Trip report - Computing & The Humanities

Yet another meeting about information technology and the humanities, with a very knowledgable set of attendees and some very attractive presentations. Most interesting: John Unsworth's questioning of how much progress we're really making, and Richard Baraniuk's proposal for a new way to teach.

NINCH (the National Initiative for Networked Cultural Heritage) sponsored a meeting at the National Academy of Sciences (now 500 Fifth St NW) on the opportunities for the humanities researcher using information technology. Participants presented some exciting projects but were faced with social and financial barriers to doing their work and wished they could encourage more cooperation both among humanists and with the technical people. Both the University of Virginia and Rice University showed impressive breadth and depth in humanities computing; if there had been more time, I know that other universities including (but not limited to) as Tufts, Columbia, Texas A&M, Kentucky, Stanford, and UCLA could also have presented a range of attractive projects. My list of the most interesting combinations of IT and the humanities would include the work of Levoy (Stanford), Crane (Stanford), Choudhury (JHU), Kiernarn/Griffioen (Kentucky), and Allen/Murray (Columbia). (I don't mean to leave anyone out other than for space reasons).

John Unsworth gave an excellent talk reminding us that there had been at least ten years, and perhaps more like forty years, of meetings discussing the opportunities for computer tools in text analysis; I can personally remember making a concordance of Icelandic sagas for a friend in the late 1960s. The tools available are still hard to learn about and don't interwork well. John observed that interoperability and modularity are often traded off for a better user interface; we haven't learned enough about getting generality and ease of use together. Progress has taken the form of (a) much greater availability of machine-readable resources, (b) image and sound resources as well as text, and (c) the use of the Web for wide distribution of humanities information on line. We still need to convince both humanities researchers of the importance of computing and the general public of the importance of humanities research.

I have trouble understanding what sorts of questions humanities researchers pose that computing can "answer," partly because many important humanities questions don't have simple answers. There will never be a one-word answer to "What was the cause of the American Revolution?" It does surprise me though that perhaps the mostly easily recognized result from the application of information technology to the humanities is more than forty years old: the Mosteller article on the authorship of the Federalist Papers.

Perhaps more important for the public than the impact of computing on humanities research would be the impact on humanities education. Teaching has been remarkably unaffected by technology over the years, with many disappointments from earlier hype. As my friend Christine Borgman of UCLA says, she knows any number of faculty members who would not be caught dead with a computer on their desk more than two years old but who still teach by walking into a lecture room and picking up a piece of chalk.

The most interesting talk about teaching was about the Connexions project at Rice University, given by Richard Baraniuk. This project encourages faculty to create "modules" rather than courses. Modules contain something around a lecture or a week of teaching (e.g. "spatial Fourier transforms" or "digital filters"). Courses can be flexible both in the arrangement of the modules and in the ability of a professor to change any individual module to meet particular circumstances. Unlike George Tech e-classes, the individual class structure and sequence is not retained and there is no video, giving greater flexibility for re-use. The best analogy is the MIT online courseware, but Rice is again stressing the ability to change the modules and rearrange them. Some 1000 modules exist covering 9 courses (primarily in electrical engineering). I was surprised that it is possible to write these small modules not knowing what the students might have studied before, but Geneva Henry said that the students were very good about following links to modules explaining pre-requisite material. Enthusiasm from the students is high and Rice is hoping to expand the e-courses to more of their engineering curriculum. It's too early to expect an evaluation of student performance.

Now some briefer notes on other talks.

Bill Wulf began by suggesting simulation of society and human behavior to transform humanities research. I think this is premature; once at NSF I asked if we should try for a total simulation of the worm "C. elegans" and couldn't get agreement on whether we were ready for that. Even a guinea pig is way out of reach; after we can do that, we can work our way through the animal kingdom to a Second Trombone.

Janet Murray explored styles of cooperation between humanities and computer science, rejecting most of them as inadequately close or fair, and discussing the abilities of computers to create interactive media in which choices made lead to different outcomes. I couldn't really see this as an advance over plays like Alan Ayckbourn's "Intimate Exchanges" or the movie "Sliding Doors", unless there are authorship tools making this widely accessible. Personally I'm always hesitant about complex multimedia presentation systems since they tend to require too many skills and/or too much capital investment for an individual to be able to use them, and thus hampering the ability of one creative individual to produce something. There are very few people who can excel at both the narrative and visual arts (Blake, Rossetti, and William Morris come to mind, but they are unusual).

Greg Crane reviewed some historical frustrations of humanists seeking to build computer systems, and it didn't seem that so much had changed about the way computer nerds and humanists interact, or about the low quality of computer software and interfaces; personally I think it's gotten worse. Greg had a couple of dramatic examples. In his first class on Homer the professor spent six weeks on the first line of the Iliad. That's not Internet time. Computer nerds do fast food, not slow cooking. Then Greg discussed the Cambridge (Massachusetts) Civil War memorial on the Common, and how much you can find out about the people named, and the history of the memorial itself. It reminds me of my favorite example of the impact of the Second World War on the US, as compared with any war since the Civil War: a plaque on E. 7th St in Manhattan, listing the names of the men of that block only who served in the armed forces during the war. There are 180 names on this plaque, from one block of New York. I think that was roughly one per apartment in 1940.

