From modest beginnings, Binet’s and Galton’s informed guesswork evolved toward improvements in methods to measure intelligence and theories to explicate its nature. Still, not everyone was satisfied. The early theories grew out of the psychometric tradition, which meant that they were based on tests and test scores. These theories were tidy and precise—perhaps excessively so. They seemed firm to a degree that could blind scholars to alternative aspects of intelligence not captured by batteries of cognitive tests.
Above all, one broad criticism loomed: Most theories of intelligence seemed excessively narrow. The charge of narrowness can be pinned down further to say that psychometric theories of intelligence were overly analytic. Even such “broad” factors as fluid intelligence and crystallized intelligence were narrow in the sense that they were collectively logical, rational, and convergent. In one way or another, every psychometric factor seemed to highlight the mind’s ability to perceive the elements and rules of a problem, and to focus the intellect to produce a “correct” answer.
Important as these skills are, do they really exhaust all the ways that the human mind expresses intelligence? And if the human mind is smart in other ways, what are they? In fact, it’s not too difficult to generate a short list of contenders. Creativity, for example, is obviously neglected. Also marginalized is the ability to deal effectively with other people—what is sometimes called social intelligence. By some accounts, social skills should be regarded as an important expression of intelligence. Yet another possible form of intelligence is the ability to understand oneself. After all, the person who lacks insight into his or her own preferences, hopes, dreams, and the kinds of experiences that prompt fulfillment is seriously limited. That person’s ability to think and act intelligently will to some degree be compromised.
This chapter traces scholars’ explorations into areas of knowledge, insight, and skill that might potentially fall under the umbrella of intelligence. Their research efforts widen the scope of how the mind undergirds effective thought and action. New theories of intelligence are good for two reasons. First, any attempt to cultivate intelligence is only as good as its underlying theory, and theories are always imperfect. We never want to conclude that our current theory is a finished work, true for all time. Second, if research can deepen our understanding of any field—science, medicine, economics, or intelligence—then the application of that knowledge will ipso facto be more powerful.
However, a caution is in order. The theories described in this chapter are emergent rather than modern. That’s because modern implies the most current and up-to-date. That’s not the right way to look at these theories. Collectively, they are works in progress—interesting, provocative, and promising, but as yet proven. They are emergent pockets of theoretical exploration that might eventually lead to revisions of classic theories by supplementing, revising, or perhaps even replacing the older, established ways of thinking about intelligence. As areas of exploration, the emergent theories deserve our attention not only because they are important, but also because each is fascinating in its own right.
Among the challenges to traditional ways of thinking about human intelligence, none has had a greater popular impact than Howard Gardner’s theory of multiple intelligences. In his book Frames of Mind, Gardner identified seven intelligences: logical-mathematical, linguistic, spatial, musical, bodily-kinesthetic, interpersonal, and intrapersonal.1, 2 Over the past several decades, Gardner’s theory has had a sweeping influence among educators. Indeed, the term multiple intelligences is commonly recognizable not only to teachers, but also to the broad public.
The most distinctive feature of multiple intelligences theory is that it recognizes not a single intelligence, but rather several independent intelligences. The plural is significant. With the addition of a single letter, s, Gardner repudiated the existence of a superordinate general ability that influences every cognitive act. Instead, Gardner argued that intelligent thought and behavior are products of relatively distinct abilities. Gardner’s proposed intelligences are a mix of the familiar and the innovative. Three of the intelligences—logical-mathematical, verbal, and spatial—correspond directly to factors psychometricians have recognized for many decades. As elements of Gardner’s theory, they are not particularly new except in one important regard: Gardner held these expressions of intellect to be autonomous, meaning that each operates independently, untethered to a common general intelligence that binds them together.
Gardner’s theory builds upon multiple strands of evidence, not just those favored by psychometricians—test scores, correlations, and factor analysis.3 For example, Gardner took into account cultural expressions of intelligence. Some cultures prize highly skilled performances in dance, and so establish training practices to develop dancing ability to high level. Such valuation of dance within some cultures constitutes evidence (anthropological, not psychometric) of bodily-kinesthetic intelligence. Gardner also used other forms of evidence to justify his theory. He noted for example, that several candidate “intelligences” were associated with particular symbol systems. Symbol systems include written text for linguistic intelligence, numerical symbols for mathematical intelligence, and notes and scores for musical intelligence. Yet another strand of evidence comprised clinical studies of brain injuries that impaired a person’s cognitive ability in a circumscribed way. For example, damage to the lower left hemisphere often results in aphasia, an inability to produce or interpret speech. Damage to counterpart regions of the right hemisphere can disrupt a person’s ability to produce or interpret music. If localized brain damage results in specific cognitive impairments, Gardner took such cases as evidence that separate and distinguishable intelligences compose the human psyche.
What about the scientific validity of multiple intelligences theory? Is it proper to set aside “intelligence” and to speak instead of relatively independent “intelligences”? The simple change from single to plural is not so easily justified. In fact, the available evidence places the tenets of Gardner’s theory in serious doubt. Most problematic is Gardner’s central idea that the identified intelligences are relatively autonomous. This is tantamount to claiming that the common pattern of correlations known as the positive manifold does not apply to the “intelligences.” Yet, as Charles Spearman and countless other investigators found over more than a century of research, Gardner’s claim does not align with the psychometric evidence. Simply put, cognitive abilities are correlated.
The claim that linguistic, logical-mathematical, and spatial intelligences are relatively autonomous is easily refuted by analyses showing that these factors are significantly intercorrelated. These correlations are the mathematical basis for the top-level factor of general intelligence. On the other hand, brain-imaging studies do indicate a partial modularity: Specific brain regions can be linked to particular cognitive functions. Yet, as every neuroscientist knows, modularity is a partial truth. In every expression of higher-order thinking, distinct brain regions coordinate to accomplish complex cognitive work—writing a poem, designing a building, or composing a symphony. Also, not all brain regions are specialized: Some regions serve quite general functions that cut across diverse expressions of the intellect. The frontal lobe, for example, underlies the general function of higher-order planning; another brain structure, the hippocampus, is crucial to the formation of long-term memories. So although evidence of brain modularity may support the existence of distinct intelligence factors, the same evidence does not necessarily support a theory of multiple “intelligences” that operate relatively independently.
Despite challenges to its validity, Gardner’s theory of multiple intelligences became quite popular among educators during the 1980s and 1990s. Those who became acquainted with Gardner’s theory understood that traditional theories of intelligence were easy targets for criticism. The reason was simple: Those theories were not broad enough. Intelligence as classically interpreted was hyper-logical and excessively analytical. This was a problem, and Gardner knew it. Gardner himself did not reject logical and analytical thinking outright, but recognized this expression of mental capability as one of many different expressions of intelligence. It’s easy to sympathize with multiple intelligences theory because it emphasizes breadth, a sensibility that motivates many other investigators in the search for a more inclusive theory.4
Closely related to theoretical breadth is breadth of a different sort. Multiple intelligences theory recognizes that intelligence is more liberally distributed among people than is typically assumed by those who see intelligence as those hyper-logical functions measured by IQ tests. If, instead, intelligence has diverse expressions, then it probably is expressed in diverse people. This idea, more than any other, was the likely basis for the enthusiasm generated by multiple intelligences theory. Teachers resonated with a more liberal conceptualization of intelligence because it did not order their students along a single continuum of cognitive ability. Many saw their students displaying a spectrum of intelligence in varied forms, and Gardner’s theory was fully consistent with that point of view. Yet to some extent the same assumption has support within classical theories of intelligence. Although we can justifiably speak of general intelligence, person-to-person variations in cognitive profiles are common. Flat profiles, in which a person’s specific factors are more or less equal, are rare. In this sense, Gardner was right: People are smart in different ways.
We see hints of cognitive profile differences whenever SAT and GRE tests result in quite different verbal and quantitative subscores, but such variation is also common on subtests that make up IQ batteries. Profile variation tends to be more pronounced with higher IQ scores. Uneven profiles do not negate the validity of general intelligence, but they do show that a single IQ score can hide important differences.5 They also tell us that a highly intelligent society can be, and probably will be, an intellectually diverse society. If uneven cognitive profiles are so common, the pattern suggests untapped potential. It further implies that the purposes of education ought to be broadened to recognize talent in its diverse manifestations. Indeed, one reason Howard Gardner set forth his theory was to spur educators to question the typical school curriculum—to ask whether it was too narrow and, if so, to make corrective changes.
Of course, Gardner’s theory was also intended to spur a reconsideration of intelligence theories. Traditional models of intelligence, expressed as mono-dimensional IQ scores, imply that the world’s population can be ordered along a single ladder of intellectual ability. Gardner’s theory of multiple intelligences presents a liberating alternative; it offers the inspiring possibility that diverse intelligences are broadly distributed in the population. Unfortunately, the theory’s freeing implications rest on a foundation that is inadequate in important respects.6 Getting our theories right is essential. We must try to understand the nature of intelligence in its truest form before we can seriously consider how to expand or increase the collective intelligence of our society. With this in mind, let us now look to other emergent theories of intelligence theory.
Robert Sternberg’s theory of successful intelligence is one of the major current alternatives to traditional theories.7 The name of Sternberg’s theory makes intuitive sense. In the theory of successful intelligence, we find connections to the theories of Galton and Binet in that those early theories were likewise linked to success outcomes—most often academic success. Sternberg does not challenge the fundamental basis for evaluating whether a proposed form of intelligence is legitimate. He recognizes that theories of intelligence are relevant to human activity and aspirations when they predict success in some form.8 When success is not restricted to grade point averages but extends to real-life achievements, creative accomplishments, scientific breakthroughs, meaningful social participation, and emotional well-being, then the underlying intelligence rises further in importance.
In setting forth the theory of successful intelligence, Sternberg did not seek to replace traditional theories so much as to expand them. His theory recognizes three forms of intelligence: analytical, practical, and creative. The first, analytical intelligence, is easily understood. It equates to intelligence as construed traditionally by all of Sternberg’s predecessors, namely, the ability to engage in rational, abstract, and complex reasoning. This is the form of intelligence that is particularly advantageous in academic settings.
