Unlearning 101

Bet Against Yourself

 

 “Don’t wanna learn from nobody what I gotta unlearn.” Bob Dylan

 

 

Question: Which of the following characteristics has a higher correlation to the success of a Hollywood movie: the involvement of a famous movie star or the location(s) where the movie is shot?

 

The answer, to the surprise and chagrin of many Hollywood producers is the latter. The fact that Mel Gibson, Meryl Streep or any other Hollywood “A” list star performs in the movie only has a small correlation. As Ian Ayres recounts in his book, Super Crunchers, a company, Epagogix, now uses a proprietary neural network equation that relies on little more than the information found in a movie’s screenplay to predict -- with a higher level of accuracy than Hollywood’s most experienced directors and producers -- how financially successful the movie will be.

 

Not surprisingly this news has not been received well among many Hollywood elites because it highlights how little they know about which characteristics truly matter most to a movie’s bottom-line.

 

This habit highlights another behavior individuals will need to unlearn as they move into the future, and this is the idea of assigning too much weight to one’s own opinions or intuition. Extraordinary new capabilities in data mining and computer processing power, when paired with equally powerful advances in algorithms, neural networks and predictive analysis software, are proving that machines are frequently far better judges than humans at many important tasks.

 

For example, researchers have found that for patients with uncommon medical conditions Google is now more reliable at assessing those conditions than your own doctor. Considering that the average doctor has access to only two million pieces of medical information and Google a thousand times more this isn’t that surprising. What is worth considering is the idea that as computer hardware and software improves exponentially so will Google’s ability to diagnose a growing number of diseases with a higher rate of accuracy.

 

In the rarified field of oenology (the study of wine), Orley Ashenfelter has produced a sophisticated regression equation that can predict more accurately than the wine industry’s top judges which vintages of red wine from France’s Bordeaux’s region will be the most valuable in the future. Ashenfelter cares nothing for swishing wine around in his mouth in search of subtle hints of oak, cherry, tobacco or blueberry and instead relies on weather-related data.

 

For wine collectors who invest in wine futures, Ashenfelter’s information is giving them a distinct competitive advantage over those who can’t or won’t unlearn their reliance on wine “experts.”

 

And only recently have many Las Vegas hotels and casinos unlearned the idea that its most valuable customers are limited to the ranks of wealthy, high-rollers. Thanks to complex algorithms hundreds of middle-aged, middle-income individuals are now being aggressively courted. If you’ve ever wondered why your ne’er-do-well Uncle Ned is being “comped” to stay at a nice hotel in Las Vegas twice a year, it’s because a computer program has figured that it is likely to make money off him.

 

Therefore, if you ever happen to receive a free weekend in Las Vegas courtesy of a hotel-casino, my advice is to bet against yourself and stay home. It will save you money.

Homework assignment: Recalling that it took Galileo years to convince people that light objects fell as quickly as heavy objects even though they could see the results with their own eyes, how long – if ever -- do you think it will take before all doctors will be required to confirm their diagnosis of patients with a machine? Defend your answer.

Related Posts

Stop Seeing Patterns

Think Small

Ignore the Eclipse and Admire the Sunset

Grow a "Knew" Perspective

See What Isn't There

Unlearn Your Intuition (Sometimes)

Recently, I gave two presentations on the topic of my latest book, Jump the Curve, at a State Managers Conference in Minneapolis, Minnesota. I discussed a great many topics, but I would like to highlight just one: “Embracing counter-intuitive findings.”

Today, a great many people, including education, public health and law enforcement officials are certain as a result of years on the job that they know what works and what doesn’t work. For instance, an education official might be convinced that vouchers and charter school don’t work; a public health worker might lay claim to knowing which programs are most effective; and many law enforcement officials are positive that stronger criminal penalties lead to lower recidivism rates.

Over time, many of these personal opinions have morphed into “conventional wisdom” and, worse, because many of these state officials have risen to prominent places in the state bureaucracy this “conventional wisdom” has been adopted as de facto state policy.

As a result of exponential advances in the amount of information that is now being collected as well as the vast improvement in algorithms, however, savvy computer scientists and social scientists are now discovering some surprising things. I am a big fan of Ian Ayres excellent book, Super Crunchers, and throughout it he systematically demonstrates how new insights gleened from algorithms are forcing educators, public health and law enforcement officials (and many, many others) to revisited some of their most basic assumptions.

