The Happiness of Pursuit: What Neuroscience Can Teach Us About the Good Life (5 page)

Read The Happiness of Pursuit: What Neuroscience Can Teach Us About the Good Life Online

Authors: Shimon Edelman

BOOK: The Happiness of Pursuit: What Neuroscience Can Teach Us About the Good Life
9.07Mb size Format: txt, pdf, ePub
Three Things Everyone Should Know About Life, the Universe, and Everything
 
The driving force and the arbiter of innovation in biology is evolution, and so a species that invests resources in representing its environment will thrive only if computing with its representations makes it, in the final account, a better replicator. The key question that needs to be asked about brain computation is therefore one of utility: what survival and reproduction advantages does having minds—which is what brains compute—confer on animals?
It may seem that the answer to this question would depend a lot on the animal, on where it lives, and on the company it keeps—but it does not. Instead, three basic facts of life, the universe, and everything join forces in creating a simple and general explanation for the myriad ways in which mindful hominids, squirrels, chickadees, or octopuses gain an upper hand, paw, foot, or tentacle over happy-go-lucky ones.
The first basic fact is about the universe. It so happens that the universe we inhabit has built into it an asymmetry between the past and the future. The arrow of time, which is what this asymmetry is often called, emerges independently from the action of several distinct families of laws of physics. Among these are the laws of thermodynamics, which govern the fundamentals of self-organization, stability, and self-replication—in other words, the origins, maintenance, and propagation of life.
The predicament of being alive is what the second basic fact is about. Life is fragile and time is irreversible. An antelope that sees its life passing in front of it as it is being brought down by a lion cannot roll the nature documentary in which it is starring back to that critical moment, only seconds before, when it ignored the flicker of movement in the tall grass. An animal that gets eaten or dies of starvation or is run over by a car or sets itself on fire by playing with matches is gone forever. The only effective insurance against such outcomes is to recognize the threats before it is too late.
Threats are always about the future, whereas the experience on which we can base our actions is all in the past. What is a poor animal to do? Seeing that antelopes, lions, and humans are still around, it must be possible to evade or otherwise thwart at least some impending disasters. This brings us to the third basic fact, which is about everything—that is, everything that’s relevant to the life of an animal. The fact is this: the future state of affairs in an ecological niche is predictable from its past, up to a point.
The reason for this predictability is the sheer physical inertia of the universe. On an appropriately short time scale, things are guaranteed to stay as they are or to carry on changing in the same manner as they did before. There are also many kinds of longer-term regularities, such as cycles of seasons. Animals can evolve to rely upon seasonal changes in the environment and to anticipate them from telltale cues (think of migratory birds that respond to the first frost). Or, animals can evolve sophisticated brains that represent patterns of change in the environment and anticipate the future by treating it as a statistically projected extension of the past. Either way, it is the capacity for
forethought
that distinguishes, on the average, between the quick and the dead.
What are brains good for, then, and what preeminence doth a beast with a better brain have over others? As I promised, the answer is short and simple: it’s all about forethought. This conclusion hinges on the three key facts about life, the universe, and everything. Forethought is meaningful, important, and practical because time is directional, because it pays to anticipate the future, and because it is possible to do so by consulting the past.
The capacity for forethought or foresight sounds so advanced that attributing it to beasts may seem like a big stretch. In ancient Greek beliefs, forethought elevated men above beasts. According to the creation myth retold by Hesiod, Prometheus, the Titan whose name
means
“forethought,” stole fire from Zeus and gave it, along with the ability to reason and sundry other survival skills, to men (but not to women, whom Zeus created later, to balance the good brought to the world of men by Prometheus). Israelite sages held foresight in even greater esteem: the Talmud asks rhetorically, “Whosoever is wise?” and replies, “The one who sees that which is about to be born.” (In the spirit of the times, those wise men too would have laughed you out of the shul for suggesting that women have sense.)
7
In reality, forethought is found in the lowliest places on earth, and even underneath it. My favorite example illustrating that you don’t have to be male, human, or even sighted to possess a modicum of foresight involves blind mole rats (of either sex). If someone digs a ditch overnight across the path to your front door, you will most likely see it in time to avoid falling in. For that, you need only exercise foresight literally and passively, by simply looking ahead. In comparison, the humble mole rat,
Spalax ehrenbergi
, exercises foresight metaphorically (necessarily, because it is blind) and actively. Here’s how it works: a burrowing mole rat bangs its head against the ceiling of the tunnel and analyzes the returning echo for signs of obstacles such as rocks or ditches.
8
This feat of subterranean echolocation was discovered only recently, in part because researchers have tried for years to trap mole rats for study by waiting for them to fall into freshly dug ditches. It does quite raise the ante on the human bid for sophistication in seeing the shape of things to come. How are we to make sense of it? It so happens that foresight and forethought are evolvable. Having a better brain—one that computes a sharper mind—can make the life of its owner longer, or at least long enough for it to mate. Thus, evolution generally favors creatures that are being mindful about the future—whether by casting their brains about or by banging their head against the ceiling of the burrow—over the happy-go-lucky ones that are oblivious to what the future may bring to their piece of turf.
9
Promethean Probabilities and Amazing Bayes
 