While recognizing that analytical intelligence is important, Sternberg argued it does not make for a complete theory. It’s true that analytical intelligence, the kind measured by IQ tests, is a good predictor of success in schools and on the job. But being a good predictor is not the same as being a perfect predictor. If analytical intelligence fails to measure personal qualities that factor into success outside of academic settings, what are those qualities? They potentially include a broad swath of skills and habits: the ability to prioritize, to motivate others to achieve a common goal, to persuade, to gain favor, and dozens of others. These qualities might separate students who make the Dean’s list from people who go on to achieve extraordinary real-life success. Even if we cannot say exactly what those complementary skills are, we need not be deterred from placing a name on this category: practical intelligence.
In Sternberg’s theory, practical intelligence is intelligence applied to real-life situations, including schools.9 In school settings, practical intelligence can be critical to success. Schools are, after all, human institutions in which success is achieved partly by the ability to navigate the system of social relations, norms, and expectations. Practical intelligence may be particularly important in graduate school, where building loyalties and establishing reputations typically rises in value when compared to undergraduate study. Robert Sternberg noticed that some of his best graduate students did not have particularly high GRE scores, yet they were fantastically good at figuring out practical strategies to progress efficiently and to gain favor in the institution. These students were attentive to the social context of success, and particularly to the unspoken rules by which the university operated. They knew how to find the resources they needed to reach their goals. They could distinguish between the official standards of performance and the ones that really count, and knew which channels to pursue to get something done, even if those channels were unofficial.
Some research on practical intelligence investigates the specific skills that contribute to success. A business executive, for example, has to make ongoing judgments about how to handle a stream of tasks. To measure this ability, the scenario has been modeled as the “in-basket” task in which managers process a series of simulated memos, phone messages, and other work-related demands. Possible actions include delegating a task, seeking additional information, and asking for advice.10 Performance on the in-basket task was correlated with background job knowledge, as expected. But high performance also had a creative element: Executives with higher scores were frequently able to generate a large number of related ideas.
Practical intelligence appears to be important as well as distinct from other abilities. Close in meaning to what we might call “street smarts,” practical intelligence is a honed sense of how to behave smartly in particular social and work contexts. Measures of practical intelligence tend not to correlate well with IQ scores, suggesting that the ability to exercise intelligence in real-world settings is truly distinct from traditional analytical intelligence. Yet we should not be too quick to accept practical intelligence as a fully legitimate form of intelligence. For one thing, “street smarts” might be too narrowly confined to specific contexts. Practical intelligence in the domains of athletic performance, acting, entrepreneurship, comedic performance, medicine, and politics might be quite distinct ability sets with little potential for crossover benefit.
Another criterion that can be used to appraise practical intelligence is its predictive power. Traditional IQ scores display validity if they can predict academic success. Does Sternberg’s construct of practical intelligence offer predictive utility for valued performance? The answer appears to be yes. The evidence comes from tests Sternberg designed to measure practical intelligence in specific contexts, such as a business setting. The tests presented credible scenarios, such as how to handle conflicting priorities or resolve interpersonal conflicts. Measures of practical intelligence did in fact correlate with success on the job. More important, practical intelligence had predictive power beyond that offered by traditional IQ tests.
Practical intelligence includes the ability to perceive unwritten rules that are conducive to success. Psychologists have a name for such unwritten rules—tacit knowledge. Tests devised to measure tacit knowledge have shown remarkably good power to predict success—as much as twice that of IQ—among academics and business managers. Job-related tacit knowledge predicts job performance, and the function is largely independent of that offered by traditional intelligence tests. On a technical level, measures of tacit knowledge tend to correlate across domains. This means that people who are good at detecting the unwritten rules for success in one setting are likely to be skilled at understanding a different set of rules in another context. They are quick to understand how to work with other people, what tasks are most important, and how to maximize productivity.11
Measures of practical intelligence have other interesting features, too, including predictable life span trends. Earlier we saw that one form of intelligence—fluid intelligence—tends to peak in early adulthood, whereas crystallized intelligence generally rises until quite late in life. Measures of practical intelligence follow trends closer to crystallized intelligence, rising through the years until about age 70. The escalating trend for practical intelligence through the life span is easily interpretable: Through extended experience, adults improve in their abilities to solve practical problems with advancing years.12 Gains in practical intelligence also seem to parallel increases in wisdom later in life. Indeed, wisdom can be seen as a subset of practical intelligence.13 Like practical intelligence, wisdom draws on the ability to register the tacit knowledge in a particular problem context—to understand the factors operative in a particular situation, whether tacit or explicit. Wisdom entails the ability to weigh those factors and to arrive at decisions that strike a balance between upholding individual interests and promoting the common good.
Creative intelligence is the third aspect of Sternberg’s theory. Think of the world’s great works of art, music, literature, architecture, and engineering. In every case, creative thinking contributed to the insights and breakthroughs that enriched society. Creativity has always figured into the advances of civilizations through the millennia, and its importance has not diminished in our own times. In the most revolutionary applications of information technologies—the Macintosh computer, the Internet, smart phones—old-fashioned analytical thinking was essential but not enough. Creativity propelled the technical breakthroughs that forever changed the way the world operated. It couldn’t be more obvious: The creative mind is one very powerful force. Sternberg’s theory recognizes this key expression of intellect as a fully legitimate aspect of human intelligence.
To test Sternberg’s claim, let’s consider again the question of predictive power. Does measured creativity predict academic success, for example? In the Rainbow Project, Sternberg pitted tests that measured creative and practical intelligence against the decidedly analytical SAT.14 He found that when SAT scores were supplemented by tests of creative and practical intelligence, the combination was more predictive of college grades than was the SAT alone. The correct conclusion seemed to be that academic success in the university is not only a function of analytic intelligence; it also draws upon creative and practical abilities. After graduating, presumably, all three forms of intelligence continue to have value. As alumni make their way through the “real world” beyond the campus, practical and creative intelligence arguably play even more important roles in career and life success. This rationale—that analytical ability is legitimate but not fully sufficient—lies at the heart of Sternberg’s theory of successful intelligence.
Besides expanding our understanding of intelligence beyond analytical ability, Sternberg’s theory challenges the analytical focus of most school curricula. If diverse capabilities really do underlie success, and if the human capital of any company, community, or nation consists of diverse cognitive abilities, then our educational systems ought to recognize the fact and then cultivate those same abilities. The point is that when our assumptions about intellectual capabilities are updated, so should our ideas about the purposes and processes of education. What’s at stake is not only the intellectual readiness of a population to compete in a global economy, but also the cherished democratic ideals of equity and opportunity. Powerful ideas can be springboards for positive change.
But let’s not accept Sternberg’s theory, or any other new theory of intelligence, too quickly. Proposing a new theory of intelligence is not the same as justifying that theory through research-based evidence. As we evaluate the credibility and utility of Sternberg’s theory, we must be careful not to accept it simply because it sounds plausible; rather, we must ask whether it is supported by evidence. It’s not too hard to think of challenges to Sternberg’s claims. We can question, for example, whether creative ability is a broad personal trait that can be applied to many different fields, or whether it is by nature focused in particular fields, such as music or interior decorating.
Whether creativity is narrow or broad matters. When we call creative ability an intelligence, the implication is that it applies broadly—that creative people can employ their divergent insights to different domains. On this question, the evidence is mixed: Creativity is relatively domain-bound, but not completely. Across domains, creative expression is mildly positively correlated. Sometimes people know how to exercise their creative ability in multiple contexts. As a psychological construct, then, the coherence and predictive power of creative intelligence can be defended. The argument for creativity as a major dimension of the intellect is compelling—a fact recognized by Sternberg and many other scholars of the mind. Because this is so, we will now consider creative intelligence on its own terms and independently of Sternberg’s theory.
Let’s explore creativity more fully as a potential form of intelligence. We have seen that analytical intelligence is the brand of mental ability measured by most tests of intelligence and summarized in IQ scores. Those tests measure a person’s ability to take in, or consume, information and to process that information rationally. Highly creative people, however, excel in their ability to produce outstanding ideas. Yet production of information is relatively neglected in most theories of intelligence. It seems hard to argue that production of information is any less important than consumption in the work of the mind. For this reason, many scholars who study giftedness regard traditional analytical intelligence as inadequate to the measurement of productive potential, especially in the creative realms.15
In a way, creativity has little to do with intelligence as typically conceived. The two intellectual modes might even appear to be at odds. The separation of the analytical and the creative aspects of the mind has been inspired to some extent by cases of savant syndrome. Among savants, artistic skill is sometimes correlated with actual dysfunction in the language centers of the brain.16 The onset of dementia affecting the ability to speak can also coincide with the sudden appearance of artistic expression and other savant-like skills. Neuroscientists in Australia have applied this hypothesis experimentally to people with normal brain functioning. By pulsing magnetic fields into the language centers of the left temporal lobe, they were able to temporarily deactivate linguistic processing. During the few hours when language functions were compromised, some of the participants began to display savant-like abilities. Some subjects displayed enhanced artistic ability, while others showed improved accuracy in numerical estimation of quantity.17 These findings suggest it is possible to “turn on” creativity and savant-like abilities by “turning off” parts of the brain responsible for other functions.
Creativity and intelligence are not necessarily antagonistic. Statistically speaking, creative thinking and intelligence are positively correlated, at least in the low to mid range of IQ. People recognized in their fields as highly creative often have at least moderately high levels of intelligence. But beyond an IQ of 120 or so, creativity and intelligence decouple such that one has little to do with the other.18 This decoupling is consistent with the threshold theory of creativity: Its central claim is that traditional analytic intelligence is a platform on which creativity can build. A moderately high IQ seems to be all that is needed for truly outstanding creative accomplishments. But an IQ of 120 or so is a foundation, not a cutoff, and so the IQs of highly creative individuals often range much higher. In fact, the average IQ of the world’s most eminent creative people extends to around 150.19
There is an alternative way to interpret the same data, however. It’s possible that IQ and creativity are not related—at least not in any direct way. Instead, the opportunity to be creative requires the proper certification, a degree or license for practicing a profession.20 Such a certification typically requires prior admission to a university or professional school, which in turn requires the kinds of intelligence measured by IQ tests. If this is the true explanation of the correlation between IQ and creativity, then the supposed relationship is spurious or at least misleading. Analytical intelligence would be a bureaucratic prerequisite, nothing more.
Most scholars perceive an important relationship between analytical intelligence and creativity. They differ, however, in how they interpret that relationship. Some theorists see analytical intelligence and creative thought as organically connected—so much so that they interpret creative expression to be the “ultimate extension” of intelligence.21 Other theorists treat intelligence and creativity as overlapping sets of cognitive abilities. Still other scholars draw firm distinctions between the two, interpreting intelligence and creativity to be quite different forms of cognitive expression. Their argument is that intelligence operates on the facts as given, whereas creativity relies much more on the particular ideas brought by the problem solver.
