“No, you’re not wrong; I’m wrong!”
How often have you heard, or uttered these words? Seldom is my guess. In fact, you may have even misread these words.
Michigan winters are hard. Even in the lower peninsula at the University of Michigan where I went to grad school, winters are long, snowy, bitter cold, and often feature treacherous ice storms. But that made springtime that much more soul-saving. Often, when it was sunny and warm, I would teach my introduction to psychology classes outside on the lawn near Angel Hall. Nearby ran one of the “main drags” in town including a T-shaped intersection. The street going into the main drag included a stop sign for the first three years I lived in Ann Arbor. And, then, there was a change. Whatever the design rationale, the highway department reversed the situation so that traffic on the main drag now had stop signs both ways and the other street was free to turn onto the main drag. That doesn’t seem like a tsunami of a change, does it?
Yet, my classes were often interrupted by screeching tires, and honking horns. Society had not yet evolved to the point of pulling a gun and shooting someone for a traffic faux pas. That would still require years of work on the part of the NRA to convince people that they needed “protection” for road rage (which coincidentally made road rage that much more deadly). But back in the 1970’s, my classes were not interrupted with gunshots. But aside from the screeching tires and honking horns, we could hear plenty of screamed profanity.
What made that an interesting situation to discuss in the intro psych class was that it was never the people who actually had the right of way who did the honking and screaming. It was always (at least so far as I observed) the people who sailed right through the new — and unseen stop signs! These stop signs were in plain view. This was not at all like the stop sign I sailed through years later in Westchester. That stop sign was well-hidden behind trees and then made more invisible by spray paint. I guess some teen-agers thought it would be pretty cool to cause an auto accident. Sigh. But let’s teleport back again through time and space to Ann Arbor a couple decades earlier. Those new Ann Arbor stop signs were large and clearly visible to anyone. In fact, both signs were both easily visible to the psych class from 75 yards away. But they were apparently under a magic spell because these same stop signs were invisible to drivers who had driven the main drag for many years. They “knew” the stop signs were not there. They “knew” there was a stop sign on the cross street. So, to many (not all) drivers, these stops signs were under an “invisibility cloak” created by their own expectations.
Furthermore, when drivers did sail through the stop sign and then found themselves almost in an accident, slamming on their breaks and swerving to avoid the accident, it was invariably followed by a loud blaming exercise. The “blame” of course, was always on the other driver — the one actually following the law. In the 5-6 near misses we observed, we never saw someone sail through a stop sign and then realize their mistake and apologize. Nope. Not once. It was always an anger display at the “idiot” who had gone right through the (non-existent) stop sign. If you read the last blog post about “Big Zig Zag Canyon” you are already familiar with how our expectations of reality can be slow to match actual reality.
Such situations remind me a little of tether ball. As a reminder, tether ball is played with a ball that is…tethered. The ball is much like a volleyball but connected by a rope to a pole. The players try to hit the ball and wind it completely around the pole in “their” direction. (This game is made for two righties or two lefties). Anyway, as the cord wraps itself around the pole once, the cord shortens and the radius of the ball path is shorter meaning it comes around more quickly. So you need to adjust your timing. But the typical behavior, at least for beginners, is to jump up a little late because everyone bases their timing on the previous cycle rather than the next cycle. The player realizes they are late and adjusts their timing. Unfortunately, they typically adjust to the last cycle and are once again late. They do keep adjusting but always one revolution too late. As a result, the ball whips around faster and faster wrapping itself into the pole.
In attempts to build artificial intelligence systems, computer scientists encounter the “update problem.” As the world changes, so too must the reactions of the system change. But what kind of change in the environment is related to which changes in necessary reactions? In many cases, humans are pretty good at this. In other cases, not so much. Let’s say, for instance, that you routinely set your clock radio for 7 am in the morning. One evening, you go out for dinner at the Fish Market and bring home left-overs which you put in your fridge. Now, you immediately go and make sure your alarm is set for 7 am, right? No, of course not! You have a model of the world that enables you to realize without any conscious thought that putting leftovers into the fridge in the kitchen will not change your alarm setting.
