Of texts and textiles

Yesterday, if you spend as much time as I do browsing random news articles online, your eye might have been caught by a story with the headline “‘Allah’ is Found on Viking Funeral Clothes.” Similar pieces ran in multiple publications, but I’ll stick with the one in the New York Times, which I think is where I saw it first. Here’s how it begins:

The discovery of Arabic characters that spell “Allah” and “Ali” on Viking funeral costumes in boat graves in Sweden has raised questions about the influence of Islam in Scandinavia. The grave where the costumes were found belonged to a woman dressed in silk burial clothes and was excavated from a field in Gamla Uppsala, north of Stockholm, in the 1970s, but its contents were not cataloged until a few years ago, Annika Larsson, a textile archaeologist at Uppsala University, said on Friday.

Larsson says that she was examining the patterns when she “remembered seeing them in similar Moorish designs in silk ribbons from Spain. I understood it had to be a kind of Arabic character, not Nordic.” The article continues: “Upon closer examination of the band from all angles, she said, she realized she was looking at Kufic script. The words Allah and Ali appeared in the silk found in Boat Grave 36 and in many other graves—and, most intriguing, the word Allah could be seen when reflected in a mirror.” It’s “most intriguing” indeed, particularly because it’s consistent with the hypothesis, which is widely credited, that “the Viking settlements in the Malar Valley of Sweden were, in fact, a western outpost of the Silk Road that stretched through Russia to silk-producing centers east of the Caspian Sea.”

Unfortunately, this particular piece of evidence began to fall apart almost at once. I’d like to say that I felt a flicker of doubt even as I read the article, particularly the part about the pattern being “reflected in a mirror,” but I can’t be entirely sure—like a lot of other readers, I glanced over it briefly and moved on. A few hours later, I saw another story headlined “That Viking Textile Probably Didn’t Actually Have ‘Allah’ On It.” It linked to a very persuasive blog post by Carolyn Priest-Dorman, a textile historian and Viking reenactor who seems perfectly positioned to identify the flaws in Larsson’s argument. As the Times article neglects to mention, Larsson’s reconstruction doesn’t just depend on reflecting the design, but in extending it conjecturally on either side, on the assumption that portions of the original are missing. Priest-Dorman points out that this is unwarranted on the evidence:

This unexplained extrapolation practically doubles the width of the band, and here’s why that’s a problem…If you consult…a photo of Band 6, you can clearly see the continuous metallic weft of the band turning at each selvedge to enter back in the other direction.If Larsson were correct that Band 6 was originally significantly wider, you would not see those turning loops; you’d see a series of discontinuous single passes of brocading weft with cut or broken ends at each edge.

In other words, if the pattern were incomplete, we’d see the breaks, but we don’t. And even if this point were up for debate, you clearly increase the risk of subjective readings when you duplicate, reflect, and otherwise distort the raw “text.”

No one has accused Larsson of intentional fraud, but it appears that the right combination of elements—a source of ambiguous patterns, some erudition, and a certain amount of wishful thinking—resulted in a “solution” to a problem that wasn’t there. If this sounds familiar, it might be because I’ve discussed similar cases on this blog before. One is The Great Cryptogram by Ignatius L. Donnelly, who argued that Francis Bacon was the true author of the works of Shakespeare and left clues to his identity in a code in the plays. An even better parallel is the scholar William Romaine Newbold, who died believing that he had cracked the mysterious Voynich Manuscript. As David Kahn recounts in his masterpiece The Codebreakers, Newbold fell victim to much the same kind of error that Larsson did, except at far greater length and complexity:

Newbold saw microscopic shorthand symbols in the macroscopic characters of the manuscript text and began his decipherment by transliterating them into Roman letters. A secondary text of seventeen different letters resulted. He doubled all but the first and last letters of each section…The resultant quaternary text was then “translated”: Newbold replaced the pairs of letters with a single letter, presumably according to a key, which, however, he never made clear…Finally, Newbold anagrammed the letters of this senary text to produce the alleged plaintext in Latin.