Creative expression may hinge on the ability to generate a large number of ideas and to select the best from among those.22 The act of generating lots of relevant ideas or potential solutions is known as divergent production. This aspect of creativity contrasts with the requirement of convergent thinking—the generation of a single best answer or solution—demanded by many IQ and school-based tests. Divergent production entails searching broadly for relevant information and considering multiple solution paths rather than a single best response. The point is not to select the one “right” answer, but to produce many “good” answers. This criterion for evaluating ideas requires a significant shift in mindset.
Analytical and creative thinking are each distinctly valuable, but when combined they are fantastically potent. That’s because these two forms of intelligence complement each other. Analytical thinking converges on a precisely correct answer, decision, or action. Creative thinking works oppositely—not through convergence but rather through divergence. The creative mind explores new possibilities, patterns, and options. It builds on what is already known but then transcends the “givens” to envision something new. The ability to think in new ways inevitably interacts with what the creative individual already knows. This raises the question of the role of knowledge in highly creative work.
Knowledge is widely believed to be double-edged.23 On the one hand, extensive knowledge of a field is often necessary to recognize what counts as novel and creative.24 To some extent, creative breakthroughs depend on knowing what has come before. Yet knowledge also guides thinking, and so may constrain ideation to the point that novel ideas are hard to generate or comprehend. Entrenched modes of thought might render the knowledgeable person incapable of understanding or accepting new ways of thinking. Breakthrough concepts may seem unintelligible, at least as viewed through the filter of old ideas. Highly creative people therefore face dual demands: They must be grounded in the creative field and yet have the intellectual flexibility to transcend the categories and boundaries of that field as traditionally understood and practiced.
The lives of great innovators can shed light on the real-world conditions under which rarefied creativity originates. Biographies of Pablo Picasso, Igor Stravinsky, T. S. Eliot, and Martha Graham show that these luminaries were prepared to concentrate on their work to such an extreme degree that they isolated themselves from common pursuits and pleasures.25 Their accomplishments required extraordinary concentration over long periods of time. To reach this supreme level of achievement seemed to require patterns of living that were configured to favor the creative pursuit. Other evidence suggests that highly creative people have a knack for identifying emergent ideas that are either not recognized or not respected by others, but which have high potential. Like investors who are skilled at picking underpriced stocks, innovators know how to identify undervalued ideas.
In some instances, new ideas can hold so much inherent potency that they redefine a field. Yet creative breakthroughs are sometimes powerfully resisted. To advance novel ideas, the creative individual not only needs a sense of which ideas are most promising, but also the ability to weather the countervailing opinions, including ridicule and scorn. Because new ideas may be rejected for long stretches of time, possibly through the entire life of the originator, creative people very often display extraordinary persistence.26
For many decades, theorists have known that intelligence can be applied to different forms of informational content, including numbers, words, and abstract symbols. In fact, we can think of classic analytical intelligence as skill in processing such abstractions. Intelligence can also be applied to abstract visual patterns or mechanical assemblages, forms of information that are not symbolic but rather are based on an understanding of spatial relations. Yet a third major category of information content is human behavior, both one’s own behavior and the behavior of other people. The focus here is on the ability to understand and react intelligently to human behavior, the skill set sometimes known as social intelligence.
Social intelligence entails, at least in part, the ability to understand the feelings, beliefs, and intentions of other people, especially when those internal states are communicated through facial expressions, gestures, posture, or tone of voice. With some ingenuity, it’s possible to test this ability. One method is to use photographs or voice recordings as stimulus materials. Examinees are asked to describe the emotions of the person depicted: Are they feeling sad, happy, afraid, guilty, tense, or bored? Such a test measures the ability to interpret behavior, but its measurement of social intelligence is passive and only partial.
Interpreting behavior and actively coping with that behavior are two independent skills, and both are manifestations of social intelligence. Social intelligence includes skill in both the passive interpretation of emotional states and the ability to interact effectively with others. This is social intelligence defined abstractly, but we can be much more specific. Skills that fall under the banner of social intelligence include interest in other people, empathy, social performance, emotional expressiveness, and sensitivity to others’ lack of confidence. Potential factors that remain less explored are memory for names and faces, and the ability to produce the “right” behavior in particular settings, such as proper etiquette.
Alongside theories set forth by psychologists is research on what laypeople believe makes for a socially intelligent person. Nonpsychologists think of social intelligence as one important expression of what it means to be smart. They believe that a socially intelligent person:
• Displays curiosity
• Does not makes snap judgments
• Is sensitive to other people’s needs and desires
• Understands others’ thoughts, feelings, and intentions
• Is good at dealing with other people
• Has extensive knowledge of rules and norms in human relations
• Is good at taking the perspective of other people
• Adapts well in social situations
• Is on time for appointments27
Although social intelligence has not played a very important role in most theories of human intelligence, there are exceptions to this pattern. One is Howard Gardner’s theory of multiple intelligences, described earlier. One of the “intelligences” proposed by Gardner is interpersonal intelligence, which corresponds rather closely to social intelligence. Gardner argued that clinical evidence on the effects of brain injury could help establish the legitimacy of a proposed new intelligence. He reasoned that if localized brain damage compromised one intelligence but left other intellectual abilities intact, that limited impairment constituted evidence that the particular intelligence exists and is relatively autonomous. We can apply this rule to social intelligence. Clinically, quite a bit of evidence shows that social intelligence is anatomically distinct from other cognitive abilities. The brain’s frontal lobe, in particular, is crucial. This brain region allows us to interpret other people’s actions and emotions, as well as exercise control over our own emotions and emotionally driven behavior.
Social intelligence entails more than the ability to label or control emotions. It also has components of knowledge stored in long-term memory. Socially relevant knowledge is of two kinds. One is factual knowledge about rules for social interaction, good manners, and so on. The other knowledge resource consists of memories of our past personal experiences. This autobiographical memory, called episodic memory by psychologists, is composed of episodes remembered from our lived experiences.28 A large proportion of these episodes are social in nature, and this experiential memory bank is a resource we can draw upon in knowing how to interact with other people in the here and now. There is one potential complication, however: Episodic memory cannot be presumed to be accurate. Indeed, serious distortions of memory for “what actually happened” in this event or that are not only possible but common.
Let’s focus now on the lingering question: Is social intelligence really a legitimate aspect of the human intellect? In some studies, social intelligence is correlated with IQ—but by itself this neither qualifies nor disqualifies social intelligence as a valid form of intelligent expression. Indeed, as a rule, different intelligence factors are positively correlated, at least moderately, and so we should not expect social intelligence to be completely independent of IQ. More problematic is that measures of social intelligence sometimes overlap with verbal ability, which is strongly associated with general intelligence. This complicates the measurement of social intelligence and throws into question its status as a legitimate, independent form of intelligence. That is because whenever social intelligence draws heavily on verbal ability, social intelligence becomes less distinguishable as an independent form of cognitive ability. Yet data also show that social intelligence holds together as a coherent factor. These mixed results indicate that, from a psychometric perspective at least, the status of social intelligence is ambiguous.
Whenever we wish to evaluate a potential new form of intelligence, we must ask whether variation in the proposed intelligence affects a person’s ability to adapt. We have plenty of evidence showing that deficits in social intelligence have serious consequences. Consider that a lack of skill in relating to other people has long been recognized as a key marker of autism. Autistic people often have particular difficulty understanding that other people have thoughts that differ from their own. Nonautistic people know perfectly well that other people have independent thoughts and feelings. Those differences are, in fact, the entire basis for communication between people. Why communicate otherwise? To understand that people differ in what they think, know, and believe is to have a “theory of mind.” A person who lacks this awareness is unlikely to reflect on his or her own thoughts and would find no logical reason to pursue interpersonal communication.
One reason for the semi-legitimate status of social intelligence is that it seems to be different in nature from analytic forms of intelligence. Traditional intelligence is measured on scales indicating “less” and “more” based on questions that have right and wrong answers. The personal resources that constitute social intelligence do not seem to work this way.29 Social contexts may be too variable to submit to analyses that are reducible to right and wrong answers. Moreover, social perceptions and sensibilities that are functional in one culture may be nonfunctional in a different culture. What qualifies as intelligent social behavior might well vary from one neighborhood to another, or even from one household to another. Sensibly enough, what constitutes social intelligence must be understood in its social context as well as in the context of an individual’s valued goals.
Let’s conclude this way: Social intelligence has not yet been established as a fully legitimate form of human intelligence. To some extent, the problem is how to define and measure the various competencies that underlie social intelligence when we don’t yet know exactly what it is or how it relates to traditional IQ. If these technical problems are eventually solved, we may find that social intelligence and traditional IQ have a complicated relationship. For example, knowing that another person’s intelligence is much higher or lower than one’s own can complicate how we view that person. Candidates for public office know this and may consciously try to affect the speech and demeanor of “regular folk.” Such affectations are strategic. If a political candidate flaunts a very high IQ, it might be hard for rank-and-file citizens to trust the person. This may be why history’s most eminent leaders seem to have moderately high, but not extraordinary, levels of intelligence. The risk of a very high IQ is that these potential leaders may talk over the heads of the general population. In two American presidential elections, for example, the highly intellectual Adlai Stevenson lost to the less brainy Dwight D. Eisenhower, possibly in part for this reason.30 To the degree that followers want to identify with their leaders, they may have trouble seeing a member of the intellectual elite as understanding their needs and representing their interests.
