Bitwise
Page 11
At this gathering of Google employees, greens overwhelmingly dominated, and I was one of them. There were smaller groups of blues and oranges, and maybe one or two golds. The organizers told us that every Google retreat ended up being mostly greens.
This classification is not science. It does capture something about the differences between people, but what it captures is arbitrarily selected and minimal. Who drew the dividing lines? The corporate consultants behind these sorts of tests claim they harken back to the ancient Greeks and Hippocrates’s classifications of the four humors, or temperaments. Some versions map the humors like this:
GREEN: phlegmatic (water, phlegm)
BLUE: sanguine (air, blood)
GOLD: melancholic (earth, black bile)
ORANGE: choleric (fire, yellow bile)
While the theory of the four humors continues to inspire artistic works like Paul Hindemith’s The Four Temperaments and Carl Nielsen’s second symphony, we no longer look to Hippocrates for medical guidance. Why look to him for personality classifications? Because humans like simple typologies, and fourfold classifications are easy to grasp. Division into two parts is too simple.*5 Dividing things into three makes symmetry impossible: dynamics reduce to two against one.*6 But four is aesthetically satisfying as well as flexible. A fourfold grouping can be carved up into pairings and oppositions, it can model symmetrical pairings (2 vs. 2) as well as unequal conflicts (3 vs. 1), and each of the four elements can have similarities with and differences from the other three. Hippocrates’s typology stuck in part because it was memorable, and folk theories worldwide likewise embrace fourfold classifications of everything from elements (earth, air, fire, water) to causes (material, formal, efficient, final) to Indian varnas and castes to Mencius’s virtues (benevolence [rén], righteousness [yì], wisdom [zhì], and propriety [lĭ]) to Buddhism’s four noble truths.*7
Leonhard Thurneysser’s alchemical depiction of the four humors in Quinta Essentia (1574)
The virtues of a fourfold classification, however, are all artifacts of the human imposition of arbitrary, simplifying order onto a far more complex reality. There is no reason why our cognitive preferences should lead us to anything resembling an accurate representation of personality types. But we easily grasp neat fourfold classifications, and so they stick. We can’t help ourselves. Humans are excellent at conjuring meaning. We inflate the most arbitrary of symbols (counting numbers, for example) with multitudinous and conflicting associations, then hold fast to them until a new system topples them. But our cognitive biases weigh heavily in the construction of these systems.
The surfeit of fourfold personality schemes circulating today mostly derives from the unscientific, speculative, yet remarkably successful efforts of two amateur psychologists named Myers and Briggs. The Myers-Briggs Type Indicator (MBTI) test has four binary axes that allow people to classify themselves into one of sixteen types (24). There is little to indicate that this has any scientific validity whatsoever, but the system, sometimes termed the thinking person’s astrology, undeniably appeals to a great many people. At least among my generation and demographic, I’ve seen people on dating sites, on Twitter, on Facebook, on Quora, and in casual conversation invoke their MBTI type as a shorthand for their personality: “Oh, I have to do it my own way because I’m an INFJ”; “Sorry I was such a jerk; I can’t help it, I’m an ENFP.” Surely something so reductive couldn’t indicate much substantive about a person, could it? So why is it that the typology is so popular? Why do people accept it, and what purpose does it serve to them? Businesses and career development organizations use MBTI and MBTI-like typologies to predict in which role a person will fare best: research, communications, support, etc. The MBTI publisher Consulting Psychologists Press, which markets and administers the personality test to corporations, universities, and governments, claims that over 80 percent of Fortune 500 companies assess their employees with the MBTI. Despite long-standing skepticism from professional psychologists, it can be hard to avoid being classified by the MBTI.
The MBTI itself has amassed surprising authority through a long, twisty history. Its simplicity and popularity make it the ideal case study for how and why people classify themselves—a matter that becomes a great deal more important once those classifications get fed into computer algorithms. Tests like the MBTI, which offer strict division and regimentation of our self-identifications, are one way in which we make ourselves more comprehensible to computers.
