Why We Are Always Being Misunderstood on the Internet

Language does not have a fixed relationship to reality bound by logic. And neither do many humans. JL

David Auerbach comments in Slate:

It was assumed that computers could understand language by interpreting them as propositions that were either true or false. The efforts in this direction have been remarkably unsuccessful. These difficulties are exactly why Google succeeded—by ignoring semantics, sticking instead to whatever it could glean without trying to understand the meaning of words or sentences.

The best class I took in college was on the philosopher Ludwig Wittgenstein. Until that point, I had avoided philosophy of language as simply being too esoteric and hermetic to be of use. David Pears, a prodigious yet modest and approachable figure visiting from Oxford, changed my mind. In large part because of Pears’ instruction, Wittgenstein’s philosophy has been directly relevant to my thinking about computer science, artificial intelligence, and cognitive science. When other scholars were thinking that language and thought could be reduced to a universal, logical language, Wittgenstein turned the matter to practical questions and raised incredibly inconvenient questions that gained traction in artificial intelligence in the 1970s, 40 years after he was working on them.

Wittgenstein, who lived from 1889 to 1951, is most famous for a handful of oracular pronouncements: “The limits of language are the limits of my world.” “Whereof one cannot speak, thereof one must be silent.” “The human body is the best picture of the human soul.” They sound great; they are also hopelessly mysterious except in the context of Wittgenstein’s entire philosophy. Or more accurately, philosophies. Wittgenstein’s writings, broadly speaking, divide into two periods, and in the second he more or less wholly rejected the underlying conception of the first. In his first lecture, Pears began: “Some philosophers fly; others struggle to crawl.” Wittgenstein flew, then crashed to earth and crawled thereafter.

(Since pretty much no one can agree on anything about Wittgenstein, I’m going to present things in the spirit of Pears’ interpretation, with the caveat that you could probably find a philosopher somewhere who would disagree with every following sentence.)

Wittgenstein’s first period, culminating in 1921’s Tractatus Logico-Philosophicus (which Pears had co-translated), drew heavily on Bertrand Russell’s work in philosophical logic and made a huge impact on the logical positivist movement of the time, which would later in turn influence computer science, artificial intelligence, and linguistics. The Tractatus makes an ambitious and ostensibly definitive attempt to chart out the relationship between language and the world. Alongside Russell’s work, it was tremendously influential on logicians, yet Wittgenstein later ended up rejecting one of its central premises: that our linguistic statements depict true or false states of affairs, and that formal logic provided the structure that regulates our construction of these statements. Language and the world share logical form, which is also the form of reality. This attempt to regiment language as formal logic went on to be an article of faith for many computer scientists and cognitive scientists for decades, as well as exerting a foundational influence on Noam Chomsky’s linguistics.

But after a 10-year hiatus from philosophy, during which he was a schoolteacher and co-designed and built an austere house for his sister, Wittgenstein came to change his mind. Language did not have such a fixed, eternal relation to reality bound by logic. The process of “measuring” the truth of a statement against reality was neither objective nor cleanly delineated. The meaning of what we say can’t be abstracted away from the context in which we say it: “We are unable clearly to circumscribe the concepts we use; not because we don’t know their real definition, but because there is no real ‘definition’ to them,” Wittgenstein wrote. Instead, our speech acts are grounded in a set of social practices.

The idea of words having relative meanings was not new, but Wittgenstein pioneered the controversial linguistic conception of meaning-as-use, or the idea that the meanings of words, relative or not, cannot be specified in isolation from the life practices in which they are used. Instead, language should be studied from the starting point of its practices, rather from abstractions to syntax and semantics. As Wittgenstein put it, “Speaking a language is part of an activity, or of a form of life.”

Unfortunately, this makes the study of language considerably more difficult, since examining the meanings of words now requires not just verbal definitions, but analyzing the whole “language-game” of situations and practices to which they are attached. Wittgenstein introduces the idea of “following a rule” to describe what we do when we use a word in everyday life, but what it is to “follow a rule” is notoriously difficult to pin down. Pears’ interpretation was that following a rule was akin to a judge applying a law in a case: Its validity depends on the past instances of how that rule was used, but also may set a new precedent for how that rule may be used in the future. Most Wittgenstein scholars wouldn’t agree with this account, but most of them wouldn’t agree on much of anything.

The meaning of what we say can’t be abstracted away from the context in which we say it.

For the sake of investigation, let’s consider what Pears’ interpretation might mean. It means that instead of a word having a fixed definition or referent, a word is an evolving entity that carries its own history with it through time, picking up new nuances and discarding old ones as practices (linguistic and life) shift. This is trivially true in a sense, as you can see from dictionaries grudgingly accepting that literally now also means “not literally” and me grudgingly accepting that begging the question will usually mean “raising the question” for the rest of my natural life and I should just start saying petitio principii instead. But the implications are more troublesome when you get to nouns, especially as they get more abstract. The usage of dog has remained somewhat consistent over the years, but try defining love or heavy or Russia in any kind of complete or precise way. You can’t do it, yet we use these words with confidence every day. As Pears puts it, “The fixed rails on which we are supposed to be running when we use a descriptive word are a fantasy.”