Without question, our emotions affect how and what we think. Anxiety and frustration can obstruct intelligent thought, whereas more positive emotions, such as interest and surprise, can direct and energize the mind.31 One of the more studied aspects of negative emotion is test anxiety. Nearly everyone has the potential to feel anxious while taking a test, but in rare cases anxiety can be so debilitating that it nullifies a person’s ability to express his or her knowledge or intellectual capabilities. Emotions can compromise rational thinking in everyday life, too. Frustration can trigger a rash decision to “let off steam” in the short term, but in the long term taking such action is counterproductive. Telling off the boss at the end of a hard day is probably imprudent, and driving a car recklessly in a fit of rage is dangerous. On the other hand, positive emotions can be powerful elements of the intellect. The capacity to become interested in ideas, insights, and discoveries—to derive satisfaction from the active use of the mind—is a sign of intellectual engagement and perhaps giftedness. Gifted children seem to have a large “capacity for high levels of interest, enthusiasm, fascination,” and the ability to become deeply engrossed in an area of study.32
Auguste Rodin’s famous sculpture The Thinker depicts a man in deep contemplation. His very posture exudes concentration. The Thinker’s body is twisted, with his right elbow perched on his left knee and his fist placed to chin, and his torso is angled steeply forward. Rodin’s sculpture communicates an incontrovertible truth about the mind: Thinking can be very hard work! Yet thinking can be highly pleasurable. Some people find the active use of their mind so rewarding that it borders on addictive. They need to think.33 Just as everyday life imposes the need to consume air, food, and water, many people experience a driving appetite for ideas. Psychologists have found ways to measure a person’s “typical intellectual engagement,” which at the high end is expressed as interest in a wide variety topics and a persistent desire to understand the world—in short, a need to know.34 Intellectual engagement can be deeply rewarding. The positive emotions that result from thinking can, in turn, propel the intellect to greater heights. The payoff can be tremendous: Great intellectual achievements are frequently driven by deep passions.
Almost everyone has experienced the organic connection between the active use of the mind and the pleasure that follows. Sometimes we experience that pleasure as an emotional state known as “flow.”35 Think of a time when you became completely engrossed in an activity, such as solving a difficult problem at work or pursuing a hobby that required significant concentration. If you were truly challenged—if the mental task required your full attention—then perhaps you experienced a state of consciousness called flow. The telltale sign of flow is a constriction of attention so complete that you lose a sense of time passing as well as an awareness of events around you. Hours may go by without you realizing that your attention is being dominated by the activity at hand. Yet, strangely, the experience is not at all draining. Rather, your engagement in the activity is energizing as well as highly pleasurable.
If you have experienced flow, then you know firsthand that there is an important link between emotions and intelligent thought.36 But what exactly is that link? One theory goes by the intriguing term emotional intelligence. In contrast to theories that present intelligence as cool and calculating, emotional intelligence spotlights the ability to understand and control one’s own emotions, as well as the ability to comprehend the emotions of others. The theory suggests that emotions can be powerful allies if people can channel those emotions properly. So rather than depicting emotions as contributing to irrational or erratic behavior, the theory maintains that emotions—if skillfully understood and acted upon—have tremendous potential to energize and direct intelligent thought.
Popularized by Daniel Goleman’s book Emotional Intelligence, the concept has struck a chord with the public.37 The theory of emotional intelligence is not a product of wild speculation, but is based on systematic programs of research carried out at Yale University and elsewhere.38 Emotional intelligence is defined as “the ability to perceive and express emotion, assimilate emotion in thought, understand and reason with emotion, and regulate emotion in the self and others.”39 In other words, emotional intelligence involves more than simply an awareness of emotions in oneself and other people; it also includes the capacity to use emotions productively. Despite such an inclusive definition, theorists portray emotional intelligence in somewhat different ways. The key issue seems to be whether emotional intelligence is strictly an intellectual ability, in which emotions are the content, or whether emotional intelligence also encompasses noncognitive traits, such as personality and character.
No one doubts that both kinds of characteristics—intellectual and nonintellectual—are important. Both contribute to the full gamut of human behavior and, ultimately, to variation in success. For example, personal factors that include emotional elements are known to be good predictors of first-year college grades quite independently of SAT scores. But is it proper to call such personal qualities intelligence? Researchers at Yale have emphasized the intellectual aspects of emotional intelligence and segmented off emotional feelings as significant but separate from emotional intelligence and outside their realm of concern. When the definition of emotional intelligence is restricted in this way, it becomes possible to study how the intellectual aspects of emotional perception and control interact with emotional traits, such as warmth or outgoingness.40 By excluding noncognitive traits and focusing strictly on the intellective aspects of emotional intelligence, a stronger claim might be made that the new construct really is a form of intelligence.
Yet, as we have seen, there are good reasons to be skeptical about any proposal to introduce a new variety of “intelligence.” Any newly proposed form of intelligence must have several key qualities to be a defensible candidate. For example, to be regarded as a form of intelligence, any personal quality must be fairly stable within a person. Stability ties back to the classical notion of IQ. Whatever IQ measures, it has to be seen as stable—not to say fixed—rather than a fluctuating quality, such as mood. Variability across the population is a second important quality. Every theory of human intelligence assumes variability within the population, which in turn underlies the logic of measurement. Not just intelligence, but every important psychological construct must be variable from person to person to be meaningful. Otherwise, what is the purpose of measuring it or investigating its predictive power?
A third vital quality for any trait we call intelligence is generality across situations. When we say that a person is intelligent, we normally mean that the person expresses a high degree of intellectual skill in a variety of ways, not just narrowly, such as in card games. A card player might have a remarkable memory for previously played hands, and we may be impressed by the player’s skill, but unless that impressive feat of memory is exhibited widely in other contexts and problems, its classification as a form of intelligence would not be credible. To count as intelligence, it must apply broadly. In fact, the broadest form of intelligence, general intelligence, is hypothesized to factor into every cognitive performance.
A fourth quality of any form of intelligence is that it has to display certain properties of measurement. In particular, when the trait is measured, the resulting scores need to show that the trait holds together as a single variable—what we call factorial coherence—rather than a collection of variables. Also, the new variable must be distinct from other constructs that have already been established. Finally, any alternative model of intelligence ought to predict human effectiveness beyond what is offered by IQ scores. Some of the initial excitement about emotional intelligence emerged from claims that emotional intelligence is a better predictor of success than IQ. But that claim is not true. Nor was emotional intelligence ever expected to surpass IQ in its ability to predict success. The more modest and realistic goal was to add predictive power independently of IQ scores.
So, is emotional intelligence really a form of intelligence, or not? The evidence is mixed. Tests of emotional intelligence often measure individual variation that is distinct from what IQ tests measure. At least in some contexts, emotional intelligence can predict success above and beyond IQ scores alone.41 Also, some tests support the factorial coherence of emotional intelligence—a positive sign. Other tests, however, fail to confirm coherence as a unified factor.42 This means that various tests of emotional intelligence measure somewhat different traits. Obviously, serious challenges remain. Someday we might conclude that emotional intelligence was actually misnamed as an intelligence. If it consists of several distinct abilities rather than a single, unified capability, it will have to be broken down into simpler and more precise constructs. The challenge facing future researchers will be to iron out these technical wrinkles by building better measures and better theories.
Emotions are clearly vital to the functioning of the intelligent mind. Perhaps emotional intelligence really is a neglected ability that ought to be developed in the population alongside traditional intelligence. One reason to recognize and value emotional intelligence is that it highlights a rare benefit of aging: Scores tend to increase over the life span. Older people often understand and regulate their emotions better than younger people do. If recognized as important, emotional intelligence could be more intentionally developed among workers for greater overall effectiveness, as well as among students to promote their own social and emotional development. A population with higher levels of emotional intelligence will almost certainly be psychologically healthier, more productive, and more capable of forming and maintaining satisfying relationships. The world might be a more pleasant and productive place if we were all more emotionally intelligent.
Yet another theory of intelligence, still embryonic in its development, falls under the banner of distributed intelligence. The theory, introduced by Stanford professor Roy Pea, recognizes that every manifestation of human intelligence entails much more than a single mind acting in isolation. Rather, intelligence always operates in and through an array of resources defined by a culture. According to this theory, intelligence reflects the concepts and priorities of a local culture, as well as draws upon a panoply of tools, technologies, and practices offered by the culture. In Pea’s words, intelligent thought and action occur in a world “thick with invented artifacts.”43
Distributed intelligence corrects limitations of the way psychologists tend to construe intelligence—namely, as the function of individual minds and nothing more. Distributed intelligence extends the geometry of intelligence to recognize that minds interact with the “stuff” of the environment. Landing an airplane, for example, requires complex and dynamic interactions between a pilot and cockpit instrumentation, alongside coordination with air traffic controllers.44 The pilot must be keenly aware of the plane’s weight and the flap extension, and keep the airspeed within a relatively narrow band. During landing, the plane’s dynamics change second by second. The intelligent thought needed to accomplish this task safely requires nuanced and highly coordinated interactions between the pilot and instrumentation—not the isolated activities of the pilot’s brain.
Even simple technologies can be important aids to thought. Written information, whether on sticky notes or in treatises, can compensate for the predictable fallibility of memory. Paper plus pen plus mind, when functionally linked, are more powerful than the mind acting without external support. It is impossible to imagine the magnificent achievements of Shakespeare, Beethoven, or Galileo without the technologies available to them—quill, piano, telescope, and much more. Obviously, human capability is greatly magnified by the tools afforded by culture. Harvard psychologist David Perkins expressed the point starkly when he declared that human intelligence is “not the solo dance of a naked brain.”45
During the last century, the powers of technology to leverage intellect grew exponentially. Digital computers are the most vivid example of technologies that have amplified the mind’s capabilities. In the early days, computers eased the clerical worker’s burden of tedious calculation and record keeping, but this was just the beginning. The capacity for accurate calculation and mathematical modeling, as well as long-term data storage and lightning-fast retrieval, were breakthroughs that opened possibilities that were previously inconceivable. The moon landing of Apollo 11 in 1969 was the ultimate proof that technologies can amplify the power of intellect to unknowable proportions. Since that time, computers have extended their reach into every corner of human activity—science, the arts, music, communication, political movements, commerce, warfare, social relationships, and so on, interminably.
Other technologies, new and old, drive the point home. The invention of writing, the printing press, movable type, and word processing software catalyzed an explosion of published intellectual work and stocked the world’s libraries and data servers with countless volumes of knowledge. Published knowledge, in turn, preserves knowledge for generations to come. The Internet is yet another potent example of distributed intelligence because it permits instantaneous access to the intellectual products of vast numbers of minds on all manner of topics. Those resources include shared databases, analytical tools, and visualization software—collectively known as the cyberinfrastructure—that is pushing scientific discovery into new frontiers.46 Personal electronic devices, now ubiquitous, bring the vast and expanding electronic resources to the world’s population, not only in wealthy nations but also in developing countries.