The Myers-Briggs Typology
In the 1940s, Katharine Cook Briggs and Isabel Briggs Myers started with four basic types, then leavened them with the personality theories of Carl Jung, as described in his 1921 book Psychological Types, in order to arrive at the four-axis classification. Jung’s work is not as pseudoscientific as astrology, which is based on a geocentric model of the universe and requires Pluto to be a planet no matter what the scientists say. But there is little to suggest that Jung’s work is rooted in anything beyond his casual observations, dubious intuitions, and free associations.*8 Briggs and Myers, who had little background in the social sciences, joined Jung’s work with their own anecdotal experience to generate their schema. In the 1950s, psychologist David Keirsey combined the Myers-Briggs model with four “temperaments,” to create the Keirsey Temperament Sorter. The two models differ in their details but share the same four axes and sixteen types. Most subsequent classifications have been derived from either one or both of them.
I first took an online Web 1.0 version of the test sometime in college and was not at all surprised to discover that I was a judgmental INTJ, also known as the Mastermind—or less charitably, the Crackpot. OkCupid’s Brutally Honest Personality Test, a takeoff on the MBTI, has this to say about the INTJ:
People hate you.
I mean, you’re pretty damn clever and you know it. You love to flaunt your potential. Heard the word “arrogant” lately? How about “jerk”? Or perhaps they only say that behind your back.
That’s right. I know I can say this cause you’re not going to cry. You’re not exactly the most emotional person. You’d rather spend time with your theoretical questions and abstract theories than with other people.
This description makes for a needed complement to the ego-inflating verbiage usually employed. A sampling:
INTJs are ambitious, self-confident, deliberate, long-range thinkers. INTJs are known as the “Systems Builders” of the types, perhaps in part because they possess the unusual trait combination of imagination and reliability. Whatever system an INTJ happens to be working on is for them the equivalent of a moral cause to an INFJ; both perfectionism and disregard for authority may come into play, as INTJs can be unsparing of both themselves and the others on the project.
They generally withhold strong emotion and do not like to waste time with what they consider irrational social rituals. This may cause non-INTJs to perceive them as distant and reserved.
Read enough of this bombast and one will surely start displaying the arrogance of the Crackpot. The INTJ sounds like a pretty great thing to be, but all the types are described in similarly glowing rhetoric. As long as the rhetoric for my type is a little more like me than the descriptions of the other types, I have some reason to believe I’m an INTJ. Perhaps I do have more in common with Stephen Hawking and Isaac Newton (whom Keirsey.com has determined to be INTJs) than with Elvis Presley or Ronald Reagan (declared by Keirsey.com to be the exact opposite of my type, ESFPs), but I’m not quite ready to see a shared essence between Isaac, Stephen, and me.
Personality classifiers are not a purely descriptive process. We also prescribe our own type when embracing any such model, whether it’s astrological signs, Myers-Briggs, or the Enneagram. If I lend credence to any of these models, then seeing myself as a Sagittarius, an INTJ, or a type 2 will grant me a point of reference and guide me in the future.*9 Even
telling people, “I think Myers-Briggs is nonsense, but in lieu of me describing my personality, I’ll admit that the INTJ is vaguely in the ballpark” will condition them to see me in that way. If the categories are flexible (we can’t change our birth month or astrological sign), I might doubt the classification, but simply knowing my classification will condition my experience. If the classification matches me reasonably well, I’ll likely be biased to think my actions fall within that framework. My future actions might serve to reinforce the classification further. It may not be a bad thing to identify as an INTJ—it could even be healthy if it supports a positive self-image and encourages productive non-crackpot activity. But that’s very different from believing it to be a scientific diagnosis, which it certainly is not.