Let’s take another example. You drive to a golf course and park. You take out your clubs and get ready to play a round. But you realize you need a new golf glove so you buy one at the check-in desk. Fine. But now you play the entire round wondering where your car will be when you’re done. No you don’t! Of course not! Again, your model of the world allows you to realize that there is no way buying a new golf glove can cause your car to appear in a different place. This is not in actuality completely true. Someone at the check-in desk could look at the credit card you used to buy the glove, ask for ID, realize you are going to be occupied with golf for the next 3-5 hours, call their buddy at the DMV, find out your license plate and then call their car thief buddy who finds your car and steals it. That’s extremely unlikely but theoretically possible.
Anyway, what is mainly easy for humans is not that easy for AI systems. It might be configured in such a way that whenever anything changes, it needs to recheck everything. But occasionally, people are confused about the update problem as well. As AI becomes more ubiquitously integrated with the Internet of Things, our own models of what is related to what may well be as outmoded as an Ann Arbor driver. You believe putting something in your fridge cannot affect your alarm setting. And that is true for your “dumb” fridge. But what about a “smart” fridge? It might infer, based on your past behavior, that you typically eat leftovers for breakfast. Your home command center reads the bar codes on your leftovers and realizes it will take you an extra five minutes to consume the dinner-breakfast you brought home. So, it automatically changes your alarm to 6:55. Helpful? Even today, how many of us can really say for certain what the interactions might be among the remote controls and settings for the various components of our home entertainment systems?
Although humans are still much better than computer systems at solving the update problem, we still make errors. Here’s one I remember. We had a small workout room at NYNEX Science and Technology where I ran the Artificial Intelligence lab. In this small workout room was an ordinary wall clock. For years, I used the workout room at noon, and glanced at the clock to check the time. At one point, the equipment was moved around and I realized that the clock would be much easier to see on the opposite wall. So, I moved the clock to the opposite wall. I got on the treadmill and about ten minutes later glanced at the clock to check the time. Only I did not glance at the clock. I glanced at where the clock used to be. Think about that. I myself had moved the clock a few minutes earlier. Obviously, I “knew” where the clock was now positioned. And yet, I felt like a clueless Ann Arbor driver.
Another common sighting of the “invisibility cloak of expectation” came at IBM Watson Research Center. This is a place where Nobel Prize winners work. Anyway, the computer science department was housed for many years at an office building in Hawthorne. Restrooms were conveniently located near the stairwells on every floor. On three of the four floors, the men’s room was on the right. But on one of the floors, the women’s room was on the right. Whether the designers did this knowingly for a joke, I am not sure. But on the “odd” floor, men often wandered into the women’s room and women into the men’s room. Now, the doors for these restrooms were not marked in Kanji characters or ancient Greek. No, they were clearly marked in English. Although the computer science department consisted of people from all over the world, they all read English quite well. But expectations apparently trump perception. That seems to be the case for everyone some of the time and for some people nearly all the time regardless of intelligence or education. People very often see (or don’t see) based on expectations rather than the evidence of their senses.
Is there anything that can be done to help us remove our blinders and see what is really there? I think so, but it isn’t easy. The first line of defense is social. What do other people see? Chances are, if you were milling around in a park and suddenly everyone else starting running and screaming away from the swing set, you probably would too even if you saw nothing at all unusual. However, in the Mysterious Case of the Ann Arbor Stop Sign, people immediately interpreted the other driver’s behavior, not as another source of information, but as proof that the other person was a careless or demented driver. Not only did the drivers not see the “obvious” stop sign but they completely overlooked the possibility that they may have been wrong themselves.