Forethought is a wonderfully versatile tool: when properly tuned and regularly consulted, it safeguards your life and liberty, whether you are a burrowing rodent or a homeowner whose front lawn has been replaced overnight by a newly dug ditch. It also promotes the pursuit of happiness, both in a big way (after all, evolution hones it to catalyze mating, which is often fun) and in various small ways, as when a grocery shopper chooses the checkout line that he or she projects to result in the shortest and therefore happiest waiting experience. Given that the future is predictable to the extent that the world exhibits statistical regularities, how do animal brains compute it? The same way statisticians do it, in all likelihood.
Brains end up resorting to statistics for the same two reasons that governments and corporate managers do so: first, because there is always more potentially relevant raw data available than can be effectively grasped; and second, because the available data are uncertain. Amassing the kind of descriptive statistics usually reported in popular media, such as the means of some quantities of interest (say, the mean income per capita), does not address the really interesting questions arising from the data, namely, what it could all possibly mean and where things may be heading. Thus, it is the second reason that makes statistics vitally important.
Unlike descriptive statistics, which at best merely quantifies uncertainty,
inferential statistics
puts it to work by processing available data in search of trends or patterns on which predictions about future outcomes can be based. Although a hypothesis-driven search of data for patterns is a standard operating procedure of any empirical science, you don’t have to be a scientist to make it happen. In fact, just as with foresight, you don’t even have to be human.
Consider the case of an adolescent blind mole rat, whom I shall call Molly (not her real name). Molly has just decided to reject the entrenched ways of her parents and is about to start digging at right angles to her home tunnel. What does she expect to find there? Soft earth, a rock face, a ditch? Her beliefs in this matter are simple and undiscriminating: she holds these three possibilities to be equally likely.
Molly may be blind and inexperienced, but she is neither senseless nor stupid. Rather than clinging obstinately to the a priori hypothesis, she performs, as she digs on, occasional experiments to gather new data (by banging her head against the tunnel ceiling, of course—the one useful skill she admits to having inherited from her family). She processes the echo from each head-slam to glean some information about the way ahead.
As she adds this new information to the existing pool of data, one of the possibilities grows more likely and the others less so. Soon little doubt is left that the way ahead leads into an open space, but Molly wants to feel it with her own whiskers. After digging a bit further, she feels on her face the draft of fresh air that all mole rats instinctively shun. With a shrug and a sigh, Molly turns back.
Let us see how Molly’s actions allow her to learn from experience and to become better at using her echo sense to feel her way ahead. The quantities of interest are the relative likelihoods of the various possible outcomes (ditch, rock, earth). These relative likelihoods are simply the ratios of the corresponding subjective probabilities; we may think of these probabilities as expressing Molly’s beliefs about various possible states of affairs in her subterranean world. The prior probabilities are all the same (one third each for ditch, rock, earth), and so the likelihood ratios are all equal to one.
Once new echo information from a head-banging experiment comes in, it must be taken into account. What Molly needs to compute, then, is the
conditional probability
of each of the possible outcomes: the probability of ditch
given
the particulars of the echo (and likewise for rock and earth). Intuitively, this conditional probability expresses the degree to which a belief (that what lies ahead is ditch) must be modified by additional information (that what lies ahead returns a particular kind of echo). How can it be computed?
It turns out that the unknown conditional probability that Molly is after is proportional to the product of two known quantities: the prior probability of ditch and the conditional probability of the echo given ditch. The first factor is easy (it is simply the belief that Molly holds before the new data are in), but what about the second one? It can be estimated if Molly keeps track of actual outcomes that follow each kind of echo. For instance, an encounter with a ditch that follows the experience of a particular echo can contribute to the estimate of the probability of that echo, given that there is a ditch there; and while she’s at it, Molly can also update her cumulative estimate of the prior (unconditional) probability of ditch.
The mathematical foundations for updating subjective probabilities and for managing uncertainty by incorporating new empirical data into a classification or decision procedure were laid by an eighteenth-century amateur statistician, Thomas Bayes. (This makes Molly a Bayesian, even though she does not know it.) The key contribution on the part of Bayes was a theorem that expresses the so-called posterior probability (posterior to the experiment, that is; in our example this is the conditional probability of ditch given echo) in terms of the product of the prior (the probability of ditch before the head-bang) and the so-called data likelihood term (the conditional probability of the echo given ditch).
10
I decided against including the proof of the Bayes Theorem here, even though it is only three lines long. The received wisdom is that every equation in a book drives away half of the readers, and I do not want to risk losing seven-eighths of my readership. (Come to think of it, this disclaimer alone may be too quantitative for its own good.) I’ll just go on treating the Bayes Theorem as a miracle then.
Considering how useful Bayesian statistics is all across cognition, it certainly deserves to be called a miracle. In perception, the Bayesian formula for posterior probability comes in handy for updating the representation of the environment given the sensory data. In decision making, it is indispensable for taking into account new evidence while choosing the best outcome. In action, it is used for planning a sequence of motor commands that combine the current goal with prior experience.
All these cognitive functions are implicitly included in the example of Molly the mole rat, and all of them ultimately serve the supreme goal of cognition: forethought, which we are now in a position to understand as processing past experience so as to be able to deal more effectively with the future. Because prediction is tricky (especially, they say, about the future), principled management of uncertainty is absolutely crucial in this undertaking. The Bayesian framework for statistical inference and decision making is exceptionally well suited for meeting this challenge. As we will see next, it is also well suited for being implemented with the standard building blocks of brains: nerve cells, or neurons.
Minds Within Brains
 
Having found out that minds are for forethought and that forethought can be distilled from experience using statistical inference, we can now begin to understand how brains compute minds. As we know, brains compute useful things for their owners by harboring representations of the world. Generally speaking, they do it by maintaining
internal states
, which stand for various aspects of the external world in a consistent manner that captures their causal structure. The state of a brain is a fleeting, dynamical thing: being simply the activities of all the neurons of which the brain consists, the state changes from one instant to the next, driven by its past history and by external inputs.

Other books

The Strangers of Kindness by Terry Hickman
Playground by Jennifer Saginor
Trainstop by Barbara Lehman
Right Where I Belong by Krista McGee
Brown, Dale - Independent 04 by Storming Heaven (v1.1)
The Reunion by Jennifer Haymore
The Locket by Elise Koepke