The result, of course, was highly suspect. Anagramming chunks of over a hundred characters at a time, as Newbold did, could result in almost any text you wanted, and the “microscopic shorthand symbols” were nothing but “the breaking up of the thick ink on the rough surface of the vellum into shreds and filaments that Newbold had imagined were individual signs.”

Donnelly and Newbold were working before an era of instantaneous news coverage, but I don’t doubt that they would have received plenty of sympathetic, or at least credulous, attention if they had published their results today—and, in fact, hardly a month goes by without reports of a new “breakthrough” in the Voynich Manuscript. (I’m reminded of the Beale cipher, a similar enigma encoding an alleged hidden treasure that inspired an entire society, the Beale Cypher Association, devoted to solving it. In his book Biggest Secrets, the author William Poundstone examined a copy of the society’s quarterly newsletter, which is available online. It contained no fewer than three proposed solutions.) In the aftermath of the Larsson debacle, a number of observers, including Stephennie Mulder of the University of Texas, raised concerns about how the theory was reported: “It should go without saying that a single scholar’s un-peer-reviewed claim does not truth make.” She’s right. But I think there’s a more specific lesson here. Both Larsson and Newbold started with a vast source of raw material, selected a tiny piece of it, and subjected it to a series of analogous permutations. Larsson doubled the pattern and reflected it in a mirror; Newbold doubled the illusory characters and then anagrammed the result. The first step increased the amount of text that could be “studied,” while the second rearranged it arbitrarily to facilitate additional readings. Each transformation moved further away from the original, which should have been a red flag for any skeptical reader. But when you summarize the process by providing only the first and the last steps, while omitting the intermediate stages, the conclusion looks a lot more impressive. This is exactly what happened with Larsson, and when we turn to Newbold, who announced his findings in 1921, we see how little anything has changed. As Kahn writes in The Codebreakers: “The public at large was fascinated. Sunday supplements had a field day.”

The Chinese Room

In 1980, the philosopher John Searle presented a thought experiment that has become known as the Chinese Room. I first encountered it in William Poundstone’s book Labyrinths of Reason, which describes it as follows:

Imagine that you are confined to a locked room. The room is virtually bare. There is a thick book in the room with the unpromising title What to Do If They Shove Chinese Writing Under the Door. One day a sheet of paper bearing Chinese script is shoved underneath the locked door. To you, who know nothing of Chinese, it contains meaningless symbols, nothing more…You are supposed to scan the text for certain Chinese characters and keep track of their occurrences according to complicated rules outlined in the book…The next day, you receive another sheet of paper with more Chinese writing on it…The book has further instructions for correlating and manipulating the Chinese symbols on the second sheet, and combining this information with your work from the first sheet. The book ends with instructions to copy certain Chinese symbols…onto a fresh sheet of paper. Which symbols you copy depends, in a very complicated way, on your previous work. Then the book says to shove the new sheet under the door of your locked room. This you do.

Unknown to you, the first sheet of Chinese characters was a Chinese short story, and the second sheet was questions about the story, such as might be asked in a reading test…You have been manipulating the characters via a very complicated algorithm written in English…The algorithm is so good that the “answers” you gave are indistinguishable from those that a native speaker of Chinese would give, having read the same story and been asked the same questions.

Searle concludes that this scenario is essentially identical to that of a computer program operating on a set of symbols, and that it refutes the position of strong artificial intelligence, which he characterizes as the belief that “the appropriately programmed computer with the right inputs and outputs would thereby have a mind in exactly the same sense human beings have minds.” According to Searle, it’s clear that there isn’t any “mind” or “understanding” involved here:

As regards the first claim, it seems to me quite obvious in the example that I do not understand a word of the Chinese stories. I have inputs and outputs that are indistinguishable from those of the native Chinese speaker, and I can have any formal program you like, but I still understand nothing.

