Thursday, October 20, 2016

Tools that Think for Us

Section - A  — Big Data and Algorithms

In March of 2016 a computer named Alpha Go, took another big step towards showing that our new tools are capable of solving more complex problems. It beat the world’s best GO player at that ancient game which has almost an infinite number of moves. This was more difficult than just combing through tons of data like Deep Blue did when it beat the worlds’ best chess player, Gary Kasparov, in 1997, or when Watson beat Brad Rutter and Ken Jennings at Jeopardy in 2011. In this contest, Alpha Go beat Lee Se Dol four matches to one.  It showed that a software program developed at Google could learn from experience and improve its own performance.  
The program was designed to improve itself based upon it’s successes and failures, even in very complex circumstances. That’s what smart humans do, but the machine’s approach was very different.  Another major difference between the machine and human competitors, is that the machine was unaffected by fatigue or psychological pressure.  Lee Se Dol emphasized that difference in his post-match interview. "It is different, there's no doubt. First of all, its not human. It took time for me to get used to its playing style," Lee said."It's not shaken up psychologically and stays focused right until the end," he said.  
It seems that one of the reasons that Lee is such a great champion is his ability to read his opponent and to push a person to his limits. People react when they feel they have made a mistake or if they think they are beginning to lose (see: Trump).  That was not true with Alpha Go.  Observers felt the machine made a strategic error early in the last match, but that didn’t upset it  — it was unaware of it really, since it is not really “aware” of anything, and it never gets “upset.”  As play continued it was able to make the necessary corrections.  
Besides the obvious advantage of computing power, it is this  lack of emotion that is the major difference between people and our machines and tools.  Machines have no prejudices or biases; they just look at the data.  And the data is another big difference.  The start of the new millennium was the time that the term “Big Data began to gain a lot of traction.  People in many fields were realizing that the amount of information that was being produced, and the amount of different factors that computers could collect and analyze, was way beyond what any human mind could keep up with, or make sense out of.  
“Big Data” is now the term for how machines do that job for us.  Big Data is the answer to the problems of being overwhelmed by the five “Vs” of information: Volume, Variety, Velocity, Variability and Veracity.  What many people don’t realize is how pervasive the use of Big Data has already become in our lives, and it will play a much greater role in the future.   It is what every retailer uses to follow what we have looked at, what we bought and then determine what they can sell us next.  It is what every political party uses to see who we may vote for, and how they can try to change our mind.  It is how sports teams decide what players deserve a lot of money.  It is what your insurance company uses to decide how much risk there is in your life. It is what doctors are using to help with diagnoses and treatment.
For several years now I have been trying to persuade my colleagues who do psychotherapy to take advantage of the capabilities of Big Data to improve their diagnostic skills and clinical interventions.  The overwhelming response I’ve received has been that would be one of the most counter-productive and anti-clinical actions anyone could take.  It would remove the element of therapy that is most important to making it work, the “human” element.
I strongly agree with that.  Because what  the “human” element really means is to make a decision based upon intuition, and emotion.  A person’s intuition comes from his or her own experience and that experience is very limited compared to all the data that is complied by a super-computer.  Likewise an emotional decision, which is a major element in almost all human decisions, is very biased as it is based upon that person’s immediate, protective needs, or short-term desires. This is true, even for therapist (imagine that).

Big Data takes the huge volume of data that is feed into it.  Sorts through it all with great speed (velocity). It finds regular, reoccurring patters in the Variety of the data, and can check on the Veracity of those patterns.  It does this with no pre-conceived ideas ( If it had a mind it would be an open mind, but of course, it doesn’t have a mind), and it does it all much faster and more accurately than any person can.
Big Data can find early indications of depression in ways that no clinician can.  In fact it can detect signs of depression even earlier than the person who is becoming depressed. There are examples of machines showing that when a person’s activities show minor changes, that it could either be a good sign or a bad sign.  If a man has been going to Home Depot, or Auto Zone almost every weekend for five years, and then he stops going, it may indicate that he has found a girlfriend who he values more, or he is getting depressed and no longer cares about his home or car as much as he always had.
There are many such patterns that Big Data already has about all of us.  We all have generated data that gets fed into huge data bases that are kept by retail companies, credit card companies, governments, political parties, and who knows who else.  It is not hard to track what we buy, how we spend our disposable income, what books and movies we watch, how much we exercise, what kinds of foods we eat, who we spend time with, and what vices we enjoy.  From all that it is easy to generate what we are about to do next. “People who listen to David Bowie also like David Bryne” is one of the more innocuous examples. If anyone could access the data you have voluntarily complied on you phone they would know much more about you, and be able to make more accurate predictions about how you will behave than you probably could for yourself.
But the main advantage / disadvantage of Big Data is that it doesn’t care. It shows things that are counter-intuitive, or that people don’t want to accept, for example that children who are raised with two parents of the same sex grow up to be heterosexual at the same rate as children of children raised by heterosexual parents, or that Republicans have the same general IQ level as Democrats, or that it’s better to have LeBron on your team than Stephan Curry.
These computer tools we have created will only get better at what they do. As we can see from the GO game, they will improve by correcting their own mistakes. Many more of our decisions will be guided by computer algorithms. This will be true for large scale decisions such as finance, city planning, health care, and elections, as well as individual decisions, such as our choices in music, restaurants, schools, or job fitness. 

Do we want this kind of guidance?  Will we accept it and use it?  Will we even be aware of how much we are being influenced?  How many of these algorithms are made for our best interest, or are the there to promote someone’s profits?



Coming Next:  Section B: But We Are Human


1 comment:

Forsythia said...

It sounds like medical doctors who take advantage of the benefits of Big Data can be better (more "human") doctors. I'd think it would be a relief to a doctor meeting each week with dozens of patients (some of them real chatterboxes) to have reliable help in making sense of all the "noise." You'd think psychologists would be more open-minded about this. Still, I don't know that much about Big Data--have had no first-hand experience in its use. Does it detect patterns on its won or does someone program it to search for information and classify it in certain ways? Sounds good, sounds bad, sounds beneficial, sounds scary--I dunno.