by Rafe Sagarin
The problem with such associative learning is that it only deals with a specific threat. After the rattlesnake training, our daughters, who were themselves learning about the dangers of desert living, asked, “What about scorpions? Do we have to take the dogs to scorpion training now, too?”
Much of our security systems seemed based on associative learning, and this leaves us vulnerable to an enemy that can learn. For example, after the anthrax attacks of 2001 we developed an association between U.S. mail and terrorism, and we responded not unlike my frightened dogs. Years later we still irradiate mail bound for the Capitol and greatly restrict uses of curbside mailboxes. Yet it’s not hard to figure out how to deliver a letter bomb using the U.S. Postal Service. I learned last fall when I tried to send a package from Santa Barbara to Los Angeles, a distance of 96 miles. Three weeks after I dropped the package into a curbside mailbox, it appeared in my former office at Duke University in North Carolina, 2,613 miles away. A sticker affixed to it told me that the package had been returned to the sender because new security measures required any package over 16 ounces to be handed directly to a postal worker. For me, the sticker on the package represented a minor annoyance. To a terrorist, it would read like a schoolhouse primer for creating a disaster. Just put your target’s address in the return address field and the post office will gladly deliver it. Best of all, if you are feeling really cheap, you don’t even have to put postage on it.
A high level of punishment (an electric shock, a terrorist attack) or reward (hitting a jackpot on a slot machine) leads to quick retention of associative learning, but it’s not necessarily adaptable learning. Learning to adapt is a much more complex process that takes place across a number of planes—individual or organizational, passive or active, through the ritualized transmission of knowledge or as an ongoing experience in a changing world. Under this broad view of learning it appears to be a force of nature, like gravity. That is, the transmission of information, and the resulting alteration of interaction with the environment by the receiver of that information, occurs in nature no matter what you do. It may be encoded in DNA, passed on from a mother sea otter to her pup, or written in an al-Qaeda field manual, but learning is at the heart of adaptation.
Viewing learning as a force of nature leads us along the same path that we take when we recognize risk as ubiquitous in nature—whether we chose to learn well or not, our enemies will learn, so we’d better learn well and understand how learning relates to adaptation. The postal security measures fail to account for the fact that learning in nature sets off a continual process of escalating threats and adaptive defenses.
Humans tend to get a bit provincial about the subject of learning. We tend to assume that beyond some silly experiments with mice in a maze, there isn’t too much we can learn from the non-human world about learning because we are the undisputed world champions of learning. Moreover, one might argue that the vast majority of us, certainly those of us reading this book, do not live in the tooth-and-claw world of animal struggles for daily survival, so the kind of things that we assume animals need to learn are not that relevant to us. I would argue, first, that while we are remarkable learners, we are not that different from many animals in our relationship between learning and adapting to a changing world, and second that it is the long-term process of adaptable learning—conducted throughout a lifetime and across generations in all organisms—from which we can learn richly from nature.
Almost all modes of learning in humans have been identified in other biological organisms. Tool-making was once considered the province of humans alone, but then chimpanzees were discovered making specialized tools to fish out termites from rotting logs. Then those who would separate humans from other animals moved the goal line and suggested that deliberately making tools for future use (rather than just to extract a tasty treat or deal with a threat near at hand) was the barrier. But a chimp named Santino in Sweden’s Furuvik Zoo methodically breaks apart the cement in his enclosure, calmly stacks the pieces in a special place, and only later, when the visitors arrive, does he use these tools by angrily hurling them at the surprised tourists.1 Critics may take solace in the fact that this occurs in an animal that is extremely closely related to humans, but now we have film of octopuses, which are about as genetically close to us as a garden snail, collecting different halves of coconuts when they are not threatened to use in constructing a rolling suit of armor when they subsequently become threatened.
Another supposed barrier between human and non-human learning was confidently determined to be the sense of “self ” supposedly unique to humans. But primates in captive learning environments have clearly shattered that barrier. And the Indonesian mimic octopus, which was only discovered in 1998, can quickly mimic, in both appearance and behavior, at least a dozen other animals like flatfish, sea snakes, and deadly lion fish.2 Does its ability to change itself into something else entirely indicate it has a sense of itself as something that must be transformed to appear like those other selves? We’ll probably never know for sure, but the more we look, the less substance we find to the categorical distinctions between humans and non-humans.
The adaptable learning process at its most fundamental level is driven by natural selection as envisioned by Darwin. The basic tenets of Darwinian evolution are threefold and quite simple. You need variation (for example, two birds are hatched to the same parents—one is solid yellow and one is yellow with mottled gray spots); you need a selective force that literally favors one type of variation (a predatory hawk that easily spots the yellow chick but not the mottled one); and that selection needs to punish the inferior variant (the solid yellow bird dies) or reward the superior variant (the mottled bird grows up and reproduces). This process is not the only way that new species arise and adapt,3 but it is an extremely important part of the ultimate domination of a once-lifeless planet by a stunning diversity of life. This chapter is about how we use and misuse selective pressures to learn or not learn lessons from our changing security environment.