This may be “human nature” but I suspect that aspect is exaggerated by an overly competitive school system and society. In school, we are molded to try to get good grades. Ideally, “grades” would not be so much about comparing people but about realizing what you still needed to learn. In society, we have perverted such intrinsically social and cooperative activities as dancing, cooking, singing, and dating into “contests.” At work, too often, a project failure results in finger-pointing rather than problem solving and prevention. Whatever the reason, it seems incontrovertible that people in our society are bunny-quick to blame others and tortoise-slow to blame themselves.
In The Walking People by Paula Underwood, she describes the “Iroquois Rule of Six.” This is a rule of thumb they use to avoid over-focusing on the very first explanation of behavior that springs into mind. Suppose you work for a large multi-national IT company and find yourself sitting alone in meeting room P-45. You glance at the clock. 10:10. You take out your calendar, whether paper or electronic, and re-read your note: Meet Joe, 10 am, P-45. Here it is 10:10 and he hasn’t shown up! It is natural to have some thought like this trounce through your head, “What he hell? What’s wrong with Joe? I guess he just doesn’t really care about our project!” Maybe. But the Iroquois Rule of Six might get you to consider at least five alternatives such as: 1. Maybe Joe is from a culture where 10:15 is “on time” for a 10 am meeting. 2. Maybe you wrote down the wrong room. 3. Maybe you wrote down the wrong time. 4. Maybe you wrote down the wrong date. 5. Maybe you are not actually in P-45. 6. Maybe the clock is wrong. 7. Maybe Joe cares about the project but is stuck in traffic. And so on. It isn’t so much that we human beings grab on to the first thing that pops into mind. The problem is that once we do grab onto an interpretation of events, we never let go! We don’t consider other possibilities.
My grade school friend Butch had had an uncle who had fought in the Pacific in WWII. He gave Butch this really cool book about how to survive off the land. One thing I read stuck with me. Monkeys are among the easiest wild animals to catch, not because they are stupid but because they are smart. One simple technique is to put two holes in a coconut shell and hollow it out as much as possible. Then, you slip a treat like a nut or small piece of fruit inside. The monkey comes along and grabs hold of the treat. Their hand, which went easily into the hole cannot get out while their fists are balled up holding the treat. So, you walk up to the monkey and club it and cook it and eat it. Monkeys are fast. It would be easy for the monkey to let go of the treat and scamper away. But they won’t. (At least, that is what the manual claimed). How much are we like the monkey? We grab at an explanation that makes us feel good and stick with it. We cannot let go. And we cannot accept the possibility that we ourselves might be wrong. Only in that last split second before the monkey’s skull is split open does it perhaps think, “Let go. Run. Too late.” Can we do better?
The United States, among other countries, has the intellectual capacity and the urgent need to quickly and fully develop new energy sources that are cheap, reliable, renewable, clean, and not dependent on foreign wars. And we are. In a trickle. But we are giving corporate welfare to old energy oil company kingpins because they are lavish campaign donors in a torrential river of cash. If you had a huge hole in your pocket that was draining all your cash, you’d see to fixing it quickly. But the oil drain isn’t so obvious. It steals far more of your money than a pickpocket could. But it’s well-hidden. Of course, at least until lately, oil money doesn’t come right out and say, “We know we’re rich but we deserve it. Give us more!” But we are so much in the habit of using non-renewable resources that we don’t think twice about it. And, those habits and expectations are played on plenty so that we are trained to think: “EPA- who needs it?” “Climate Change – unproven science”, “Solar and wind power are great but way off in the future”, “Pollution may cause cancer and asthma but that’s the price of civilization.”
The cheap oil prize that we so greedily grabbed hold of is now the trap that will get us killed, quite literally. It’s what we’ve been doing for many years. Why let go now? Instead, it’s easier to scream at others: “There is no stop sign here!!” Eventually of course, people change and civilizations change. But to change too slowly means you could be the cause of an accident; you glance on the wrong wall to see the time; you miss the tether ball on every cycle. Or, it could just mean complete annihilation. Maybe you could at least let go for a little while. Maybe you could at least let go with one hand. Maybe you could just forget the prize and the coconut and get away before it’s too late. I hope so.