Monday, February 16, 2015

Armand Marie Leroi: Digitizing the Humanities into a Science?

Armand Marie Leroi, professor of evolutionary developmental biology at Imperial College, London, inadvertently alarms us to the threat of scientism's frontal assault on the humanities in "One Republic of Learning: Digitizing the Humanities" (New York Times, February 13, 2015). Inadvertently? Yes, because he's describing scientism without realizing it, not even when he cites the term only to dismiss it:
In the Republic of Learning[,] humanities scholars often see themselves as second-class citizens. Their plaintive cries are not without cause. When universities trim budgets it is often their departments that take the hit. In the last 10 years, however, there has been one bright spot: the "digital humanities," a vast enterprise that aims to digitize our cultural heritage, put it online for all to see, and do so with . . . scholarly punctilio . . . . The digital humanities have captured the imaginations of funders and university administrators. They are being built by a new breed of scholar able to both investigate Cicero's use of the word "lascivium" and code in Python . . . . But the true promise of digitization is . . . . the transformation of the humanities into science[,] . . . mean[ing] using numbers to test hypotheses. Numbers are the signature of science; they allow us to describe patterns and relationships with a precision that words do not. The quantification of the humanities is driven by an inexorable logic: Digitization breeds numbers; numbers demand statistics. The new breed of digital humanists is mining and visualizing data with the facility that bioinformaticians analyze genomes and cosmologists classify galaxies. All of them could, if they cared to, understand each others' results perfectly well . . . . [The triumph of numbers is] easy to see . . . . A traditional, analog, scholar will make some claim about the origin, fate or significance of some word, image, trope or theme in some Great Work. He'll support it with apt quotations, and fillet the canon for more of the same. His evidence will be the sort that natural scientists call "anecdotal" - but that won't worry him since he's not doing science[, except that] . . . . then a code-capable graduate student will download the texts - not just the canon, but a thousand more - run the algorithms, produce the graphs, estimate the p values, and show the claim to be false, if false it indeed is. There will be no rejoinder; the analog scholar won't even know how to read the results. Quantification has triumphed in field after field of the natural and social sciences. It will here, too . . . . [even though h]ard words such as . . . "scientism . . . will be flung about . . . [because] the vocabulary of anti-science is rich, well-honed.
This scientistic triumphalism sounds 'too bad to be false'! But it isn't, because it isn't true, as Leroi himself admits, sort of:
If the rudiments of a new cultural science are visible, so are its limits. There is one great difference between human and natural things: The former have meaning; the latter do not. That is why the humanities are filled with critics and the natural sciences are not: Critics tell us what artifacts mean . . . . [But] deep-learning algorithms are becoming very good at extracting meaning from data; and, as art becomes data, it is always possible that new meanings may be revealed by algorithmic microscopes yet unbuilt. That said, it would take a very clever algorithm to flag up irony in Jane Austen. More fundamentally, the truth of art criticism is not the same kind as scientific truth.
This concession is deeply at odds with Leroi's earlier triumphalism and undermines the whole scientistic enterprise! Quantification and the mining and visualizing of data will have their uses, but the 'influence' of some text upon other texts cannot - at the most fundamental level - be measured, for the mere occurrence of some word or other says nothing about whether the 'influence' favors the later writer's acceptance or rejection of the earlier writer's view. Influence is thus intrinsically ambiguous! And that is precisely where the question of meaning gets raised.

Or so I interpret. What do others reading this post think?

Labels: , ,


At 11:19 PM, Blogger Kevin Kim said...

I'm still just trying to wrap my head around what all this might even mean.

"deep-learning algorithms are becoming very good at extracting meaning from data"

Really? Are they? I think a certain Arizona-based philosopher would beg to differ, and as much as I'm a Kurzweilian when it comes to strong AI and functionalist views of intelligence, I'd have to grudgingly agree with said philosopher... for now, at least.

"...the truth of art criticism is not the same kind as scientific truth."

I'm also not sure what this means. In what sense is criticism—assuming "criticism" is a species of opinion—true? I've heard of a "legitimate criticism" before, i.e., a criticism that is rational because warranted and justified, but I've never heard of "true criticism."

My own feeling is that we're not even close to making machines and/or computer algorithms that can appreciate meaning—at least not on anything like the deep, emotional, anthropic level of meaning-appreciation. A self-driving car might "understand" what red on a traffic light "means," but that's a very brute, superficial take on the concept of meaning—it's more on the level of pure action-reaction, stimulus/response. Is meaning quantifiable? Good question. For now, though, I'd say the answer is a definite "no."

At 6:05 AM, Blogger Horace Jeffery Hodges said...

Leroi doesn't exactly seem to be thinking about AI, but even androids who dream of electric sheep would themselves confront questions of meaning that couldn't be transformed into science through quantification. The answer to the meaning of life, the universe, and everything isn't really 42!

Jeffery Hodges

* * *


Post a Comment

<< Home