Learning is intricately related to adaptability in both the process of how it comes about and in the results of the learning process. One of the fundamental properties of selective learning in animals is learning to recognize signals from the environment such as the color of the sweetest flowers, the appearance of edible versus nonedible prey, or the songs of potential mates. This learning is selective because those organisms that simply can’t identify what is good to eat, what is good to avoid, and who is good to mate with will die out or fail to pass on their genes.
But simply recognizing these signals is not enough, because it doesn’t account for the ubiquitous variation and change in nature. An organism that can only identify a particular signal will starve or get eaten or fail to mate if the critical signals related to feeding, predation, and mating change. So organisms critically learn to distinguish small differences between signals—the sweeter sweetness of a deeper red flower, the most palatable prey among a herd, or the more robust call of a stronger male mate. And remarkably, it is from the development of this finely tuned discriminatory ability that animals learn the most important skill of all—the ability to generalize their sensory abilities and be able to respond to novel stimuli from environments they have not yet experienced.4 In other words, through the process of learning in detail about everyday threats and opportunities, organisms gain an adaptive capability that serves them in responding to future unknown threats.
Even the process of how animals learn is not immutable. Darwin, who fought against his contemporaries’ notion that animal species were immutable—unchanging since the dawn of creation—would not be surprised to read recent studies demonstrating that within lifetimes and across generations, animals can learn to learn differently. That is, animals have some basic capacity for learning, but they can learn in accelerated ways depending on the environment they are put in. Monkeys, which are generally considered to have the learning capacity of a human two-year-old, can be trained in experimenta
l settings to learn like a nine-year-old, including understanding a sense of their own self as a unique entity interacting with and affecting the world around them.5
The animal world thus tells us that learning is itself a selective adaptation; that learning to discriminate variation in the immediate world helps to build an adaptive capacity to respond to novel threats; and that the process of learning can change, given the right conditions. This powerful suite of properties should be seen as a warning, reminding us that no security adaptation should be assumed to be a safe everlasting solution, because there is always the potential for an adaptable enemy to learn how to overcome it.
Nonetheless, the degree of difference between humans and others is sometimes impressive and bears special attention here because human learning plays such an important role in protecting, and threatening, our security. Humans possess a wide variety of learning abilities, and they learn fast, especially when working with or against other humans.
HOW HUMANS LEARN
Michael Kenney, who has extensively studied the ongoing battles between narco-traffickers and international law enforcement efforts in Colombia and has applied his knowledge to the battle between terrorists and counter-terrorists, believes that learning is the key mechanism for adaptability on both sides of these battles. The different ways in which learning occurs among individuals and organizations is the most telling indicator of how well they adapt to changing environments and to the activities of their adversaries.
Kenney uses the Greek terms techne and mētis to differentiate two different modes of learning. Techne refers to formalized knowledge—facts, figures, techniques, plans, and other data that can be conveyed through lectures, field manuals, and other academic training. Mētis, by contrast, refers to knowledge that is gained through experience and interaction, learning through failures and successes during day-to-day operations, and spread through networks of practice. Kenney finds that while both techne and mētis are essential forms of knowledge transfer, successful criminal or law enforcement agents primarily rely on mētis to maintain the capability to adapt to continually changing environments:Like drug trafficking, counter-drug law enforcement is a mētis-based activity that requires improvisation and responsiveness if it is to be successful. Effective agents often supplement their formal training with practical know-how and intuition that comes from performing their activities repeatedly in local environments. Like the narcos they seek to apprehend, law enforcement “narcs” cultivate mētis in carrying out investigations and adapting their activities in response to unanticipated circumstances.6
But where do these advanced forms of technical and experiential learning derive from? Are we born with them, or do we learn to learn? Alison Gopnik, a psychology professor at UC Berkeley who studies childhood cognitive development, argues that while both technical learning and complex experiential learning are made possible by the brain morphology and physiology that all humans are born with, they only attain their true potential through an adaptable learning process that begins in the rapidly changing brain of an infant. According to Gopnik, one of the signatures of human learning is the extreme capacity for change that accompanies every level of human development. In contrast to the popular belief that babies possess incomplete or quasi-defective adult minds, Gopnik argues that babies have a fully functioning, but radically different, mind than that of a toddler, a teenager, or an adult. This mind is protected—by a heavy degree of parental care—from having to do all the everyday things like hunting, gathering, and balancing checkbooks that adult minds do, and instead it is hard at work mapping how the world works and developing adaptable learning strategies that will allow it to respond to completely novel experiences it will ultimately face. Other species that have extended periods of parental care—crows, for example—also show adaptive capacities to learn, but no species spends as much time caring for its young—and thus allowing adaptable learning systems to develop—as humans.7
As babies grow up and catalog more experiences, the human mind increases its store of memories. But memories aren’t just nostalgic photo albums of the mind. They help us quickly understand the present and even predict the future. Our memories are essentially chunked as packages of related observations. These chunks aren’t held in a specific address in the brain but in a networked set of neurons that fire in a certain pattern when the memory is invoked.8 These little programs form recognizable patterns that can be recalled even in completely new situations. A chess Grand Master has on the order of 50,000 positions he can instantly recall, and he can also have the ability to remember significant setups—the moves that deliver players to that particular position.9 We now have computers, which through sheer computational muscle and through being dedicated solely to the process of playing chess, can competitively play against Grand Masters, but they don’t play like Grand Masters.