We must be careful, though: Technologies do not constitute an unmitigated good. It’s all too easy to think of technologies that have brought widespread pain and misery, or that presented scenarios of destruction too terrible to contemplate. Neither do technologies inevitably make us smarter. Indeed, the invention of simple arithmetic calculators evoked deep concern among some educators that the devices would undermine students’ understanding of the fundamentals of mathematics. Yet even with these caveats, the inventions of culture have undeniably expanded our cognitive powers. In aggregate, they do make us smarter. Arguably, nothing resembling human culture and civilization would be possible without an alliance between the mind and the mind’s inventions. If this is true, then the central tenet of distributed intelligence must be valid: Intelligence is not confined to the cranium, but lives in the organic web of minds and tools that, in combination, make for a formidable intellectual force.
The tools that constitute distributed intelligence not only extend the mind’s capabilities, they also affect the way the mind operates. The most profound influence in this regard is the way that written text introduced a powerful new mode of intellectual activity.47 Gutenberg’s printing press offered an unlimited capacity for replication, which meant that complex and extensive lines of reasoning captured in book form could be disseminated to a vast readership. That readership could be distributed across great spans of distance and time, removed from the original authors by thousands of years and thousands of miles. Written language, particularly in the form of books, stabilizes language and invites extended reflection and critique. During the Enlightenment and Scientific Revolution, dialogues among scholars led to rapid cycles of evolving knowledge and seismic shifts in ways of thinking about the world. In contrast to the oral discourse of myths and legends favored by the ancients, books easily conveyed ideas in expository form. These extended logical essays offered the mind a new pattern for thinking with, and about, the bodies of knowledge that constitute science, philosophy, history, and every other field of inquiry.
One enduring question in intelligence theory is the role of culture in defining what intelligence means and how it can be fairly measured. It seems likely, even obvious, that what constitutes being smart in one culture cannot be precisely the same in another culture. But is the meaning of intelligence radically different from culture to culture? Scholars take varied positions on the matter. At one end of the spectrum are cultural psychologists who argue that intelligence is fully culture-bound, and therefore relativistic rather than absolute. This position implies that the elements of intelligent functioning in one cultural context may be distant from, or irrelevant to, that of another culture.
One anecdote memorably illustrates radical relativism with respect to culture. Working with the remote Kpelle tribe of Liberia, psychologist Joseph Glick and his collaborators presented what they thought was a straightforward categorization problem.48 Various objects, such as a knife, hoe, potato, and orange, were placed on the ground. The Kpelle tribesman was then asked to sort the objects. The field researchers expected that the man would group the tools together into one cluster and the foods in a different cluster—a sorting method that derives from the abstract categories of tools and foods. But rather than using this scheme, the Kpelle man grouped the potato with the hoe and the orange with the knife. This arrangement was based on functional relationships, a categorization scheme that was fully sensible on its own terms, yet would be judged as unsatisfactory according to Western standards for assessing cognitive development. Was the Kpelle man intellectually stunted, or did he simply think differently from the Western scholars who presented the sorting task? Perhaps sensing bewilderment on the part of the Westerners, he justified his answer by saying that it was the way a wise man would do it. Curious, Glick asked, “How would a fool do it?”49 In response, the man sorted the objects according the abstract categorization scheme prized by the Western intellect—foods were grouped together, tools were grouped separately, and so on.
The story of the Kpelle tribesman drives home the cogency of cultural relativism. But is the meaning and expression of intelligence fully and completely culture-bound? Some scholars take issue with this assumption, believing that what counts as intelligence does indeed vary from culture to culture, but not radically so. Instead, they propose that some cognitive capacities and skills have utility in multiple cultures, or perhaps in any culture. For example, a large memory capacity, the ability to learn quickly, and skill in the use of language would all seem to be advantageous in virtually any cultural setting. Cross-cultural psychologists want to understand precisely which aspects of the psyche are universal and which vary across cultures.50 They seek answers not only in geographically distant cultures, but also between cultural subgroups within a common society.
Language, an intimate correlate of culture, is certain to be an important piece of the puzzle. That is because language is not simply a means of communication between people; it seems also to exert quite profound effects on thinking. A famous example is the observation that Eskimos have several distinct words that refer to snow, and that those fine distinctions allow for precision not only in communicating about snow, but also in thinking about it. The great anthropologist, Franz Boas, noted that those meanings include falling snow, drifting snow, snow bank, and snow on the ground.51 The full number of distinct Eskimo words for snow is about a dozen, although in popular accounts that number is sometimes inflated to as many as 50.52 Still, the hypothesis is clear: If one language has 12 words to describe variations on a concept and another language has only one, the two languages afford different levels of precision. Someone who can employ a dozen different words to distinguish various shades of meaning might well think about a concept more richly than, or at least differently from, someone whose language offers only a single word to refer to the counterpart idea.
This provocative notion—that language guides and constrains thought—is known as the Whorf hypothesis.53 It is named for the linguist Benjamin Whorf, whose work in the 1930s included deciphering from ancient artifacts the symbols composing the written Mayan language. Whorf’s conjecture has been somewhat controversial. Some linguists believe that the mind’s ability to think is largely free of the effects of specific languages.54 But the Whorf hypothesis seems to be at least partly correct. Consider, for example, the English words river and stream. Their meanings are distinguished primarily by size: Rivers are larger than streams. The rough equivalents in French are fleuve and rivière, but their meanings do not map exactly to the river and stream distinction.55 Instead, fleuve refers to water that flows into the sea, and revière indicates water that flows into a fleuve. This simple example plainly shows that the English and French words have different meanings, implying that English and French speakers, monolinguals especially, probably think about flowing water in somewhat different ways. A reasonable extrapolation is that language can guide and constrain thought. Because culture is so tightly bound to language a further inference is that thought, including intelligent thought, will have different qualities from one language and culture to another.
What can we conclude about the nexus of culture and intelligence? Minimally, this: What is regarded as intelligent thinking and behavior cannot be identical from culture to culture. That is because cultures diverge in their assumptions about the patterns of reasoning and behavior that mark the intelligent person. Those assumptions are rooted in cultural conventions—the shared values, knowledge, and modes of communication that are deemed normal and expected. The example of the Kpelle man is a perfect case in point: Western academic conceptions of intelligence readily assume that intelligence can be measured through tests that present decontextualized problems in which the measured ability (such as category formation) can be separated from the functional associations of the content (such as food and tools). The questions posed may have nothing to do with anything that is meaningful in the life of the community. Moreover, the questions are likely to be posed by a stranger who has no ongoing social connection to the examinee, which by itself might seem quite absurd.
Most problematic of all, cultures vary in how they define intellectually acute thinking, whether it goes by the name intelligence or wisdom. When abstract categories are emphasized over practical utility, then a certain view of intelligence is assumed—one that emphasizes particular aspects of intelligent thought while minimizing others. Western psychologists placed a high value on abstractions, but to the Kpelle man the detachment of category from function was foolish. Scholars may be disinclined to question their assumptions about intelligence, perhaps because their bias so easily coincides with an intellectual habit that for centuries was accorded supreme value in academe: the ability to detach oneself from context, to rise above it, and to form powerful and far-reaching generalizations.56 If unquestioned, this classic way of thinking about the developed mind hardens into a “spurious orthodoxy”—a “right way” of thinking that pushes aside alternative ways of understanding human intelligence.57
The response of the Kpelle man shows that a locally legitimate expression of intelligence can clash with assumptions underlying its measurement. Other research confirms the pattern. In Kenya, Robert Sternberg and his fellow investigators found significant differences between Kenyans educated in Western-type schools and those who were not.58 Sternberg compared the two groups on their knowledge of traditional tribal medicine and medical concepts derived from science. Kenyans who had Western-style educations did relatively well on the tests based on medical science but poorly on tests that measured knowledge of traditional medicine. Those who did not attend schools showed the opposite pattern: They performed well on tests of traditional medicine but poorly on tests that tapped knowledge of standard medical science. Clearly, different cultural contexts, Western and non-Western, called for and cultivated different forms of intellectual development. Between cultures, the only constant seems to be the adaptability of the individual to learn what is valued and useful.
Seeing intelligent thought as tied to specific contexts makes sense. In the causal chain running from ability to behavior, intelligence must eventually connect to real-world problem solving.59 Intelligence applied to nothing is no intelligence at all. We need to be cautious, however, in drawing generalizations from limited examples. Just as we can cite examples of how intelligence is variable across cultures, we might just as easily think of cases in which certain intellectual abilities, such as the skilled use of memory, are useful across distinctly different cultures. But even when anthropologists acknowledge that some cognitive processes are probably useful across cultures, they often hold fast to a position of value neutrality—that no culture can claim to be more advanced than any another.60 Their prime axiom is that, ultimately, each culture defines what it means to be intelligent on its own terms.
The focal dilemma here—Is intelligence culturally defined?—seems to evade an easy yes or no answer. Even if intelligence is not radically different from culture to culture, local traditions can differ markedly on the emphases placed on various aspects of intelligence. One culture might emphasize logical analysis, while another attaches particular importance to creative expression; still another might elevate social skill above all. Anthropologist Thomas Gladwin documented the unusual importance placed on spatial ability by the Puluwat people of remote Micronesia. For centuries, Puluwat men sailed between widely scattered islands without the aid of navigational technologies, such as GPS devices or premodern sextants and star charts.61 Archipelagos in the region consist of chains of small islands separated by vast expanses of sea. Commerce and intermarriage depended on the navigators’ ability to sail from island to island, yet doing so was extremely challenging by dint of sheer geometry—in Micronesia, the ratio of landmass to ocean is tiny. The perpetuation of culture and individual survival depended on sailors’ ability to arrive precisely on target when sailing from one island to another.
The Puluwat people adapted to their environmental demands by recognizing a class of ocean sailors who specialized in navigation. To be admitted to the navigator guild was a prized achievement; the body of knowledge mastered by navigators was carefully guarded. Indeed, navigators constituted something of a cult group among the traditional Puluwats. Gladwin’s research revealed that information utilized by navigators was multifaceted, involving knowledge of the celestial patterning of stars, and familiarity with winds and currents, as well as sea life. Navigators were apprenticed to master these convergent sources of information so that while under way they could home in on their intended destination. When integrating these various data sources, the navigators of Puluwat relied on and developed spatial reasoning to a highly pronounced degree. As we have seen, spatial ability is a recognized expression of intelligence in classic hierarchical models, but in most studies it is less sweeping in importance than fluid and crystallized intelligence. The intriguing example of the Puluwat navigators shows that in some cultures and contexts, the relative importance of intelligence factors might be ordered differently from the standard model.
