From Jane Austen to Chatbots: Using Conversation to Judge Intelligence

Naomi Michalowicz //

Chatbots, those little pop-up virtual assistance you encounter at the bottom of every page on every website of every company, who cheerfully ask “what can I help you with?”, are not as smart as we’d like them to be. Often frustratingly obtuse, the virtual assistant is incapable of answering your questions or directing you to the page you are looking for, leaving you longing to speak to a real, flesh-and-blood human being, who—however dense they may be—might actually be able to tell you how much you owe in toll fees, for example, and where to go in order to pay them. According to a recent article in The New York Times, in costumer-service circles this phenomenon is given the appropriately ominous title, “the spiral of misery.” 

Why are we so frustrated with the stupidity of chatbots? Our smartphones are often not as smart as we would like, and this stupidity may be inconvenient—but not in the same order of magnitude, and definitely not as existentially frustrating as the spiral of misery. The names we give them suggest that we expect smartphones to be smart, and chatbots, not necessarily. Why is it, then, that our expectations of our smartphones are not as high as they are of chatbots?  

The answer lies, I believe, in the context: it is not so much the presumed smartness of the virtual assistant or device which dictates our expectations, but rather, it is the cultural significance of conversation as a particular type of interaction, and the conversational pretense of the chatbot, which creates such high expectations—and thus leads to such depths of disappointment.

//

Other than formal assessments, such as standardized tests, I think it’s reasonable to state that conversations are the primary arena in which we form judgment of other people’s intelligence. After all, the skills required for making good conversation—the ability to make causal and logical connections, the ability to comprehend and respond to ideas, having a working knowledge of multiple topics and the ability to recall details from memory, having a sense of humor, an extensive vocabulary, and, importantly, thinking and responding quickly—are all mental skills and attributes we think of as constitutive of a person’s intelligence.

In 1984, Australian cognitive psychologist Jacqueline Goodnow suggested that intelligence may be more productively understood if we were to think of it not as “a quality people possess in varying degrees” but rather as a “judgment—a quality attributed—when people display ‘intelligent behaviours’” (391). This approach helps us think of intelligence as something historically and culturally contingent, rather than inherent and distinct from social circumstances, and so calls on us to pay attention to the values underlying our assumptions about the nature and makeup of intelligence. Perhaps the most potent such assumption is that of the mind-body distinction, and the assumption that intelligence is strictly a property of the mind: this makes conversation, the “oral exchange of sentiments, observations, opinions, or ideas” (Merriam-Webster), of things that live in the mind, a much more convenient platform for the demonstration and assessment of intelligence than, say, a soccer game, despite the latter also requiring a high degree of skill and sophistication.1

As with more formal assessments of intelligence, some people prove more or less capable, and so judged more or less intelligent. And perhaps it is the inverse of a successful conversation, the dull, prolonged, stilted conversation, which sheds light on which qualities we really do value and which less so: if in the context of, say, standardized testing it is the quick recall of a lot of stored information that will lead to a high score—and therefore, a positive judgement of a person’s intelligence—in conversation we are more likely to appreciate someone who is thoughtful and selective in the information they present. No one wants to get trapped talking to that person, the one who will bombard you with endless facts and bits of information of questionable relevance, however impressive may be their knowledge of Harry Potter fan culture, or engine-building, or medieval ships. 

Because we tend, as linguistics professor Jonathan Culpeper as noted, “towards trying to fit … information about people or characters into pre-formed social schemata” (266), we form impressions more easily when traits are cohesive rather than contradictory. Just as we would find it hard to form a unified impression of someone who is very socially active but also not very talkative, because this particular mix does not match the pre-formed schema of an outgoing person, so we would be more likely to describe as intelligent someone who is thoughtful and selective, than someone incapable of reading the obvious signs of boredom evinced by his listeners—a particular type of stupidity which might detract from what would otherwise be a positive assessment of that person’s intelligence. This is particularly important because in formal contexts of assessment, inability to respond to boredom might not play into the overall evaluation of a person’s intelligence, or, at best, be relegated to different category altogether, such as “emotional” or “social” intelligence.

//

When reading novels, despite the fact that we are often granted access to the interiority of characters’ minds, we are often called to make judgments about characters in the same way that we do about real people: by observing their behaviors, listening to them speak, and forming impressions based on our observations. And in the same way we rely on conversation to assess the intelligence of our friends, coworkers, and that one person who corners us at a party, we rely on fictional characters’ conversational skills in our assessment of their intelligence. 