A computer can grind out calculations, but it can’t predict the future. However, experienced observers who have chunked their memories can essentially do this. Before he was famous for writing books like The Tipping Point and Blink, Malcolm Gladwell wrote an article for The New Yorker on “Physical Geniuses,” people like the great hockey player Wayne Gretzky and the cellist Yo-Yo Ma.10 What Gladwell found was that these people put in an enormous amount of time practicing their craft, and all that practice gave them a huge storehouse of chunked memories that could be recalled consciously or subconsciously and used, even to predict the future. When Gretzky played hockey, he had an uncanny ability to make the perfect pass to set up a goal (although Gretzky holds the all-time NHL records for both goals and assists, he made nearly twice as many assists as goals). Gladwell showed Gretzky a film of one of his perfect passes that set up a goal and asked The Great One to explain how he knew his teammate would be in the right place. Gretzky struggled to explain it, but basically acknowledged that given all the conditions and positions of the players on the ice right then, he knew his teammate would be where he was sending the puck. It is difficult to describe this kind of learning consciously with words, both because chunked memories live in the province of the subconscious—we use them before our conscious mind even knows we’re using them—and because we generally don’t feel too comfortable talking about an ability to predict the future.
ORGANIZATIONAL LEARNING
If individual humans can learn in complex ways, it would seem to follow that organizations created by and composed of individual humans would also have the capacity to learn. In nature, learning occurs at multiple levels of organization. An immune system within an individual organism learns each time it interacts with a pathogen. Individual organisms then learn through a combination of their inherited genetic code and changes they make throughout their lifetimes in response to changes in their environment. Their individual learning may then impart a higher-level learning as the species they belong to becomes enriched in or depleted of individuals that operate a certain way. This multi-level aspect of learning suggests a tight correspondence to individual and group learning in society. Just as species “learn” from the collective successes and failures of lower orders of biological organizations, it would seem that organizations should learn from the actions of their individual members.
Yet we often decry how societal organizations and bureaucracies don’t seem to learn any lessons. When high gas prices followed by a collapsing economy led American car buyers to abandon the once popular and enormous gas-guzzling SUV trucks and giant automakers like GM were brought to the point of needing a massive government bailout, people wondered if they had learned anything from the gas crisis of the 1970s, when smaller fuel-efficient Japanese cars began to flood U.S. markets and gained a permanent stronghold. Organizational inertia in government seems to prevent us from responding even to known threats, such as terrorists’ clear intentions to do harm to American targets, until a catastrophe occurs.11 But there are also many examples of nimble organizations that appear to learn from their own past as well as from the experiences of other organizations, and
there is a strong interest in understanding how they do it. What makes an organization learn or not learn is the stuff of CEO and shareholders’ dreams and nightmares, and accordingly, a vast popular and academic literature has arisen around the topic of how and if organizations learn.
At first blush, much of the organizational learning literature seems to take pieces from a biological playbook, and sometimes it comes tantalizingly close to identifying a kind of organizational learning that is essentially biological. Most of the papers on organizational learning that I’ve come across take some riff on the variation theme, noting that the business environment is constantly changing and businesses must change their practices accordingly. This literature also generally acknowledges that some force of selection—typically in the form of enlightened managers (presumably those who suffered through that particular article or organizational learning seminar) who must screen employees and departments to make sure that they are learning—is necessary to separate out learning organizations from those that don’t learn. And a subset of the literature also pays deference to replication of successful variants—noting that rewards to departments and individuals for learning effectively are necessary to keep an organization learning in the future.
Yet the literature quickly strays from the organic world into theoretical constructions and flow charts full of various boxes representing entities like “organizational routines” linked by arrows representing actions like “organizational double loop learning”12 that supposedly map out how organizations should optimally learn.