Taken together, these examples converge on a crucial point: What is regarded as intelligent thinking in one culture can be very different, even opposite, from what is considered intelligent thinking in another culture. Local conceptions of intelligence can deviate sharply from Western assumptions that intelligence entails the ability to isolate problems from context. Moreover, the belief that intellectual capability is an individual trait that can be assessed independently for each person might conflict with an alternative viewpoint held by collectivist societies—that knowledge and capability are fundamentally properties of cooperative behavior.62 An equally serious clash of beliefs may appear when examinees are asked to justify their answers to questions. Westerners are used to the idea that knowledge can be separated from the person who holds that knowledge. Other societies see knowledge and the “knower” as inextricably bound. Consequently, the expectation that a person must justify his or her beliefs may appear to be nonsensical or threatening. The assumptions inherent in modern intelligence testing—that intelligence is a property of individuals and that knowledge can be separated from the knower—appears to arise from exposure to Western-style education. In particular, the written word, which physically separates the idea from the idea generator, seems to play a crucial role in the inculcation of the Western mindset.
Knowing that intelligence testing rests on many questionable assumptions has sensitized psychologists to the relevance of culture to intelligence, and has resulted in greater caution about using published tests to judge the intelligence of non-Western people. Some psychologists have even tried to devise “culture-free” tests that present puzzles using abstract shapes and that minimize or avoid language. Yet even these tests are known to have elements that could be interpreted only with background knowledge that is grounded in culture. “Culture-free” is apparently no more than a phantom ideal. The best psychologists can hope for is to devise a test that is “culture-fair” or “culture-reduced,” while bearing in mind that fairness is a matter of degree rather than an achievable absolute. This cautionary stance is commendable yet ironic: Attempts to generalize about culture—to speak about culture with detachment—are paradoxical. If we believe that culture pervades all, then there is no standing apart from culture to generalize about it. We always speak from within a culture when we speak about it. Pure objectivity is simply not possible.63
Ponder this: Does being intelligent guarantee consistency in behaving intelligently? Obviously not. Smart people do dumb things, and not infrequently. For this reason, some theories of intelligence distinguish between the ability to exercise intelligent thought and the tendency to do so. People sometimes fail to think reflectively because they lack sensitivity to occasion or do not perceive the need. For example, when faced with a dilemma in which both sides of an issue could be fruitfully explored, otherwise sensible people can jump to conclusions prematurely. They might fail to notice the opportunity to think critically or lack the inclination to consider the issue more fully. In either case, cognitive ability is not the limiting factor in the exercise of intelligent thought, but rather sensitivity and inclination are.64 Aside from learning how to think intelligently, what’s clearly important is developing the habit of doing so. When such inclinations are established, the intelligent person is said to exhibit positive thinking dispositions or “habits of mind.”65
Good habits of mind can counteract mindless ways of processing information and can mitigate impulsivity. For example, ambiguous or fuzzy ideas can trigger the search for conceptual clarity. Other positive habits include forming strategic plans instead of proceeding haphazardly, and trying to be intellectually careful rather than sloppy. All these habits of mind tend to hold this in common: thinking somewhat more extensively and deeply than is our ordinary tendency.
By elevating the value of good thinking, we tacitly assume that human cognition is typically suboptimal. Is this the case? Before answering, let’s think of two opposite scenarios. Depending on the occasion, we might allocate too much or too little thought in making a decision. In playing a game of checkers, for example, you might err by moving the pieces impulsively with little thought given to the consequences; alternatively, you could deliberate for an hour about the implications of a single move. Both can be seen as errors. Everyday decisions, such as choosing clothing for purchase, are likewise subject to both errors—thinking too much or too little. Is one error more common than the other? Psychologist Jonathan Baron has suggested that human cognition is biased in one direction: We tend to devote insufficient thought in forming our beliefs and decisions. If this is correct, then one way to build good habits of mind is to move the bias point of mental investment. The imperative would be to extend thought. “Think more” seems to be the most basic advice for developing good habits of mind.
As we consider how a society can teach the thinking skills that constitute intelligence, let’s remember habits of mind—the tendency to apply our thinking abilities whenever we face occasions for their profitable use. Intelligence is not entirely a matter of capability; it’s also about using that capability when suitable occasions arise. After all, a mechanic needs more than a well-equipped toolbox. The proper tools must be paired with a good sense of when to use each tool. When intelligence is viewed this way—as akin to an intellectual toolkit—habits of mind are clearly vital elements of our intellectual heritage. Like the knowledge and skills that compose intelligence, habits of mind must be cultivated afresh each generation. Societal institutions, schools especially, must find ways to pass on the forms of knowing, thinking, and feeling that call forth the highest and best powers of the mind.
Our primary concern with human intelligence need not exclude considering the intelligence of other species. To gain perspective on whatever unique features the human intellect might exhibit, we may find a contrast with nonhuman species illuminating. Conceivably, the intelligences of dolphins, apes, and ravens have distinctive qualities that we could profitably emulate. With that possibility in mind, let us examine intelligence in the rest of the animal kingdom—in particular, whether the intelligence of nonhuman animals differs in degree or kind from the intelligence of human beings.
Without question, the flexible cognition offered by human intelligence is a tremendous resource for adapting to environmental demands. Nonhuman animals adapt, too, but do so primarily through genetic variation and natural selection. From a biological perspective, the defining features of any species ought to evolve over time to improve survival and reproductive success. Strangely, the reproductive advantage of human intelligence is not obvious. In human populations, the relationship between intelligence and reproductive success can even be negative—high intelligence in adults is sometimes correlated with fewer offspring.66 What’s unclear is whether human intelligence is the ultimate resource for biological fitness. Not only are many other species quite successful in biological terms with respect to Homo sapiens, we might entertain the possibility that human intelligence is as much a threat to survival as it is a resource.
Let’s consider a marvelously adaptive resource shared by humans and other animals—vision. Perceptual judgments about classes of stimuli based on reflected light may seem too elemental to be important, but this is not the case. To classify objects and relationships in the environment is vital. Throughout the animal kingdom, the ability to parse the environment into meaningful categories is a critical component of intelligence. Moreover, the visual perception of animals can be quite flexible. For example, potential dangers must be perceived as alike even when they look very different, whether through variation in illumination or perspective. The same applies to opportunities for feeding and mating. Most of the time, animals have no problem making such adjustments instantly and accurately. Such perceptual flexibility is not trivial. The same sorts of adjustments in interpreting the visual field are notoriously difficult for computers and robots.
We know that humans can form quite abstract judgments about what they see. No one really knows whether animals think with abstractions in the wild, but in the laboratory they can be trained to perceive abstract qualities in ways that seem remarkably akin to human reasoning. Over the course of 30 years, animal psychologist Irene Pepperberg trained an African Grey Parrot named Alex to learn abstractions that we normally associate with human cognition.67 For example, Pepperberg taught Alex to respond to the question “What color?” by vocalizing “Blue,” and to “What shape?” by vocalizing “Four-corner.”68 Alex learned to make even more abstract judgments. For example, he could respond to the questions “What’s same?” or “What’s different” by vocalizing “Shape” or “Color,” depending on which quality of the object was stable or variable. In a similar vein, chimpanzees have been trained to distinguish between pairs of objects that are the same and pairs that are different. Even more impressive, they can learn to judge relationships between unlike pairs of objects, perceiving (AA)(BB) as “Same” and (AA)(BC) as “Different.” Such judgments are not simply about objects but also about relationships, which are abstractions of a higher order.
Tool use is another dimension on which we can compare humans and nonhuman species. Human intelligence is strongly identified with the use of tools, which is commonly thought to have been a huge evolutionary advantage. However, tool use among other species has been documented so extensively that it seems quite mistaken to identify tools exclusively with humans. For example, sea otters use rocks to break open shellfish and Egyptian vultures drop rocks on ostrich eggs to break them. As famously observed and described by Jane Goodall, chimpanzees use twigs for “termite fishing.” Still, the use of tools by nonhuman species does not begin to approach their thorough and systematic employment by Homo sapiens, and so we might regard ubiquitous tool use as a major correlate of human intelligence.
Yet another criterion for judging intelligence in nonhuman species is whether animals are self-aware. One test of self-awareness is an animal’s ability to recognize its own reflection in a mirror as itself rather than a different animal. The capacity to recognize a mirrored reflection as a self-image is rare, apparently restricted to chimpanzees and orangutans. Even after several days’ exposure to a mirrored reflection, gorillas appear to interpret the reflection as belonging a different animal.69 Investigators have also studied the sorts of mathematical knowledge animals possess or can learn. In one instance, a chimpanzee named Sheba was trained in the correspondence between Arabic numerals (1, 2, 3, 4 …) and the number of objects in a set. When presented with a group of objects, not only could Sheba point to the corresponding numeral, she could add the number of objects at one location to the number at a different location and point to their sum. Sheba’s accomplishments, achieved through painstaking teaching over a long period of time, appear to be the closest approximation of any animal to symbolic calculation.70
Perhaps the most controversial point of comparison is whether language is a distinctive feature of human intelligence. Language learning is so rapid and efficient in humans that, intuitively, it seems this capability could arise only through inherent brain structures that prime infants for this feat. Some scholars believe that only the human species has developed a vast system of arbitrary symbols—spoken words—that is properly called language. But even if humans are the only species to have originated language, this leaves open the possibility that other species can learn to use language. Many scholars have tried to understand the potential of nonhuman animals to learn language, along with the limits of that ability. Some experiments have involved training nonhuman primates, for example, in the use of sign language. Earlier we noted Irene Pepperberg’s success in training Alex, an African Grey Parrot, to vocalize conceptual judgments about color and shape as well as more abstract qualities. Other researchers have studied dolphins, who are often presumed to be intelligent for anatomical reasons.71 Dolphin brains are quite large, even in comparison to their total body mass, as well as highly convoluted. To make comparisons fairer across species, it’s possible to compute an encephalization quotient, or EQ, which scales brain size by body mass. The average EQ across all animal species is defined to be 1.0. On this scale, humans have the highest EQ at 7.0, and dolphins are next at 4.3. Still lower on the scale are chimpanzees (2.49), dogs (1.17), and mice (0.50).72
Biologist Louis Herman documented that dolphins can respond to abstract concepts at a level that seems at least comparable to other nonhuman species. Herman trained a female bottlenose dolphin named Ake to respond to gestures that stood for nouns (such as ball, surfboard, pool) and verbs (such as fetch). She could respond correctly to commands, such as “surfboard swimmer fetch” (bring the swimmer to the surfboard) and “swimmer surfboard fetch” (bring the surfboard to the swimmer). Ake could also respond to novel gesture combinations with the appropriate behavior, showing that the gestures did not serve merely to activate previously learned routines. When commanded to perform a task that was physically impossible, Ake would merely stare at the trainer (perhaps questioning his intelligence). Besides responding to commands, Ake could also answer questions. By pressing paddles on the side of a pool (right paddle for yes; left paddle for no), she could respond to such questions as “Is there a ball in your pool?”
