Thread by @alanyliu

Just made a 1st pass through Nan Z. Da's worthy "The Computational Case against Computational Literary Studies," just out in Critical Inquiry. A blast against the cultural/text analytics side of DH based on statistical & related kinds of critique. bit.ly/2XX8Ku0

It may be a bit too single-minded for its own good, like a missile-camera view of its target (we used to say "tunnel-vision" before our POV went aerial in the age of drones). I think it may be holding distant reading up for (very hostile) comparison against the wrong ideal.

The piece critiques distant reading by comparison with an ideal of perfect statistical validity and methods of inference. And it also forecloses at its start any "incrementalist" argument that DH is trying to improve in that regard.

But, honestly, the valid domain of comparison is that named in the article's title: "Literary Studies." The meaningful comparison in the current juncture of literary studies is not perfect statistical validity but the kind of thing that I used to say as a literary scholar:

e.g. (generic example): "Wordsworth uses 'joy' a lot in important poems like 'Tintern Abbey'." Evidence of that sort underlies much of literary studies, going back to close reading. Let's compare the statistical validity of _that_ to DH's attempt to make it, if not right, better.

Another way to make the point is to think about Da's critique of the adequacy of distant reading's inferences of "aboutness, influence, relatedness, connectedness, generic coherence, & change over time" (p 605). Really, I'd like to see such criteria assessed for regular lit crit.

By the way, it was about here near start of Da's piece that I began to suspect it might be a screed. "It is in fact true that data mining text labs are given institutional resources disproportionate to what they offer and how little computing power…their work...requires." (603)

It would be futile to ask for the statistical validity of this statement based on the single anecdotal example offered in a note and the undefined variable of "what they offer." Also, I find it amusing that I am identified in the piece as "literary sociologist Alan Liu."

Had another day to think about this. Da's piece is worthy, as I started by saying. My fundamental problem with it has less to do with its truth value (which will be adjudicated over time) than its fundamental lack of generosity and so its dishonesty to larger truth.

Here is a generous critique: 'X does this well in trying to do Y, & we can see why this is important in the context of larger issues in literary studies and society such as Z. But some of its ABC's are wrong. Or the way it frames its larger context is off-angle. We can improve.'

Here's a screed: 'I'll list in local context only every imputed error by everyone in this community with whom I have no relation but hover over like a drone; & I'll reference nothing else. And I'll do that unrelievedly. Plus I'll close off all exits, such as improveability.'

'Plus I'll overreach in my claims, as in my single data point about overinvestment in DH.' (By comparison with what? the gazillions invested over the years in the libraries and departments needed for normative literary studies?)

'Plus I'll not acknowledge the generosity of my targets in making their data either open or available so that I can test for reproducibility, by contrast with the inexplicit methods and basic unreproducibility of normative literary reading, especially close reading.'

'Plus I'll not paint my own vision of what statistically improved machine learning for lit. crit. might be, including why it may be needed when lit. tangles with other media, culture, & networks at the scale & quality level that I (Da) acknowledge such methods were made for.'

P.S. Not subtweeting. As mentioned earlier, I just can't find and verify a Twitter handle for Nan Z. Da.