As an exercise, I provide here two snippets of conversation taken from Jane Austen’s Pride and Prejudice, the first between Mr. and Mrs. Bennet, and the second between their daughters Elizabeth and Jane, following the first meeting of the Bennets with Mr. Darcy and Mr. Bingley. Because conversation is an event packed with information about the speakers’ mind, we can rely on these excerpt to form an assessment of the intelligence of Mrs. Bennet, and compare it to that of her daughter Elizabeth (the first speaker in the second excerpt):

1.

“Oh! my dear,” continued Mrs. Bennet, “I am quite delighted with him. He is so excessively handsome! and his sisters are charming women. I never in my life saw anything more elegant than their dresses. I dare say the lace upon Mrs. Hurst’s gown—”

Here she was interrupted again. Mr. Bennet protested against any description of finery. She was therefore obliged to seek another branch of the subject, and related, with much bitterness of spirit and some exaggeration, the shocking rudeness of Mr. Darcy.

“But I can assure you,” she added, “that Lizzy does not lose much by not suiting his fancy; for he is a most disagreeable, horrid man, not at all worth pleasing. So high and so conceited that there was no enduring him! He walked here, and he walked there, fancying himself so very great! Not handsome enough to dance with! I wish you had been there, my dear, to have given him one of your set-downs. I quite detest the man.” (216-217)

2.

“Well, he certainly is very agreeable, and I give you leave to like him. You have liked many a stupider person.”

“Dear Lizzy!”

“Oh! you are a great deal too apt, you know, to like people in general. You never see a fault in anybody. All the world are good and agreeable in your eyes. I never heard you speak ill of a human being in my life.”

“I would wish not to be hasty in censuring any one; but I always speak what I think.”

“I know you do; and it is that which makes the wonder. With your good sense, to be so honestly blind to the follies and nonsense of others! Affectation of candour is common enough;—one meets with it everywhere. But to be candid without ostentation or design—to take the good of everybody’s character and make it still better, and say nothing of the bad—belongs to you alone.” (217)

These two short passages are enough to convey to the reader the gulf between the Elizabeth and her mother. Whereas Mrs. Bennet is completely uninterested in her interlocutor, Lizzie directs her comments to her sister by making her sister the object of her conversational attention; where Mrs. Bennet is indignant, Lizzie is wryly humorous; Mrs. Bennet is focused on the superficial behavior of Mr. Darcy, describing it with little imagination or nuance and drawing limited conclusions from it—“he walked here, he walked there, fancying himself so very great”—while Lizzie pays attention to the nuances of her sister’s character, and is interested in inferring abstract principles from her observations. Deductive reasoning, emotional attunement and consideration of the listener, sense of humor—all signal that Lizzie is smarter than her mother, an impression corroborated by the narrator: Mrs. Bennet, we learn in the very first chapter, is “a woman of mean understanding” and “little information,” while her daughter, on contrast, is a girl of remarkable “quickness of observation.” 

It is worth noting that in Mrs. Bennet’s case, her stupidity manifests as vulgarity, and is inseparable from her inferiority of social station, exemplified in her obsession with having her daughter marry up, and her violent response to whatever may stand in the way of that goal. Because conversation is an encounter between two or more people—or, in the case of fictional conversation, between two characters and a reader—the relative position of each participant must inform, to a certain extent, the judgement. As Pride and Prejudice goes to great lengths to show, class and social standing are not external characteristics, but shape the way the very minds of the characters work. Mrs. Bennet is a bad conversationalist because she is narrow-minded and vulgar; Mr. Darcy, Elizabeth’s stiff-upper-lipped suitor, can afford to be a bad conversationalist because his aristocratic position means there are no material consequences to his stilted and abrasive manner of speaking (until of course, there are; this is why the novel is a romantic comedy and not a naturalist drama). 

In 19th century novels the key word in describing intelligence is not “intelligent” or “smart” but rather “clever”—a term that, unlike the preceding examples, evokes the inseparability of the judgment of intelligence fro class connotations. And so we find cleverness as both a marker of an aristocratic mind and as an invaluable tool in the social climber’s toolbox: the wit of Lord Henry in Oscar Wilde’s The Picture of Dorian Gray, with his delightful quips and scandalous paradoxes, relies on the irreverence only someone of complete financial and social security can afford. And on the opposite end of the social ladder, pretty and ambitious Beck Sharp of William Thackeray’s Vanity Fair relies on her cleverness and conversational charm to make her way in the world in one of literature’s most fascinating accounts of social climbing. 