I’ve never been convinced by this argument, in part because I approached it through the work of Douglas R. Hofstadter, who calls it “a quintessential ‘bad meme’—a fallacious but contagious virus of an idea, similar to an annoying childhood disease such as measles or chicken pox.” (If it’s a bad meme, it’s one of the all-time greats: the computer scientist Pat Hayes once jokingly defined cognitive science as “the ongoing research program of showing Searle’s Chinese Room Argument to be false.”) The most compelling counterargument, at least to me, is that Searle is deliberately glossing over how this room really would look. As Hofstadter notes, any program capable of performing in the manner described would consist of billions or trillions of lines of code, which would require a library the size of an aircraft carrier. Similarly, even the simplest response would require millions of individual decisions, and the laborious approach that Searle presents here would take years for a single exchange. If you try to envision a version of the Chinese Room that could provide answers in real time, you end up with something considerably more impressive, of which the human being in the room—with whom we intuitively identify—is just a single component. In this case, the real “understanding” resides in the fantastically complicated and intricate system as a whole, a stance of which Searle dismissively writes in his original paper: “It is not easy for me to imagine how someone who was not in the grip of an ideology would find the idea at all plausible.”

In other news, a lawsuit was filed last week against John Searle and the Regents of the University of California, where he has taught for decades, accusing him of sexual harassment. The plaintiff is a twenty-four-year-old woman, Joanna Ong, who was employed as Searle’s research assistant for three months. The complaint states:

On or about July 22, 2016, after only a week of working together, Searle sexually assaulted Ong. On that date, he asked his previous research assistant to leave his office. He then locked the door behind the assistant and then went directly to Ong to grope her. Professor Searle slid his hands down the back of her spine to her buttocks and told Ong that “they were going to be lovers,” that he had an “emotional commitment to making her a public intellectual,” and that he was “going to love her for a long time.”

When Ong took her story to the director of the John Searle Center for Social Ontology, she was allegedly told that Searle “has had sexual relationships with his students and others in the past in exchange for academic, monetary, or other benefits.” No further attempt was made to investigate or respond to her claim, and the incidents continued. According to Ong, Searle asked her to log onto a “sugar daddy” website on his behalf and watched online pornography in her presence. The complaint adds: “On one occasion, when Ong”—who is Asian-American—“brought up the topic of American Imperialism as a discussion topic, Searle responded: ‘American Imperialism? Oh boy, that sounds great honey! Let’s go to bed and do that right now.’” When Ong complained again, the lawsuit states, she was informed that none of these issues would be addressed, and she ultimately lost her job. Earlier this month, Searle ceased to teach his undergraduate course on “Philosophy of Mind,” with university officials alluding to undisclosed “personal reasons.” As far as I know, neither Searle’s attorney nor anyone at the university has commented on the allegations.

Now let’s get back to the Chinese Room. At its heart, the argument comes down to a contest between dueling intuitions. Proponents of strong artificial intelligence have the intuition, or the “ideology,” that consciousness can emerge from a substrate other than the biological material of the brain, and Searle doesn’t. To support his position, he offers up a thought experiment, which Daniel C. Dennett once called “an intuition pump,” that is skewed to encourage the reader to arrive at a misleading conclusion. As Hofstadter puts it: “Either Searle…[has] a profound disrespect for the depth of the human mind, or—far more likely—he knows it perfectly well but is being coy about it.” It reduces an incomprehensibly complicated system to a user’s manual and a pencil, and it encourages us to identify with a human figure who is really just a cog in a much vaster machine. Even the use of Chinese itself, which Searle says he isn’t sure he could distinguish from “meaningless squiggles,” is a rhetorical trick: it would come off as subtly different to many readers if it involved, say, Hungarian. (In a response to one of his critics, Searle conceives of a system of water pipes in which “each water connection corresponds to a synapse in the Chinese brain,” while a related scenario asks what would happen if every Chinese citizen were asked to play the role of a single neuron. I understand that these thought experiments are taking their cues from Searle’s original paper, but maybe we should just leave the Chinese alone.) And while I don’t know if Searle’s actions amounted to sexual harassment, Ong’s sense of humiliation seems real enough, which implies that he was guilty, if nothing else, of a failure of empathy—which is really just a word for our intuition about the inner life of another person. In many cases, sexual harassment can be generously viewed as a misreading of what another person needs, wants, or feels, and it’s often a willful one: the harasser skews the evidence to justify a pattern of behavior that he has already decided to follow. If the complaint can be believed, Searle evidently has trouble empathizing with or understanding minds that are different from his own. Maybe he even convinced himself that he was in the right. But it wouldn’t have been the first time.