Brain Trust (19 page)

Mathematician and computer games expert Jonathan Schaeffer at the University of Alberta solved checkers. After using up to two hundred computers running simultaneously for ten years to consider 10
¹⁴
possible board states, his program, Chinook, eventually discovered a path of optimal play that never loses. And after solving checkers, Schaeffer turned his
cerebral and computational firepower to another game—poker—specifically, two-player, limit Texas Hold ’Em.

In limit poker, you can only bet so much and so the game becomes a fairly mathematical exercise—based on your hole cards and the community cards, what’s the chance of winning? (If this makes no sense, you can familiarize yourself with Texas Hold ’Em rules online.) Generally, if you have a more than 50 percent chance of winning, you bet.

But even this simple version of poker is tricky to study because “you can get lucky and unlucky and bad luck can last for a very long time,” says Schaeffer. Bad luck can make even the best players look like rookies, so in a study it’s hard to disentangle good play from good luck. Schaeffer and his research team found an interesting solution: They played two human-versus-computer pairs, each pair playing the same cards, but with the hands reversed. This way both sides get lucky or unlucky to the same degree. By comparing win/loss rates, Schaeffer and his collaborators discovered what strategies won over time.

“The best way to play against a computer is mathematically,” says Schaeffer. Aggression or conservatism are trends that a computer will recognize and exploit.

“But playing against a human, aggression is correlated with success,” says Schaeffer. “Pushing a lot of money into the pot forces your opponents to make many tricky decisions,” and especially against weak opponents, these decisions are likely to result in mistakes.

When playing suckers, push chips.

In a surprisingly fun and interesting paper
subtitled “Human Perfection at Checkers,” Schaeffer tells the story of his program Chinook’s 1994 battle against the human checkers champion, Marion Tinsley. In thirty-nine games, there were thirty-three draws, four wins for Tinsley, and two for Chinook. This was a great triumph for Chinook, considering that Tinsley only lost five other games between 1950 and his death in 1995 (paper linked from Schaeffer’s faculty bio). What made Tinsley so great? According to Schaeffer, it was Tinsley’s uncanny memory that even toward the end of his life allowed him to quote move sequences from games dating back to 1947, and a sixth sense born of experience. Yes, when playing checkers Tinsley “just knew” the best move—but, according to Schaeffer, it was because Tinsley had painstakingly added these moves to his mental Rolodex over many thousands of hours of play and study.

There are 43,252,003,274,489,856,000
possible states for a Rubik’s Cube. Researchers at Kent State opened up a can of supercomputing whoopass on these states, showing that the max number of moves needed to solve the cube at any time is twenty. This task would’ve taken a desktop computer thirty-five years, but took supercomputers at Google about a week.

What happens when you get a decent meal at a great restaurant? Or pay fourteen dollars to see an Oscar-winning film that turns out to be so-so? Or get a 3 percent raise when you expected 5 percent? You’re disappointed, that’s what. You expected greatness, you got mediocrity, and you’re pissed. But look at it another way: Dude, you got a raise! That’s awesome.

This is what economists call gain/loss reference dependence—human happiness really isn’t about the amount of liquid in the glass, it’s about that old half-full, half-empty thing. Or, more precisely, it’s about how much liquid you expect to be in the glass. More than you expect and you’re chuffed; less and you’re disappointed.

How strong is the effect? It’s hard to tell because it’s extremely tricky to create emotionally charged expectations in the lab—a prerequisite to smashing these expectations and seeing how mad people get (or exceeding these expectations and looking for happiness).

So instead, Gordon Dahl, economist at the University of California–San Diego, turned to football. Before a game starts, there exists a very definite measure of expectations in the form of Vegas odds. You know who’s supposed to win and by how much. And you know how emotionally charged a game is—is the team in playoff contention? Is the game against a traditional rival? Finally, there’s an unfortunate but telling measure of how people are affected by football outcomes.

“We find that upset losses lead to a 10 percent increase in
domestic violence in the losing team’s home city,” says Dahl. If the losing team has an unusually high number of sacks and turnovers, make that a 15 percent spike. And if the upset loss is against a traditional rival, domestic violence in the team’s home city increases by 20 percent. You can see it in police reports: As an upset starts to look likely in the game’s final hour, domestic violence starts to climb, peaking just after the game, and returning to normal about two hours after the final whistle.

But it’s not the loss that does it. If a team is expected to lose and then does … there’s no spike. It’s only when a team favored by four or more points chokes that football fans form fists. “The more salient the emotional shock is to you, the worse it is for your spouse,” says Dahl.

And the rosy side isn’t nearly as true. A home team’s upset win does little to lower domestic violence. Or in other terms—sure, getting an unexpected great meal at a questionable restaurant makes you happy, but it doesn’t nearly balance the unhappiness of a great restaurant missing. It’s as if a happy surprise is (for example) +3 while a disappointment is -8. Over time, betting on restaurants is a losing proposition—this is likely one reason we get stuck in safe ruts, eating at the same decent place every time we go out. “But if we learned to manage our expectations, we’d all be better off,” says Dahl.