For instance, educators are learning how different policies effect different populations in different ways. The same is true for numerous other public policies. For too long it has been too easy to simply adopt a black-and-white, one-size-fits-all view of the world. There is now so much information available and it is being “crunched” in so many different ways that it is no longer sufficient to rely on intuition or “gut feeling.” Unless you can back up your adherence to “conventional wisdom” with hard facts, my recommendation is that you should “crunch” your assumptions against the data because there is a solid chance you might have to unlearn your intuition.

New Voice Recognition Technology Mandates Unlearning

Late yesterday, Google released a very cool new mobile application which employs voice recognition technology. The question is not so much what the technology can do today, the question is what will the technology be able to do in the near future—and how might it require professionals in the education, health care, and a host of other businesses to unlearn?

I’d love to hear your thoughts on how this technology could unfold and how it might alter people’s behavior. Below are my initial thoughts from my other website: www.jumpthecurve.net:

Unlearning Your Daily Commute

I live in a major metropolitan area and the traffic situation drives me crazy. If someone then told me that they were going to close a road in order to reduce traffic congestion, I’d likely think they were either crazy or stupid. As I mention in my book, Jump the Curve, emerging technologies can, however, often throw people—myself included—for a loop. This story is a perfect case-in-point.

One would think that perfect information on traffic conditions would lead to an optimization of people’s personal travel time. Alas, this is not always the case. In fact, people’s commutes in major cities such as Boston, London and New York can be reduced by 30%, 24% and 28% by selecting the “social optimum” choice instead of the "personally optimum" choice. And, if people are reluctant to unlearn this finding because it defies conventional wisdom, local governments or traffic departments can always choose to close a road or two and force them into a better (i.e. faster) choice . Of course such a scenario smacks of Big Brother-like socialism and makes the civil libertarian in me cringe, but if you read the article you’ll at least understand the logic behind the idea.

If You Want to Win: Practice Your Unlearning Game

On December 30, 1936, over 17,000 fans packed into the old Madison Square Garden to watch Long Island University, the country’s top-ranked basketball team, take on Stanford. It was slated to be a great game. Long Island was putting up its 43-game winning streak up against the reigning Pacific Coast Conference champions.

More than that, though, people were there to watch Hank Luisetti, Stanford’s 6 foot 2 inch star sophomore. He was the only player known for shooting the ball with one hand while jumping up in the air. At the time, every other player shot with two-hands or took hook shots.

Luisetti’s “jump-shot” was so radical that it caused Nat Holman, the legendary coach of City College of New York—and a man known as “Mr. Basketball” – to remark, “That’s not basketball! If my boys ever shot one-handed, I’d quit coaching.”

Stanford went on to crush Long Island University and, two years, later, Luisetti became the first college player ever to score 50 points in a game. Today, it is impossible to find a single player at either the collegiate or professional level who shoots two-handed.

It took a long time for basketball players to unlearn the two-handed shot, but the story serves as a good reminder that unlearning is not easy and that—like most sports—it is something which must be practiced continuously.

In fact, the sporting world is an excellent place to study the practice of unlearning. On May 6, 1954, Roger Bannister unlearned the “four-minute mile” barrier when he circled the track at Oxford in a then record of 3:59.6. Dick Fosbury similarly unlearned the traditional straddle method of high-jumping when he invented the now famous “Fosbury Flop.”

Alas, unlearning is not so easy. In the insightful book, The Wisdom of Crowds, James Surowiecki, recounted a study conducted by David Romer, an economist at Berkeley. In the study, Romer reviewed 1,100 first-quarter fourth-down plays where the teams would have been better off going for it. Instead, they kicked the ball 992 times. In blunter terms this means that in 90% of the cases, the team’s coaches were actively working against their team’s prospects of winning.

Is such insanity limited to the gridiron? No. In the book, Moneyball, Michael Lewis demonstrated how the Oakland A’s general manager Billy Beane unlearned his reliance on crusty old baseball scouts and instead turned to young, Ivy League educated “number crunchers” to help him spot the top prospects to draft. The book was published in 1999. To date, and in spite of the A’s consistently fielding one of the best teams with the lowest payroll, only one other team has adopted the strategy in a widespread manner—the Boston Red Sox.

Perhaps it is a coincidence—and a healthy payroll—but in case you haven’t been paying attention the Red Sox have won two of the past four World Series. It is also, however, possible that Boston has been practicing its unlearning skills. In 1947, Jackie Robinson became the first African-American to play in the Major Leagues. The Boston Red Sox were the last team to integrate—doing so only in 1957. During that time it won no American League pennants and, of course, no World Series.

Bottom-line: If you to want to win, practice unlearning.

