methods enable knowledgeable individuals to make better predictions. Ayres tries to proove that data-driven decisionmaking is improving education policy, health care, and even tax regulation and government. Ayres does as statisticians do: PRESELECT the input data until the outcome is as desired. Ayres says that humans put far too much trust in their intuition and would often be better off listening to the numbers. Right, but these are mostly given to them by large corporations and the government. It is statistics, so it MUST BE true!
This book promotes – with a few examples but no proof whatsoever – the idea that our life is being improved by ‘experts’ who use statistical analysis and prediction. But worse is to come as in chapter 8 Ayres writes about statistical standard deviation and how it applies it to IQ, student grades and the stock market. I propose that IQ measurements and student grades are dehumanizing and tell us nothing about how our children will fare in life! When will we finally stop to discriminate humans in this horrible way?
To understand the stupidity in stock markets I suggest to read “Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets” by Nassim Nicholas Taleb. If the ‘experts’ using Super Crunching are so smart Mr. Ayres, why do we have a crisis with bad real estate loans in the US?
There is nothing wrong with using statistics to find out what happened in the past. One has to be very careful to collect that information and interpret it. Ayres uses numerous mostly disjointed examples that are supposed to show how powerful and accurate statistics are. He mixes statistical data mining and interactive random testing and other methods and calls it Super Crunching. None of these methods provide better decision making (a.k.a. as being SMART), rather the opposite. For statistics to be useful, data filtering is the key. Therefore a human conscious decision has to be made beforehand about what data is going to be used. After the data has been collected human intuitive (experience based) decisions have to be used to act. The same intuitive decision can be made by simply exposing the decision maker to a subset of that information. Those decisions can be made in real time while statistics are about the past. I am not talking about normal just-in-time warehousing logistics as used by the likes of Walmart. The most successful methods to improve decision making limit information gathering to the relevant detail and do not provide more data.
From my perspective, Business Intelligence and Data Mining produce an unintelligible mess of data that has little to no meaning for business decisions, automated or not. Imagine that we create codified decision-making knowledge based on abstract statistical illusions? That will ruin each and every business. Humans solve this problem of decision-making on inaccurate data by use of an emotionally weighted pattern-matching ability of the brain – its called INTUITION. Read “Gut Feelings: The Intelligence of the Unconscious” by Gerd Gigerenzer. Emotions play an important part in that process. Are not strong managers and entrepreneurs mostly very emotional, even unreasonable people? Antonio Damasio – today Professor of Neuroscience at the University of Southern California – has long researched neural systems for memory, language, emotion, and decision-making. In his 1994 book, “Descartes’ Error: Emotion, Reason, and the Human Brain” he documents his discovery that “humans with dysfunctional emotional centers face grave difficulties in decision-making.”
To improve decision making with the help of computing we need to invent technology to model emotional human decision-making far beyond logical rules. While Ayres mentions Neural Networks and Pattern Matching he fails to understand that both do not need masses of data but REAL decision-points linked to real data. It is not important how many people took a certain decision, but what data pattern was used by each individual to come to the decision. The Google-Ad testing Ayres did for the book title is very basic decision point sampling but he misses to distinguish it.
Ayres asks a few critical questions but then claims that Super Crunching is better than intuiton. Decision making based on abstract statistical data illusions is one of the reasons why governments and large corporations are so inept. The sizes of organizations in business and government have simply become too large to be managed well. The reason is greed for money and power and not the wished for economies of scale. Statistical data analysis is no more than a desperate try to solve that. The ‘expert elite’ claims they know what they are doing. Read Thomas Sowell’s “Knowledge and Decisions” and “The Vision of the Anointed: Self-Congratulation As a Basis for Social Policy” to understand that it is doomed to fail.