Here’s one example. The French equivalents for here and there are ici and là respectively. But if I point to a pen and say, “The pen is here,” the French equivalent is not “Le stylo est ici,” but “Le stylo est là.” In French, là is always used to refer to a specific place or position, while in English here or there can both work. This rule is so obscure I never learned it in French classes, but obviously all native speakers learn it because no one ever uses it differently. It could just as easily be the other way round, but it’s not. The situation is not arbitrary, but the way in which language carves up the interaction between mind and world varies in such a way that French speakers recognize certain practices as right or wrong in a different way than English speakers do. This may seem a trivial point, until you have to program a computer to translate “I pointed to Paris on the map and said, ‘She is here.’ ” into French—at which point it becomes a nightmare. (If you are a translator, on the other hand, this is great news.)

Wittgenstein’s later work is an ongoing, minute examination of these kinds of practices and all the ways in which they can possibly work: in everyday life; in mathematics; and especially when it comes to referring to our thoughts, sensations, and feelings (whatever the hell those are).

Sometimes these endless pages of investigations, which Wittgenstein never did assemble into a completed work (Philosophical Investigations comes closest but was not finished), can seem like so much sophistry. Wittgenstein’s later work often comes in the form of gnomic fragments, near-aphoristic pronouncements, and bedeviling dialogues between two (or perhaps three) voices in which you frequently aren’t sure whether Wittgenstein is arguing for a position or against it. Frustration can ensue when, contemplating the difference between someone in pain and someone pretending to be in pain, Wittgenstein declares that the sensation of pain is “not a something, but not a nothing either! The conclusion was only that a nothing would serve just as well as a something about which nothing could be said.” There is a cogent thought here, but the approach is opaque.

Philosopher Jerry Fodor parodied Wittgenstein with a page of elaborations on Michael Frayn’s own parody, “Wittgenstein on Fog-like Sensations,” of which this is the best:

“I can’t see a thing in this fog.” Which thing?

So, language is quicksand—except it’s not. Unlike the parlor tricks of the deconstructionists who bloviate about différance and traces, there clearly are rules that shouldn’t be broken and clearly ways of speaking that are blatantly incorrect, even if they change over time and admit to flexible interpretations even on a daily basis. It’s just that explicitly delineating those boundaries is extremely difficult, because language is not built up through organized, hierarchical rules but from the top down through byzantine, overlapping practices. Some things can be pinned down with practical certainty, just not in isolation and without context.

Artificial intelligence was quite slow at learning this lesson. Well into the 1970s, it was still assumed that computers could understand natural language in more or less the same way that they could understand formal logic: by interpreting them as propositions that were either true or false. The efforts in this direction have, on the whole, been remarkably unsuccessful.

Wittgenstein’s philosophy also accounts for the disastrous state of Internet discourse today.

And these difficulties are exactly why Google succeeded—by ignoring semantics as much as possible, sticking instead to whatever it could glean without trying to understand the meaning of words or sentences. Google could count the popularity of a word, see which words co-occur with others, figure out which people where use which words—anything as long as it didn’t require determining where and how one should use a word. In very limited, circumscribed situations, like asking questions of certain specified forms, computers can figure out what you mean, and even then things are very limited. Google can answer, “How many ounces in a pound?” but still can’t tell me “How many years has Obama been in office?” Picking up on “Obama” and “years” and “in office,” Google returns some data about his 2012 re-election, but that’s as far as it gets in

understanding” my question. The problem, as summed up by Wittgenstein: “Understanding a sentence means understanding a language.”

Wittgenstein’s philosophy also accounts for the disastrous state of Internet discourse today. The shift to online communication, textual interactions separated from accompanying physical practices, has had a persistent and egregious warping effect on language, and one that most people don’t even understand. It has made linguistic practice more limited, more universal, and more ambiguous. More people interact with one another without even realizing they are following different rules for words’ usages. There is no time or space to clarify one’s self—especially on Twitter.

It is this phenomenon that has affected political and ethical discourse in particular. To take some hot-button issues, use of the words privilege and feminism and racism is so hopelessly contentious that it’s not even worth asking for a definition—even if you get one, no one else will agree with it. In situations where misuse can get you savaged on the Internet, I’ve simply stopped using a word. Let me know when everyone else has worked it out.

There are few easy ways into Wittgenstein’s work; I was fortunate to have David Pears as a guide, and I’ll be forever grateful to him. (He died in 2009.) Ray Monk, who penned an excellent biography of Wittgenstein (The Duty of Genius), also wrote by far the most accessible introduction, How to Read Wittgenstein. I think some of its analysis is seriously off-base, but it does give some idea of what Wittgenstein was up to. After that, I find Marie McGinn’s exceptional Routledge Guidebook to Wittgenstein’s Philosophical Investigations to be the clearest and most readable overview of the later work, though it is significantly heavier going. (It also leads off with the apt Flaubert quote, “Stupidity consists in wanting to reach conclusions.”) Equally strong work by David Stern, Stanley Cavell, William Child, and David Pears himself will be rough for those without a philosophical background. Or you could just dive into the Philosophical Investigations and see what you make of it.

There’s no doubt in the end that Wittgenstein remains a real pain. I wouldn’t have read him had I not been convinced that the answers I was getting from every other field, and even from every other philosopher, were unsatisfactory. Wittgenstein didn’t give me the answers, but his philosophy did teach me to ask the right questions. Even still, it took me years after my class to get a grasp on these questions in any kind of systematic way. Today, however, they inform every single piece I write without exception, though often invisibly. “Philosophy is a battle against the bewitchment of our intelligence by means of language,” Wittgenstein wrote. And that lesson is worth repeating on a daily if not hourly basis.