Cliff Johnson’s 1987 puzzle game The Fool’s Errand, based on the iconography of the tarot
* * *
—
I discovered the tarot deck through Cliff Johnson’s wonderful and maddening 1987 computer puzzle game The Fool’s Errand. Several years later I read Italo Calvino’s The Castle of Crossed Destinies, which uses a two-dimensional layout of cards to construct stories by reading sequences of cards across and down.*10 The mythology of the cards intrigued me—as myths, not as magic. So rather than having a reading done, I picked the cards that I felt were most appropriate for me. If I was going to let my fate be guided by a mythology, I wanted to guide how I applied the mythology to myself. So I decided on the enigmatic Hanged Man, the melancholic Hermit, and the truthseeking Ace of Swords.*11
The Hanged Man, the Hermit, and the Ace of Swords, from a nineteenth-century Italian Minchiate deck
Is it cheating to pick tarot cards instead of drawing them at random? Only if you think the game itself is meaningful. I would argue that it’s far healthier to choose one’s cards instead of having them be dealt out from a shuffled deck. Classifications like the MBTI are closer to tarot than clinical diagnoses. They aren’t as random as tarot readings nor as contrived as picking tarot cards yourself. They exist somewhere in between those two approaches, not as any sort of diagnostic test. MBTI tests usually offer a couple dozen questions that slot test-takers into one of the sixteen types, but the shorter ones just contain four questions, one for each binary axis, allowing the test-taker to pick and choose. It’s not a diagnosis, it’s a selection.
In contrast, a clinical psychological test like the Minnesota Multiphasic Personality Inventory is constructed by normalizing around a sample population. In its classic incarnation, the MMPI-2, the MMPI offers over five hundred yes/no questions, including some duplicates, that assess the test-taker’s personality on dozens of individual axes and around ten synthetic scales, including hypochondria, gender presentation, depression, and paranoia. It also includes “validity scales,” which measure the honesty and consistency of the test-taker.
I took the MMPI when I was in my early teens and registered as normal on most scales save for one—depression. This came as a relief, as depression was one of the more innocuous scales next to mania, paranoia, schizophrenia, and the like. I was told I had been very honest on the test, which I suppose I had been. The test, I concluded, had been made for people exhibiting more severe indications of mental and emotional maladjustment than myself, and so I hadn’t had much cause not to be honest in the first place. But the MMPI appeared empirical in a way that Myers-Briggs did not: it was assessing me relative to a chosen sample population that exhibited certain personality traits. That’s not to say the MMPI is flawless: the normative component of such a test carries with it some serious complications, though far subtler than those of Myers-Briggs tests. Unlike the MMPI, Myers-Briggs never calibrated responses. It presents its dichotomies independent of any empirical basis and demands that you throw yourself into them.
The upshot is that Myers-Briggs and typologies like it are inherently self-classifications. To embrace a test like the Myers-Briggs is, in effect, to say, “Here is how I wish to see myself!” and “Here is how I want to be seen!” Anyone dissatisfied with their Myers-Briggs result should have little difficulty in skewing a subsequent test to obtain a more satisfying result. This is cheating only in the sense that it’s cheating to look in the mirror to see if you missed a spot while shaving. As with astrology, the choice to use Myers-Briggs is more meaningful than any specific result. We care a lot about what classifications we put ourselves into.
Bitwise and Byte Foolish
I will not be pushed, filed, stamped, indexed, briefed, debriefed, or numbered. My life is my own.
—No. 6, The Prisoner
I dwell on Myers-Briggs for three reasons. First, it’s awfully popular. It or some variation of it is used in corporations and training across the country as a productivity tool and a means for workers to better understand themselves. Myers-Briggs shapes a substantial portion of social interactions. Second, Myers-Briggs serves as a simple and relatively non-loaded example of the sort of classifications that happen in far more controversial and high-stakes arenas: gender, race, class, sexuality, political affiliation, diseases and afflictions mental and physical, and criminality.