You can’t really adjust your surprise happiness/unhappiness payouts of +3/-8, but imagine lowering your expectations so that you’re happily surprised at more than two-thirds of the new restaurants you visit. Now, in the long run, you’re better off exploring.

Restaurants aren’t the only medium in which expecting less allows life to frequently exceed your expectations. I grew up a Seattle sports fan and reflecting on how I now watch sports, even when the Mariners or Seahawks are ahead, I’ve ingrained the fatalistic attitude of “well, they’ll probably blow it in the end.”
My expectations stay low and so I’m happily surprised more than I’m disappointed (OK, with Seattle sports this barely allows me to break even).

The trick is to do this without becoming Eeyore. First, remember that your goal is to adjust your expectations without blunting your payoff. You can still root like heck, just imagine chopping a touchdown or three runs off Vegas’s prediction for your team. And keep your lowered expectations to yourself—you don’t want expecting less to lead to getting less.

Gordon Dahl also explored how violent
blockbuster films affect violence in the neighborhoods surrounding theaters. Do violent films create violence? “Surprisingly,” says Dahl, “during the movie, violent crime goes down.” Dahl attributes this to temporary, voluntary incarceration: For the film’s duration, violent people are off the streets. And the rate stays down for a couple hours after the film because after three hours in a theater, these violent people are sober.

GDP, per capita income, unemployment
, educational performance: these are the measures of national well-being. But the United Kingdom hopes to add one more—a national happiness index. What’s cool is that the prime minister imagines its use in driving and evaluating policy decisions. Just as an increase in GDP might be a reason for reform or an indication of an initiative’s success or failure, so too could change in the happiness index drive decisions in government.

“I eat a lot of popcorn,” says B. J. Fogg, experimental psychologist and founder of the Persuasive Technology Lab at Stanford. “I cook it in oil and I eat it at night. It’s a kind of addiction.” But Fogg broke this addiction by announcing to his social network that for the rest of the month, he would become a popcorn teetotaler. This is self-manipulation—by putting his social reputation on the line, Fogg forced himself to change his snacking habits. And he makes a career out of designing technology that does the same to you.

“Can we be manipulated by robots and code into doing things we don’t want to do? The answer is clearly yes,” says Fogg. “But you can’t just grab techniques from Alcoholics Anonymous and apply them to getting people to sign up for Flickr.”

Instead, Fogg’s Behavior Model (
behaviormodel.org
) lists three things that need to be true to change behavior: high motivation, high ability, and a trigger. Think about the person who bought this book. He/she must have wanted to buy it, had the ability to do so—money in the pocket, an
Amazon.com
account, etc.—and then something happened that actually made this person reach for his or her wallet. But exactly how to create these three things depends greatly on what kind of behavior you want to change.

Fogg’s chart of fifteen types of behavior change (
behaviorgrid.org
) crosses five types—do a new behavior, do a familiar behavior, increase an existing behavior, decrease an existing behavior, and stop an existing behavior—with three durations: once, for a duration, and from now on. Fogg applies codes to each of the fifteen types of behavior change, like “BlueDot,” which is performing a familiar behavior one time—for example, buying a book on
Amazon.com
. The type of change coded “BlackSpan” is stopping an existing behavior for a period of time—not eating popcorn for
a month. “PurplePath” describes increasing a behavior from now on, like exercising more.

Fogg’s Behavior Wizard (
behaviorwizard.org
) asks questions that help you define the code of the behavior you’d like to change, and then lands you in the appropriate resource guide. Simply click through the wizard for concrete, usable strategies.

For example, let’s take a look at getting yourself or others to do a familiar behavior one time—a BlueDot behavior. If it ain’t happening right now, ability, motivation, or trigger must be too low (or some combination thereof). Fogg recommends attacking triggers first—they’re the easiest to manipulate and could be a quick fix. For example, if you want to make sure you go for a run this afternoon, schedule a text message for after work saying “Go for a run now!” If you want employees to do the ergonomic wrist stretches they’ve been taught, you can have a manager walk around and encourage an immediate two-minute time-out or send a quick e-mail memo.

If triggers don’t do the trick, Fogg’s next step is to adjust ability, which he divides into the categories time, money, physical effort, mental effort, social deviance (is the behavior unexpected?), and nonroutine (is it out of the ordinary?). For example, if customers are still not ordering movies online even after being bombarded by your e-mail spam campaign (in a benevolent way), perhaps you need to streamline the ordering process and thereby decrease the time or mental effort needed to make a purchase. Or perhaps even with a trigger to go running, you’re stymied by the inability to find a matching pair of running socks. In this case, increase ability by sorting that pile of clean clothes.

Finally, and only finally, does Fogg recommend working with motivation. (To Fogg, going here first is the sure sign of an inexperienced designer.) This is because motivation’s tricky to measure and tricky to adjust in a uniform way. For Fogg, putting his social network reputation on the line increased his motivation to
abstain from popcorn. And maybe for you, imagining toned calf muscles would increase your motivation to run. But others might not care about their calves or mind backsliding on Facebook and may be more motivated to run by the thought of a healthier heart. So it’s tricky. Fogg suggests thinking about motivation in terms of sensation (pleasure/pain), anticipation (hope/fear), and belonging (acceptance/rejection).