Unlearning the Future

The future is unknowable. There are far too many variables for even the most foresighted individual or powerful supercomputer to accurately forecast what tomorrow—let alone next year or the next decade—will look like with precision. Nevertheless, this fact neither discounts the importance of forecasting, nor does it diminish the work that those individuals (myself included) try to do in discerning what the future might hold in store.

I would, however, like to submit that anyone inclined toward thinking about the future should be open to the idea of unlearning, which I define as “the unique skill of jettisoning old knowledge in order to accommodate newer and more relevant information.”

A case in point is this new study suggesting that global warming may not be worsening hurricanes. Now, before anyone gets too not and bothered by the real or perceived flaws in the study’s methodology, I’d like to make clear that it is not my contention that this study is the final word on the topic. Rather, I simply want to highlight it as an example of how continued advances in the development of better and more sophisticated supercomputers, algorithms and ubiquitous sensors are likely to lead to more situations in the future where scientists and researchers produce results that question and challenge conventional wisdom. (To this point, ever since Hurricane Katrina many people have come to believe that there is a direct correlation between global climate change and the frequency and severity of hurricanes, and often this belief has lead them to predict that more hurricanes are in our future.)

The job of forecasters and futurists, however, is to be receptive to contradictory information—especially when it challenges fundamental beliefs or assumptions about the future.

History is littered with examples of yesterday’s dogma being mocked and ridiculed by the next generation. There is no reason to think that many of our most cherished beliefs won’t be similarly mocked and ridiculed in the future.

One way to avoid this fate is to have the courage to “unlearn” things whenever new and compelling information becomes available.

Unlearning Choice

It is easy to believe that having more choices will lead to a greater level of happiness. However, as this new study suggests, people may be more satisfied when selecting from a smaller set of options.

There are a couple of reasons for this somewhat counter-intuitive finding, but the most poignant seems to be that when people select from a wider array of choices they often experience a greater sense of regret (after their selection) because they aren't as confident that they made the optimal choice.

Trying to convince people that having fewer choices is a good idea will be difficult endeavor, but smart companies may be able to enhance their customers feeling of satisfaction by presenting information in new and innovative methods. For instance, think of on-line dating companies. Instead of presenting customers with an ever widening array of potential partners, they might instead employ sophisticated algorithms to narrow the list of prospective mates from which a person selects their choice.

Unlearning Computer Programming

One of the reasons I have established this blog on unlearning on Typepad is because it allows me to maintain and modifiy the site easily without knowing a lick of computer programming or software development. And, trust me, I don't know jack about those fields.

Nevertheless, I recently stumbled upon this old article, The Joy of Unlearning, which was written by Andy Hunt -- who is knowledgable about software. The article encourages programmers to embrace unlearning. I especially liked this quote:

Old habits are hard to break, and even harder to notice. The first step to unlearning--and the hardest part--is to realize that you're using an outdated approach. The next hardest part is actually letting go. Mental models and patterns of thought are built and refined at great cost, and one doesn't discard them lightly. It takes awareness, conscious effort, and hard work.

It's great advice and it doesn't just apply to programmers. It applies to all of us.

Unlearning Models

Chris Anderson, the editor of Wired, has written an excellent article entitled "The End of Theory: The Data Deluge Makes Scientific Method Obsolete" in which he convincingly argues that massive amounts of data, in combination with sophisticated algorithms and super powerful computers, offers mankind a whole new way of understanding the world.

Anderson believes that our technological tools have now progressed to the point where the "old way" of doing science -- hypothesize, model and test -- is becoming obsolete. In its place, a new paradigm is now emerging whereby scientists, researchers and entrepreneurs simply allow statistical algorithms to find patterns where science cannot.

If Anderson is correct -- and I believe he very well could be -- this will take unlearning to a whole new level. In short, instead of having to do the unlearning ourselves (e.g. when our hypothesis doesn't work out or is proved wrong) scientists and researchers can instead rely on intelligent algorithms to do the heavy lifting.

The implication for a field such as biology which, as Anderson points out is actually becoming more difficult to model as learn more about it (due to our limited understanding of how genetics, microbes, personal behavior, the environment and a host of other factors work in partnership to determine a person's health), could be profound.

More specifically for the field of unlearning, this means we will be able to analyze data without allowing hypotheses (which may be wrong) to cloud our view of what the data is really showing us.

Before this vision can be achieved, however, it will require a great many brilliant scientists to unlearn the idea that their "model-based" method of trying to make sense of today's increasingly complex world is the only way to search for new meaning. Increasingly, we will be able to allow computers to crunch the data and extract the meaning for us.