Finally, classifications like Myers-Briggs are extremely computer-friendly. If we accept the framework provided by the MBTI, I can describe my personality to you on the fingers of one hand, each finger representing one axis, extended or retracted. By seeing which fingers are extended, you will know who I am. In the Myers-Briggs schema, a person’s personality type can be represented with a mere four bits of data; that is enviable compression for a description of a personality. (One’s astrological sign requires four bits as well.) When computers are introduced, they work with these sorts of typologies to classify data. Here’s how I could encode a person’s Myers-Briggs type in C++:
struct {
bool fExtrovert;
bool fIntuitive;
bool fThinking;
bool fJudging;
}
Each Boolean variable above is set to either true or false. A true indicates that a person’s type is that of the name of the variable. A false indicates that a person’s type is the opposite of the name of the variable. It’s only four bits of information, but each variable occupies at least one entire byte (a byte is almost always eight bits long, and the smallest, atomic unit of computer storage), so this is at least thirty-two bits of storage here. I can improve on this by using a single byte-length variable (the one-byte char, the size necessary to store a single alphanumeric character), with each particular bit designating one of the Boolean flags. It would look like this:
#define MBTI_EXTROVERT 0x01 // (or 1)
#define MBTI_SENSING 0x02 // (or 1 << 1)
#define MBTI_THINKING 0x04 // (or 1 << 2)
#define MBTI_JUDGING 0x08 // (or 1 << 3)
unsigned char MBTIType;
So for an ESTJ, all the bits are on, yielding the binary number 1111. For an INFP, all the bits are off, yielding binary 0 000 (or just plain 0). For my type INTJ, the last two bits are on and the first two are off: 0011. The variable MBTIType still has bits left unused after the first four, but we’ll leave that be.
If I want to assign and read the type in one go, I could enumerate all sixteen types and assign them consecutive numbers beginning with 0, all the way up to 15. This is aptly called an enumerator:
enum MBTIType {
INFP = 0,
INFJ = 1,
INTP = 2,
INTJ = 3,
ISFP = 4,
ISFJ = 5,
ISTP = 6,
ISTJ = 7,
ENFP = 8,
ENFJ = 9,
ENTP = 10,
ENTJ = 11,
ESFP = 12,
ESFJ = 13,
ESTP = 14,
ESTJ = 15
};
What you may not have noticed—and wha
t computers certainly would notice—is that these last two representations, the enumerator and the four-bit variable, are identical. I have ordered the enumerated type such that the value of any given MBTI type is the same as that which you would get by putting the respective bits together in the four-bit representation. The enumerator contains four bits of information. INFP is 0, just as it was when all four bits were off in the preceding example and we obtained binary 0000. ESTJ is 15, which translates to binary 1111. (Instead of powers of 10 for places, as in decimal, binary uses powers of two, the first four being 1, 2, 4, and 8. 1 + 2 + 4 + 8 = 15.) INTJ is 3, which translates to binary 0011 (1 + 2).
For the computer, these representations are identical, because in machine language, the program works exclusively with numbers. The code that I write to compare the value of an MBTI variable against the code for INTJ is comparing that variable against a value of 3 (or 0011, more accurately, since it’s all just 1s and 0s at the bottom). Whatever meaning I gave it, even in the code, has disappeared by the time the CPU is executing the code. Numbers are truly meaningless to computers.
The reduction of all data to numbers (bits, specifically) makes the computer a truly neutral arbiter of data. While we may enjoy playing around with personality types and going to corporate retreats, there’s something far more unsettling about seeing a personality type coded into a fixed computational representation and having that representation stored indefinitely on a centralized server. “Hello, I’m an INTJ!” doesn’t carry with it the sense of being indelibly marked, but having a centralized database (whether corporate or governmental) store a 3 designating me as an INTJ certainly does. The computer does not understand what it means to be an INTJ, but we do. And once that 3 is in there, the issues of who controls the value, who sees it, who gives it meaning, and how it is used become significant far beyond the limits of the machine’s memory. Switching up my Myers-Briggs type should not be as difficult a process as changing my legal name, but as we lose personal control of our data, we also lose part of our ability to define and redefine ourselves.