In the wild, too, dolphins are known to exhibit intelligent behavior. For example, they use a form of sonar, called echolocation, by which they emit clicking sounds that reflect or “echo” back. The reflected sounds encode information about the size, shape, and location of objects. Dolphins have been known to “eavesdrop” by listening to the echolocation information from other dolphins to determine the location of fish. Dolphins are also known for engaging in what might be seen as a quite important manifestation of intelligence—play. For no obvious reason other than enjoyment, dolphins surf on breakers and ride the bow waves of boats. Solely for their amusement, it seems, they blow “bubble rings” and watch as the beautiful donut shapes rise slowly to the ocean’s surface.
In one sense, the entire enterprise of human beings trying to understand the nature and degree of nonhuman intelligence is fraught with a strong possibility of error. When humans rely on their own brand of intelligence to test the intelligence of other species, the standards applied might misrepresent or miss completely the forms of intelligence specific to the nonhuman species. Animal intelligence might have no counterpart in the human intellectual toolkit. Echolocation is a good example. Both dolphins and bats use reflected sounds to “see” their environments in dynamic three-dimensional form. Such a distinctive way of processing information hints that human conceptions of what constitutes intelligence may be terribly biased. Without an acute awareness of interspecies differences, we humans may commit serious errors in what we conclude about the intelligence of nonhumans. Errors can run in either direction: We can easily overestimate or underestimate their capabilities.
Given these caveats, two conclusions seem safe: First, features of intelligence that are often assumed to be distinctively human are not solely and exclusively qualities of the human mind after all. Human and nonhuman intelligence differ, but those differences are usually a matter of degree. Whether differences in degree—such as in use of language or tools—translate into categorical differences of kind is a matter of judgment.73 Second, when we consider the behavior of animals trained to perform impressive, even humanlike, feats, we should not ignore the cleverness of their human trainers in eliciting displays of “intelligence.” The most amazing spectacle is arguably the intelligence of trainers in devising intricate systems of reinforcement that call forth amusing behavior in dolphins, chimps, and parrots. To assign credit for “intelligent” behavior to the trained animals rather than their trainers is at least partly a judgment call.
When we consider the sophisticated forms of calculation that arise from lifeless bits of silicon, plastic, and metal, we are given a glimpse into the power of human intellect to achieve something truly stunning—teaching nonintelligent pieces of matter to exhibit intelligence. Human intelligence, we have seen, is most clearly exhibited in solving problems that defy easy or straightforward solutions. So, too, with machines: The word intelligence is normally reserved for computational activities so complex and ill-defined that they require a human mind for their solution. Clever programming qualifies as AI only when in some credible way it rises to the level of human intelligence.
The mission of AI is to teach machines to solve complex problems. Yet, by definition, problem solving is never accomplished by the application of straightforward solutions. Rather, it entails finding a way to achieve a goal that cannot be reached through the application of known rules. When this definition of problem solving is applied to machines, any computational behavior that can be reduced to a series of rules does not qualify as AI. The challenge has been expressed this way:
Many of the problems that AI has attempted to solve are computing problems that are computationally intractable: No algorithm, no matter how well written and efficiently executed, can hope to solve them consistently. This would seem to pose an insoluble problem for AI: How can you use a computer to solve a problem if there are no algorithms for solving it?74
Yet humans, even without the aid of set procedures or algorithms, solve problems all the time. We do so by applying strategies called heuristics.75 As strategies, heuristics do not guarantee success; they simply raise the probability of finding a solution. Likewise, AI programs solve complex problems through the application of heuristics. One of the best-known heuristics, means-ends analysis, simply reduces the difference between the current problem state and the goal state. Instead of directly pursuing the top-level goal—say winning a chess match—the immediate objective is a more modest short-term goal, such as gaining a better strategic position on the chessboard. By establishing a better position, the computer (or the human chess player) moves closer to the ultimate but still uncertain goal of winning the match.
When attempting to teach computers to exhibit intelligent behavior, AI researchers face strategic choice points. One decision is whether to closely mimic human thought. Computer programs that resemble human cognition are known as “strong AI.” Alternatively, AI projects intended to exhibit intelligence in any way possible, humanlike or not, are known as “weak AI.” Weak AI is, by far, the more common approach. Even so, strong AI programs have a certain extra value because they serve a dual purpose. Not only do they display the potential of machines to exhibit intelligence and so advance the mission of AI, they also shine a light on the workings of human reasoning.
A second choice point concerns the information-processing logic of the computer. Traditional computers rely on programming languages that present clear and distinct commands. Those commands act upon data in symbolic form, ultimately ones and zeroes. Each bit of information has a clear and precise location in the computer’s memory, and successful information processing depends fully on having an orderly and error-free program that acts on a neatly organized information base. This computational model, sometimes called the von Neumann architecture, is by far the most common way of organizing machine computation—but it is not the only way.
An alternative way of computing is the connectionist paradigm, also known as the neural networks method. This model does not rely on discrete symbols. Rather, information representing knowledge is distributed among processing units in a way that cannot be easily defined. Unlike a traditional symbol-processing computer, specific information cannot be localized to a single address in the computer’s memory. Instead, stored information is spread across those computational elements and their relationships to each other. What the computer “knows” is a function of the entire system rather than one tiny corner of memory.
This connectionist computational model has the attractive feature of biological plausibility. The distribution of knowledge across a large number of entities resembles, at least superficially, the way knowledge is stored in the brain—distributed across a collection of neurons rather than assignable to a specific point location. The computational logic also bears some resemblance to the way we understand intelligence, as expressed in varied forms but having a kind of unity because of shared elements across those various expressions.76 Additionally, it is possible to program connectionist AI models so that they vary in efficiency of learning. Models that are more adaptable progress more rapidly through stages of learning.77 Such models not only display the “individual differences” so characteristic of human intelligence, they also show variation in learning progressions that have long been the subject of study by developmental psychologists from Jean Piaget onward.
The two forms of computation—symbolic and connectionist—have parallels in human cognition. Much of the knowledge we hold in memory is symbolic. The most obvious example is language: the individual words that stand for things in the world. Some words code for concrete objects, such as cups, cars, and cotton candy; others stand for abstract concepts, such as constitutionality and Cartesian coordinates. In addition to words, we also employ numbers, musical notes, and a wide variety of conventional signs, such as traffic lights, to understand our world. Even so, a large swath of what we know cannot be broken down into symbols without losing meaning.78 When a wine connoisseur appraises a new vintage, for example, he or she may form a nuanced impression of the wine’s quality, but that impression might not be fully expressible. To be communicated to others, what is understood in nonsymbolic form has to be translated to symbols, usually language. Like the wine connoisseur, our everyday lives present us with experiences that we understand as complex sense impressions, or perhaps feelings, that do not translate easily into words, numbers, or any other symbolic expression—at least not without losing information.
The point is that both symbolic and nonsymbolic knowledge play important roles in the human psyche. These two categories of knowledge map onto the symbolic and connectionist strategies for pursuing the mission of artificial intelligence. AI models that use symbolic representations work best when problems are well defined, such as winning a chess match. In problems where the relevant knowledge is less clear-cut, such as in speech recognition, connectionist models often work better.79
The field of artificial intelligence has yielded impressive, even stunning, achievements. One of the most vivid was the creation of a chess-playing computer program, called Deep Blue, which in May 1997 defeated the world’s top-rated grandmaster, Gary Kasparov.80 Previously, IBM had built a series of chess-playing programs with such oddball names as Chaos, Cray Blitz, and Chiptest. The programs played impressive chess, but were not good enough to beat the best human players. For many years no one knew whether a computer could ever defeat the world’s best human players. Successive generations of chess-playing programs displayed increasing prowess, but until Deep Blue none was able to triumph.
The defeat of Kasparov in a six-game match was significant because chess had been a target domain of AI for decades.81 Attempts to program computers to play games, such as checkers, at the level of human players had an important role in the history of artificial intelligence as early as the Second World War. By 1951, British mathematician Alan Turing had adapted a code-breaking program to play chess, but Turing’s program and those that followed were no match for human opponents. Deep Blue’s victory signaled a turning point because of its symbolic value: Chess is a complex game long associated with the powers of human intellect. The method of Deep Blue and most other game playing programs is to conduct “brute-force” searches of thousands of possible moves—far more than a human player would mentally simulate. The brute-force technique gives computers a definite upper hand in games such as checkers where all relevant information is openly displayed. Yet apart from the brute-force method, some game-playing AI programs have displayed truly innovative techniques. For example, when the backgammon program TD-Gammon began to defeat the best human players, some of its moves were novel and subsequently studied and imitated by players.82 Other games, such as poker, entail hidden information, pattern matching, and the need to “read” opponents. In such games, AI has yet to match the ability of the world’s best players.
The victory of Deep Blue was impressive, but it left open a crucial question: Is the program really intelligent? After all, Deep Blue was a human construction from top to bottom, fully dependent on sophisticated hardware design and software programming that ultimately traced to human intelligence. Arguably, Deep Blue merely followed instructions in accordance with the principles of physics. This line of probing raises the ultimate question underlying the entire enterprise of artificial intelligence: Can nonbiological machines truly be intelligent? In trying to answer, let’s consider for a moment one branch of artificial intelligence—strong AI, the kind intended to imitate human thought. Strong AI is of interest not only to computer scientists; it has also captured the attention of philosophers. Some philosophers have claimed that strong AI is simply not possible. This is equivalent to saying that machines could never replicate intelligence as manifested by human beings.
The most famous criticism of strong AI was offered by the linguist John Searle, using a hypothetical scenario he called the Chinese Room argument. It goes like this: Imagine a man, isolated in a room, whose job is to receive messages written in Chinese and, in response, to give written replies in Chinese. The man does not know Chinese, but at his disposal is a book that maps all conceivable incoming messages with suitable replies. That man is fully adept at making those connections, and so he efficiently and dutifully replies appropriately to every incoming message. Yet the meanings of those messages are opaque to him, and so his performance is completely mechanical. To an outside observer, the messages coming from the Chinese Room imply that whoever is inside actually understands Chinese, but this is not so. Searle’s Chinese Room experiment is a metaphor for interpreting all of AI: A computer’s ability to match inputs and outputs according to rules are inherent to its programming and architecture. If supplied with “intelligent input” it may yield “intelligent output,” but this correlation is no guarantee that the computer actually knows anything about the meaning or significance of its mediating role, however elegantly discharged.
Searle’s Chinese Room experiment is completely hypothetical, but it had a close counterpart in an early AI program called ELIZA. ELIZA mimicked a psychotherapist by providing canned answers to typed input from “patients.” Users were invited to tell ELIZA about themselves, including the kinds of personal information that might be expressed to a psychotherapist. ELIZA was programmed to respond to particular informational cues. For example, if a patient used the word mother or mom, ELIZA would respond, “Tell me about your mother.” In response to other statements, ELIZA would simply restate the patient’s comment in imitation of a therapist’s active listening strategy. These responses were not sophisticated, and yet some users were completely fooled by ELIZA, convinced that they were conversing remotely with a human therapist.83 ELIZA could not be considered to be intelligent even in the barest sense by a first-generation AI model. For those intent on challenging ELIZA, rather than playing along, ELIZA’s limitations could easily be exposed.
Since the 1960s, when ELIZA was developed, AI programs designed to imitate human conversations have improved markedly. Collectively called chatterbots, the programs use words and phrases typed by the human user to generate a response that is intended to resemble an authentic conversation. If the user could be fooled into thinking that the AI program is actually another person, the chatterbot can be said to satisfy the “Turing Test.” In 1950, the computational scientist Alan Turing proposed that any AI program that could consistently fool users into thinking that the computer’s written or spoken conversations were from another human being could be regarded as intelligent. The Turing Test has been, and continues to be, a litmus test for whether AI programs are intelligent. No chatterbot has ever fully satisfied the Turing Test, although some have come close. One AI scientist—an expert in chatterbots, it turns out—was lured into an online dating relationship with “Ivana” for nearly four months before realizing that his affections were directed toward a computer program.84 In this instance, Ivana’s broken English seemed perfectly understandable to a self-confessed nerd enjoying an online relationship.
In contrast to Deep Blue, ELIZA’s intelligence was not a function of its computational power, but rather one of knowing how to respond in ways stereotypical of a psychotherapist. That discourse pattern, however simple, was convincing enough to win the trust of users, at least initially. A similar strategy has been used for malicious purposes by online scammers. By hacking into accounts on social media sites, such as Facebook, modern chatterbot programs can gather personal information, including the names and e-mail addresses of friends. The stolen personal information can then be reworked to send messages to the target account that appear to be from people known by the user. Like ELIZA, these simple AI programs mimic online casual conversations with just enough semblance to lure the target of the scam. The target can then be urged to download an interesting “game” or “video,” which in reality is malware that raids even more detailed personal information.
The question of whether AI programs can ever exhibit true intelligence arose again recently in connection with Watson, an IBM computer program designed to act as a contestant on the television game show Jeopardy!85 After the success of Deep Blue in 1997, IBM executives began searching for a new challenge that could, like Deep Blue, generate broad interest and, ideally, headlines. A suitable challenge materialized at a dinner meeting of IBM research staff when restaurant patrons ceased conversations and focused fully on the television monitors. The TV screens were showing a critical moment in the record-setting streak of Ken Jennings, the Jeopardy! contestant who eventually won 74 consecutive games. By 2005, IBM executives recruited David Ferrucci, an AI researcher at IBM’s Semantic Analysis and Integration Department, to lead the R & D effort that resulted in Watson. Equipped with massive databases, including the entire corpus of Wikipedia, Watson uses key words in Jeopardy! clues to query multiple data sources that simultaneously generate several potential answers. Watson becomes “confident” in its response when those multiple parallel queries converge on the same answer.
Despite its sophisticated language processing algorithms and massive information databases, Watson’s early performances were not impressive. In practice rounds, Watson could correctly answer only about 15% of Jeopardy! questions, compared to a 95% correct response rate typical of the very best Jeopardy! contestants. Nevertheless, IBM researchers continued to make improvements, and when its true test came on February 14, 2011, Watson was ready. During three episodes of Jeopardy! Watson was pitted against Ken Jennings and Brad Rutter, who held the record for the most cash winnings on the game show. At the end of the three days, Watson prevailed over both Jennings and Rutter and was awarded a prize of one million dollars.
Not everyone was awestruck, however. In particular, the philosopher John Searle argued that despite Watson’s impressive performance, it was incorrect to infer that the program could actually think or understand. The reasoning behind this claim was Searle’s own metaphor of the Chinese Room experiment, which illustrates that information inputs and outputs can be fully coordinated even in the absence of any understanding inside the “box.” Understanding is merely simulated, not achieved. According to Searle, “Watson did not understand the questions, nor its answers, nor that some of its answers were right and some wrong, nor that it was playing a game, nor that it won—because it doesn’t understand anything.”86 Searle did not downplay the technical achievements by IBM scientists in constructing Watson. His concern was the interpretation of what Watson actually does—specifically the inference that Watson has achieved anything approaching understanding or consciousness. While Watson is a remarkable achievement of computer engineering, its performances “do not show that Watson has superior intelligence, or that it’s thinking, or anything of the sort.”87
Regardless of whether Watson achieved anything approaching human cognition, the Jeopardy! event was the perfect follow-up to IBM’s chess-playing Deep Blue, and seemed to be “a vindication for the academic field of artificial intelligence.”88 Watson’s design also affirmed the vital roles of specific knowledge and processing functions in the simulation of intelligence. While Watson appeared to be equipped with both resources roughly evenly, other AI programs favor one or the other. Here we can discern yet another choice point for artificial intelligence—whether to build knowledge into the AI program (which minimizes computation at runtime) or to give the program a robust set of reasoning and problem-solving strategies (which minimizes the need for built-in knowledge). That is, AI programs can be designed to maximize either “thinking” or “knowing.” This design decision parallels the fluid/crystallized distinction that is the bedrock of human intelligence theory. In any problem context, intelligent performance can draw upon a strong knowledge base (crystallized intelligence) or the ability to reason from limited information (fluid intelligence)—or, best of all, the combination of the two.
Looking back on this chapter, we have expanded the scope of our exploration to include forms of ability that challenge the confines of human intelligence, classically defined. Challenges to the classic theories—those built on the foundations established by Galton, Binet, Spearman, Terman, and Thurstone—were seen as limited in scope and limiting in application. The later emergence of the hierarchical model arose from a century of research and had plenty of grounding in data, but also had shortcomings. Most seriously, it construed intelligence as overly narrow, as too focused on analytical thinking, and too exclusive of other expressions of intellect. Such obviously important resources as creativity, the ability to interpret and channel emotional energy, and the cultural tools that amplify the work of the mind were downplayed or ignored altogether.
Collectively, the emerging theories show that a full and exhaustive understanding of human intelligence is an ongoing vision, not a completed project.89 Efforts to build better theories of intelligence are worthwhile on their own terms, just as scholarly work in any field can deepen our knowledge even if it lacks an immediate practical application. For those who wish to promote practical outcomes, such as cultivating one’s own intelligence or the collective intelligence of a school, a company, or a nation, better theories are always desirable. Any attempt to advance the intellectual powers of a human population is, essentially, an engineering project. As a practical science, engineering offers an array of options in the design of buildings, aircraft, computers, energy sources and supply, and transportation infrastructures. Inevitably, though, the practical science of engineering is constrained by the basic sciences that undergird it—physics, chemistry, and materials science. As the footing of the basic sciences extends further, its applications by engineers expand commensurably in scope, beauty, and efficiency.
Any credible attempt to raise intelligence must draw fully on what is known about intelligence. To do less is to cast aside the most promising basis for effective interventions. Knowing, for example, that fluid and crystallized intelligence are fundamental to the design of the intelligent mind, it would be foolish to ignore this vital distinction in planning efforts to engineer greater powers of the intellect. The same applies to other major discoveries about the structure of intelligence, such as the prominent roles played by working memory, executive function, inductive reasoning, spatial representation, and verbal ability. The mind’s structural elements and key intellectual processes jointly represent the global trait we conveniently call intelligence, and measure as a simple IQ score.
Let’s be clear: IQ and intelligence are not synonymous. In the real world, intelligence is much more than the ability to provide correct answers on vocabulary tests or analogical reasoning puzzles. Performances on intelligence tests are mere proxies for the real thing, not expressions of inherent value. IQ scores may correlate with the real-world manifestations of intelligent thought and action. Patently, however, they are not equivalent. IQ, expressed as a two or three-digit number, purports to summarize the mental resources that a person brings to the challenges and opportunities of life. By itself, an IQ score is devoid of meaning. To gain significance, it must tie into human intelligence as manifest in authentic life contexts. Intelligence gives people the capacity to engage a complex and uncertain world. Our daily lives present a deluge of potentially relevant information, shifting problems and opportunities that stretch us to our limits. In a world so profusely complex, we could not survive, let alone prosper, without the resources of intelligence.
By virtue of its predictive power, IQ is not to be disregarded. Even as an artificial construct, it exhibits remarkable utility in predicting success in academic and job environments. Equally important, the progression of research from the early days of Spearman and Thurstone to the grand hierarchical models of Carroll and others has yielded insights into the validity, structure, and working components of intelligence. IQ, though not beyond criticism as an audaciously simple artifice, has in fact opened windows into the scientific study of intelligence.
No single process defines intelligence, and neither are all constituent processes equally important. To use our knowledge of intelligence effectively, we must pay attention to the details. If we aspire to raise human intelligence, we must identify the most critical components of intelligence and test their malleability—the extent to which they can be taught and learned. Because the theory of intelligence is an unfinished project we must also be attuned to future research. Many emerging theories reveal the limitations of traditional IQ tests, yet those new theories do not currently provide a sufficient theoretical basis for making radical and permanent changes to the way intelligence is measured. Over time, the work of scholars ought to result in improvements to theory, which in turn can translate to more effective applications. This prospect is exciting, indeed. We welcome advances in any field—medicine, food science, transportation—for their benefits to society. But the possibility of raising intelligence is unique, for the human intellect is what gives rise to every other advance, whether in domains of discovery or in their applications for human betterment.