One would think that conversationalist artificial intelligence would be exempt from this complex of class associations. In some obvious ways, it is indeed exempt: we do not think of AI as participating in human society, and so we do not expect of it to utilize conversation in order to demonstrate class superiority. We are not likely, for example, to accept a statement such as Lord Henry’s comment “I like persons better than principles, and I like persons with no principles better than anything else in the world” as a paragon of aristocratic wit, should it come from a chatbot. However, we do expect AI to be able to possess specific kind of conversational skills: of listening, interpretation, and quickness of response—those that characters in lower social standings often have to rely upon (such as poor Fanny Price in Austen’s Mansfield Park).2 In the words of that NYT article, the improvement of chatbot technology means that “customer service chatbots will become more intelligent, more conversational, more humanlike and, most important, more helpful.” 

//

That equivalence between the intelligent, the conversational, and the humanlike, so casually stringed together in the NYT article, reflects a deep commitment to conversation as the domain in which intelligence is being assessed. This is why conversation is also the foundation of one of the world’s most popular thought experiments, which still serves as the ultimate test of AI’s “true” intelligence. Alan Turing’s “imitation game,” popularized as the Turing Test, offers a pragmatic way of testing the successful creation of artificial intelligence. If we cannot distinguish between human and machine intelligence, goes Turing’s argument, then we can say that the machine, for all intents and purposes, is as intelligent as the human. And of course, the way this test takes place is through conversation: an examiner engages in two conversations, one with a machine and one with a human, both of which attempt to convince the examiner, through they manner in which they converse, that they are the human participant. So far, no AI has consistently succeeded in passing the Turing Test.3 And that is the whole point: it doesn’t matter if AI is capable of cognitive feats matching, or surpassing, those of humans—if it wins games against chess masters, or even make scientific discoveries4—as long as it cannot hold a conversation, we will never accept it as intelligent. 

//

06.27.22 Edit: In response to the recent claims by a Google engineer about a sentient AI, the news website The Conversation published an article which calls attention to the human tendency to equate conversational ability with intelligence. This article elaborates both on the risk inherent in this equation in terms of attributing intelligence where there is only fluency, and the bias to which it gives rise, as people tend to associate regional and foreign accents, or stuttering, which a lower intelligence.

Image: Augustin Théodule Ribot: The Conversation. Wikimedia Commons.

Notes:

  1. A lot has been said about the cultural bias which overvalues linguistic and mathematical skills and undervalues other forms of intelligence, starting with Howard Gardner’s 1983 Frames of Mind: The Theory of Multiple Intelligences.
  2. Fanny is the example used by cognitive narratologist Lisa Zunshine in her discussion of the class inequality of what she calls “mindreading”—the ability to accurately gauge another person’s state of mind. 
  3. At least by March 7, 2022, according to an article in Big Think, “no computer has ever passed the Turing test.”
  4. See “Planarian Regeneration Model Discovered by Artificial Intelligence” Tufts Now, 2015.

Works Cited:

  • Austen, Jane. The Complete Novels. Penguin Classics, 2006.
  • Gardner, Howard E. Frames of Mind: The Theory of Multiple Intelligences. Basic Books, 2011.
  • Goodnow, Jacqueline J. “On Being Judged ‘Intelligent.’” International Journal of Psychology, vol. 19, no. 1–4, 1984, pp. 391–406. Wiley Online Library, https://doi.org/10.1080/00207598408247538.
  • Johnson, Stephen. “The Turing Test: AI Still Hasn’t Passed the ‘Imitation Game.’” Big Think, 7 Mar. 2022, https://bigthink.com/the-future/turing-test-imitation-game/.
  • Lohr, Steve. “Ending the Chatbot’s ‘Spiral of Misery.’” The New York Times, 3 Mar. 2022. NYTimes.com, https://www.nytimes.com/2022/03/03/technology/ai-chatbot.html.
  • Wilde, Oscar. The Picture of Dorian Gray. Edited by Robert Mighall, Revised edition, Penguin Classics, 2003.

Keep reading

%d bloggers like this: