Can computers understand whether comprehension comprehends understanding?
The Financial Times announces a milestone in the history of education in particular and humanity in general. After triumphing over grandmasters in chess and expert players of Go, artificial intelligence (AI) now outperforms humans in a recently invented human activity our schools call reading comprehension. Here is how the FT describes the achievement: “Now computers are taking a step into a very human territory: the reading comprehension test, scourge of schoolchildren everywhere.”
Not only has computing evolved, but so has reading itself, which means it’s time to provide its new 3D definition:
A human activity formerly encouraged in education to develop the mind, students’ critical faculties, their cultural sense, rhetorical skills and mastery of style, currently simply to memorize facts and emulate the performance of computers
The culture of competition that, in recent centuries, has accompanied the rise of capitalism and the progress of the industrial revolution has spread so broadly and deeply that it has exhausted the public’s interest in humans competing with other humans, except of course in the domain of sports. To make the news, outside of physical sports and presidential elections, the thrill of competition can increasingly be found in the rivalry between humans and computers.
Our sociopolitical culture has elevated meritocracy and the idea of competitive success to the highest place in our scale of values. Competition accomplishes two major goals: It sells and selects. It attracts eyeballs and sells products in a commercial culture increasingly dominated by media and it selects the stars and talents whom an adoring public of consumers will passively celebrate. Nothing fascinates the public more than competition, which at the same time is designed to distract its attention from the true, underlying issues. We see it in politics every day. We can also see it in stories like this one about artificial intelligence.
When a competitive test tells us computers are outperforming humans, the public will usually respond with delight or disappointment, depending on whom they were rooting for. But, distancing ourselves from the thrill of the competition, if we still possessed even a vestige of critical thinking, we should be wondering what this tells us about reading comprehension tests in themselves and the skill they claim to be testing. The Financial Times was correct to call it “a scourge,” whose original meaning is “a whip with which to punish people,” in this case children.
Reading comprehension tests are a by-product of the Industrial Revolution and the institution of mass education. They operate on the unstated assumption that a text is essentially a corpus of information that can be analyzed into discreet units rather than an expression of complex notions and concepts embedded in a web of intricate cultural connections and spoken or written by a sentient human being. It is no wonder that literature — once the foundation of education — has now been demoted to the status of a trivial art, more or less a subcategory of entertainment. An art that, according to the new orthodoxy, fails in its triviality to compete with the high seriousness of STEM (science, technology, engineering, mathematics), the only subjects that will enable the next generation to get jobs as programmers in Silicon Valley or stock market analysts on Wall Street.
Always interested in competition, The Daily Devil’s Dictionary would therefore like to propose a new reading comprehension test for artificial intelligence. The text to be “comprehended” would be a novel six words-long, which, according to what is in most likelihood an apocryphal anecdote, Ernest Hemingway produced after betting friends he could do it: “For sale: baby shoes, never worn.”
It is conceivable that someday artificial intelligence will integrate forms of logic that enable it to speculate about recondite meaning, but it would first have to identify the presence of recondite meaning.
For centuries, academic institutions and literate society as a whole — at least from the Vedas in India to the Enlightenment in Europe — were so backward that they blithely neglected what has now been defined as the vital modern skill of reading comprehension. It could be claimed that modern education began when testing the young for reading comprehension began, though it isn’t even a century old. Having used it for a good 70 years to test generations of students, we are now moving on to using it to test the computers that will be replacing on the job market students no longer capable of reaching the computers’ level of comprehension.
In the meantime, it might be worthwhile to step back and think about what we mean by reading itself. We have clearly forgotten about the role reading played in building the complex cultural foundations on which our current increasingly binary techno-industrial civilization still depends. By reducing the aim of reading to comprehension rather than understanding, we cast the reader into the role of an auditor examining the books rather than a mind exploring the meaning and capacity for resonance of a book.
But aren’t comprehension and understanding the same thing?
The difference between these two words tells the story of a massive cultural shift. Comprehension contains the idea of an act of “prehension,” which means taking or grasping in Latin. How appropriate in our consumer culture! We seize and grasp what interests us and then take it away, maybe to consume it, maybe to store it in a safe place.
By contrast, understanding literally means to stand in a position where we can see the complete text, similar to the German verb, verstehen¸ meaning to stand before. Standing under implies exploring the full breadth and the depths of the text, following its roots, gazing at its branches, not just snagging the tempting fruit in the tree.
The real paradox — and the reason such tests are felt as a scourge by students while leaving computers indifferent — lies in the fact that while all texts contain information, few texts that actual human beings produce are designed as repositories of facts. Even dictionaries and encyclopedias are culturally complex artifacts, as our Devil’s Dictionary attempts to prove on a daily basis. Any text we attempt to comprehend is composed of layers of meaning always conveying multiple associations, implying stories, emotions, states of being, forms of awareness and enlightenment that only use the details of the text — the facts — to produce their effect.
So how big is the Amazon rainforest? So big that its health concerns the planet and it’s big enough to house hundreds of tribes, with an equal diversity of languages and customs. Oh, the test was expecting a number. Probably in square miles, not kilometers, because that’s what the original text said. I’m sorry, I’ll just have to go back to my textbook. I may have understood but I certainly failed to comprehend.
And remember, it’s the state of the baby’s shoes that tell the story. As for you, AI, I’m sorry, the story isn’t in fact about footwear.
*[In the age of Oscar Wilde and Mark Twain, another American wit, the journalist Ambrose Bierce, produced a series of satirical definitions of commonly used terms, throwing light on their hidden meanings in real discourse. Bierce eventually collected and published them as a book, The Devil’s Dictionary, in 1911. We have shamelessly appropriated his title in the interest of continuing his wholesome pedagogical effort to enlighten generations of readers of the news.]
The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.
For more than 10 years, Fair Observer has been free, fair and independent. No billionaire owns us, no advertisers control us. We are a reader-supported nonprofit. Unlike many other publications, we keep our content free for readers regardless of where they live or whether they can afford to pay. We have no paywalls and no ads.
In the post-truth era of fake news, echo chambers and filter bubbles, we publish a plurality of perspectives from around the world. Anyone can publish with us, but everyone goes through a rigorous editorial process. So, you get fact-checked, well-reasoned content instead of noise.
We publish 2,500+ voices from 90+ countries. We also conduct education and training programs on subjects ranging from digital media and journalism to writing and critical thinking. This doesn’t come cheap. Servers, editors, trainers and web developers cost money.
Please consider supporting us on a regular basis as a recurring donor or a sustaining member.
Support Fair Observer
We rely on your support for our independence, diversity and quality.
Will you support FO’s journalism?
We rely on your support for our independence, diversity and quality.