More generally, Greg talked about the possibilities if the newest language technology was applied to humanities text. I'm not sure searching would get all that much better; the history of phrase matching and the like is very discouraging. But I agree that on-the-fly translation would be of great assistance, and perhaps some of the "named-entity" handling. I think we need a lot of research on knowledge representation before that's going to help much; humanists could probably do this as well as nerds.

Personally the part of this talk I liked best was the conversion of drawings of 19th century London streets into a VR model. There are several such projects trying to recreate an city, ranging from modern Los Angeles to medieval Kyoto and classical Rome. Greg's version does this with a mixture of modern maps and older drawings and demonstrates a kind of data fusion we're going to have to get better at, given the widely disparate kinds of geographic information that we are likely to have for different sites.

Doug Greenberg then presented the work of the Visual History Foundation, which is has more than 52,000 videorecorded interviews with Holocaust survivors. They are experimenting with a mixture of access methods, including a minimal amount of structured data for each interview, a much longer detailed index by time segment, and transcripts produced by speech recognition. It's a huge job to do the indexing (and a very stressful one, given the nature of the content). Usage is constrained by privacy issues as well as by practical matters. As of now there isn't really enough usage either by educators or researchers to make clear what forms of access will be most important. Exploiting a major oral resource like this is as yet badly understood, and yet of immense importance since more and more of our "history" is going to be recorded as spoken and/or video material rather than just text.

The most visually attractive talk was by Steven Murray describing his measuring and modeling of French cathedrals, so that 3-d representations can replace ordinary slides in teaching the history of art and architecture. His examples of Amiens cathedral and some churches in Burgundy were very convincing for the advantages of this technology for presentation, even if I don't quite accept all the numerology of the dimensions of the churches (thinking that surely the site topography must have constrained the details of the design as much as some belief that a square drawn on the floor should match an elevation).

Will Thomas described the creation of a complex web structure about the role of slavery before the Civil War in two Shenandoah Valley counties. There is lots of provocative and interesting data, for example maps showing the ubiquity of slaveholders in Augusta County, Virginia. Most of the discussion was not however about the value of this material for research or teaching, but about the problem of rewarding the 18 months of work that went into the site, which turned into a discussion of whether this should be called an "article" or a "book". You can't get tenure in a good history department, for example, if you write articles at the rate of one every 18 months. I don't know whether I'm more depressed at the length of time it is taking for universities to accept online publications that don't conform to conventional models, or at the survival, even when considering traditional publishing, of an evaluation model that considers the number of articles more important than their content (as in the old line that "Deans can't read, they can only count").

Perhaps a useful higher level discussion was the role of narrative in electronic publication. Will the ability of computers to provide interactive and small-unit presentations destroy the traditional task of presenting a long and coherent linear explanation of historical phenomena? It reminds me of a remark I saw recently that to succeed in today's corporate management you have to be suffering from attention deficit disorder.

Cliff Lynch talked about the problems of moving these techniques into the mainstream. We need better tools, and re-usable data; and we also need social changes, to include the necessary technology in courses and training for graduate students. A particularly touchy social issue is the credit system for data exchange. The sciences have a variety of models for sharing of data, and the humanities have to create theirs. It has always puzzled me that one of the fields with the most generous data sharing is molecular biology, with its protein and genomic data banks, despite the enormous potential financial value of this data; while astronomical data, of no practical use on earth whatever, is kept for longer by the researcher who created it. A friend remarked that it was precisely because the biochemists could get money that they could be generous with sharing prestige and credit; this bodes ill for the humanities.

Overall the conference reminded me of Jim Gray's line "may all your problems be technical". We are doing pretty well at introducing images, 3-D graphics, and sound recordings into the research corpus; we're starting on simulations and interactive media. What we haven't figured out how to do is make this mainstream and rewarded by the conventional academic structures. Humanities departments are not closed groups: they have in recent decades accepted new fields of study such as the literature of the European colonies in Africa, or feminism, or film. Technology may be difficult to study at times, but the Perl manual is a children's book compared with deconstructionist literary theory. So why don't humanities departments take up IT? I can only think the answer is in university politics: fear of domination by the technical departments, who already wield lots of power in university circles. And thus the solution may be shown by universities like Virginia, where the top administration has made clear through funding and the creation of an institute that they will reward humanities scholars in their own world and not make them a sub-department of computing.

Our contribution needs to be the exploration of interesting problems via IT, to show both the scholars and the public what can be achieved. We need another flagship project that is easily described and dramatically successful, like the authorship studies of the Federalist papers. Perhaps keeping Beauvais Cathedral from falling down will be such a project. We also should look at how to change education. The students are leading the way: they do all their research on the Web today. Faculty and researchers should catch up.

About the Computer Science and Humanities Initiative

About Key Partners of the Computer Science and Humanities Initiative