Lies, Damned Lies

The elevators in our office building have these little monitors built into them, on which are displayed random tidbits of pseudo-news and other glossy distractions. Because god forbid we should be bored on the ride to the third floor.

Anyhow, the other day as I was leaving the office, the monitor was showing a little infographic that showed a steep decline in the number of hours per week Americans are working — from 40-point-something in 2009 to 34-point-something in 2011. The other person on the elevator and I looked at one another and both said “that can’t be right,” but there was no context other than “the number of hours per week Americans are working” and a series of numbers associated with years. We wondered at the time whether “Americans” meant all adult Americans or adult Americans with jobs, such that the steep decline indicated more people out of work. As I wandered off toward the subway, it hit me that even if the figures referred to adult Americans with jobs, the steep decline could indicate the growing part-time-ification of the workforce — in which case the drop suggests a growing under-employment problem, and not that Americans are opting to spend more time in the Hamptons.

This stupid infographic has annoyed me since I saw it, not least because it demonstrates everything that’s wrong with the ways that statistics get used by the mainstream media: Look! A number! It must Mean Something! What gets missed, of course, is that the gap between numbers and meaning can only be bridged by interpretation, and that such interpretation requires serious critical and analytical skills. That while numbers may have a demonstrable basis in empirical reality, what they mean is not at all evident, and many interpretations of them may simply be wrong.

All of this has me thinking about some of the claims that have been leveled at the digital humanities in recent days, in particular that it’s a mode of “processing” texts that attempts to bring literary studies fully into the empirical, by counting rather than reading. And sure, there is some work in the field whose results look awfully numeric. But by and large, while the wide range of work covered by the umbrella term “digital humanities” has one foot in the digital — in the kinds of tools and methods many have associated with scientific or engineering or other empirically-oriented fields — the best of it has the other squarely in the humanities, and in the kinds of questions and concerns and modes of analysis and interpretation that arise out of those fields. And it seems to me today that one key role for a “worldly” digital humanities may well be helping to break contemporary US culture of its unthinking association of numbers with verifiable reality, by demonstrating the ways that such numbers only open the process of meaning-making.

11 responses to “Lies, Damned Lies”

  1. A short post from @kfitz on interpretation, quantitative analysis and DH:

  2. #DH #MLA12 //MT @rgfeal I certnly agree w @kfitz that numbers don’t tell stories. Narrative does. Context does. Analysis does.

  3. A role for DH may be to break “US culture of its unthinking association of numbers with verifiable reality” – @kfitz

  4. I’ve been thinking a lot about this lately (the first part–less the “digital humanities” bit), and it is putatively a big piece of the fabled “next book.”

    The problem with this is that you can push for more analysis but it isn’t going to happen, at least at the “end user” level. I think you can and should expect journalists to be more critical, but the structural problems in journalism argue against this. So what is needed is more quantification and distributed filtering. The latter of these was traditionally performed by good journalists, but again, that model is broken.

    I suspect quite a few New Yorkers get a good chunk of their news while riding elevators. You aren’t going to fit a lot of nuance on that screen. The trick is knowing that numbers are attractive (shiny! numbers!) and designing metrics that are biased in the right kinds of ways. All metrics are biased in some ways, even when the bias is unintentional. So, they should be (a) strategically biased and (b) how they are constructed (how the bias is built in) should be easily discoverable.

  5. Kimon Keramidas (@BGCDML) Avatar
    Kimon Keramidas (@BGCDML)

    I agree whole-heartedly with the need for interpretation, but I would argue that one problem with digital humanities right now is that there also a decided charge away from the theory and interpretation you are saying is so central the umbrella term. That charge away I think is what alienates a lot of yet-to-be-digital scholars, partially because to them it reeks of just the kind of focus on number gathering and fetishization of algorithms that you are calling out against. That is why I have always been wary of the “tools for the sake of tools” argument that was actually quite prominent in one of those big NYTimes articles on digital humanities ( It actually provides justification for resistance against the uptake of the use of the very tools the tool-builders are making.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.