Tito Orlandi (Univ. di Roma La Sapienza e Acc. dei Lincei).

Conference
"The future of the humanities in the digital age",
Bergen Sept. 25 - 28, 1998.
 


Minutes from workshop 1:
"Formal Methods in the Humanities and their Teaching",
Bergen, 28 September 1998.

[Written by Tito Orlandi on the basis of notes taken by Lou Barnard (God bless him!)]

Present: Tito Orlandi (chair), Ingo Kropac, Allen Renear, Elisabeth Burr, Lou Burnard, Manfred Thaller, Jacques Souillot, Harold Short, Dag Tjemsland, Espen Ore, Joseph Bell, Jon Gunnar Jørgensen.

The workshop was organised by the ACO*HUM working group in "Formal Methods for the Humanities", and intended:

  1. to define and explore some of the problems raised by the use of formal methods in the humanities;
  2. to discuss the academic perspectives of future humanities;
  3. to discuss the role of the working group (see below).

The items for discussion were:

Orlandi introduced the discussion: The questions to be debated are of theoretical type, but we may start from real life situation. When one scholar has to do with a problem the solution of which may be helped by means of the computer, the first thing he must do is to describe the problem itself in a suitable way, so that the right procedures (software) may be chosen or created, by himself or by others.

There is a difference here between humanities and (hard) sciences: in these last it is obvious to formalize a problem using the mathematical language; humanities did not use any formal language before the advent of computers. One solution might be to use mathematical language also in the humanities. But this is contrary to the personal attitudes of those who have chosen humanities as their scholarly field, and to the sound principle to respect traditions, even in new circumstances. The trend towards statistical linguistic or quantitative history or statistical archaeology does not seem an acceptable answer to the problem, because it would confine the use of computers to restricted branches of those disciplines.

Besides, what the scholar needs is not just to find tools, but to describe the problem so that an appropriate automatic procedure can be found. We want to retain tradition, and on the other hand to learn to formalise in a good way, by means of metalanguages suitable to the humanities.

We may remember that all recent work of Jean Claude Gardin has been devoted to the possibilities of transformation of humanists problem-defined language into a formal language. His achivments, though far from conclusive, show the way: we must (1) understand what formalization in the Humanities really means; (2) enquire which parts and problems of the Humanities can be formalized; and (3) enquire how formalized problems can be treated through automation.

Orlandi concluded suggesting that discussion lead to arrange a plan for the future work of the the workgroup.

Burnard: encoding and hermeneutics are what distinguish the humanistic approach. It is «non problem oriented», so it is difficult to formalize its methodologies.

Short: this workgroup was set up by the Acohum Board as service to oher committees: other workgroups' concerns with methodology should be centered here.

There has been little work on formal methods, but there is some; so the workgroup can get something achieved in next 12 months in order to help the whole project to develop. Emphasis should be put on teaching and learning. New theory might contribute, but may be left for later development.

Practical problems: application of tools in ignorance of the underlying formalism. Inappropruate use of computer techniques. Lack of formal documentation, so lack of continuity. Formal descriptions needed at analysis, at encoding/implementation stages. On which formal methods are the descriptions based?

Textual scholars 10 years ago: variety of tools, methods; no formal method underlying their activities. Now we have «tei».

It's becoming impossible to be a linguist without being a computational linguist; Kenny foresaw a day when every humanities scholars would have to have formal statistics as part of their job.

Souillot presented some results of the Erasmus CAMEEL project: it surveyed degree of knowledge expertise in universities doing non-euro (africanist and japanistics) languages and literatures, with predictably depressing results. Working towards european masters in multilingual engineering.

Thaller: Gardin participated in the first Humanities Computing conference ever held in '62, and has been in field since then. He was archaelogist sorting artefacts. Key sentence in later work: production of results is not what Humanities Computing is about, but rather to find models of thinking. Long careers usually lead to stronger interest in methodology.

At two periods HC was reshuffled completely: in mid '80s -- formal methods in history developed; then came pc. usage expanded inversely with methodological focus. Last 5 yrs, same process. Methodology is what remains if technology develops further. MT's 87 paper on potential for fuzzy sets could be reprinted in 96 because method still not applied. What we should not do:
- do the same as me because i am well funded
- assume that other disciplines will do work for us (computer scientists do not have something to learn).

Start from intellectual issues; do not assume computer scientists will work procedural questions. Need forum for discussn of methodology independent of applications. The application of computers in the Humanities has frequently been seen as a field, where ready made tools should be applied for short term profit. This approach has obvious implications for the teaching of the Humanities: when the purpose of the application of IT is the reaching of short term goals, to teach such applications has necessarily to be tied to practical courses.

If one reviews the various attempts to teach "Humanities and Computing" courses during the last twenty years, one easily sees, that a majority of them disappeared within a few years, because the subjects taught were after that time to be considered part of a general computer literacy, which students either acquired outside of university teaching or considered to be too trivial, to be bothered with.

It is highly significant in that context, that the only field in the traditional Humanities which has been succesful in spawning its own IT related sub-discipline, computer linguistics, has gone a completely different way. While the situation of a few years ago, when computer linguists considered practical computing as a matter too trivial to bother a highly trained scientist, has disappeared, the discipline is still using a very high level of abstraction in its approach to computing.

The paper proposes, that in other areas of the Humanities the definition of similarly abstract sub-disciplines can be undertaken successfully. It argues further, that only such an approach can define a topic which can be usefully taught to students over the short term changes in IT fashions. It explores, what the curriculum of such courses would have to be and ends by comparing a number of current curricula in Europe, which provide "Computer and Humanities" degrees on a high level of generality. Putting things on web isnt going to work. web pages dont form part of a cv.

List:
- dont believe in common humanities. individual disciplines may have something in common
- pedagogic aspects: problem is that HC depts are too much results oriented. training on purely practical basis (e.g. hc in US) becomes self-defeating
- loss/disparition: infrastructure need : data archives, not a question of methodology
- technology specialists: dont just hire them, either aim at joint degrees, or train them in humanities.
- teaching correct use of comms: this is not methodology
- evaluation: hc has an underdog feeling; humanities take us more seriously if we talk their lang
- organize repositories: use classic orgs, dont create niche orgs like data archives of 70s

Buzzetti: There are different questions proposed, and different points of view.

Acohum context: emphasis on teaching and learning is right, but for us, it's necessary that this be a discipline before we can teach it. We need to wear two hats. We should issue recommendns for political reasons.

Humanities disciplinary methods are more complex than those in physical sciences. In any cas, using digital data is already formal. Computer Scientists dont have as sophisticated way of thinking as humanities: e.g. poet expresses emotion; scholar analyses emotion.

Discussion

Tjemsland(?): what do we mean by formalization? Language, written or digital: info it contains is always the same. In digital form, formal methods are more easily used. How do we get info from digitized text. TEI formalizes something in language, but at one level, too roughly to be useful in computer analysis. Every morpheme and wordstem, minute level up to syntax, semantics. CL started by formalizing from smallest part, having control. Once this is done, can go on doing what you did before but with larger results.

Joe (Bell?): total formalization of text is about severe restriction of texts we can have. much lower level of standrdization is necessary. We are interested in searching for ideas and concepts across huge amounts of text, not microanalysis. "Reasonably correct and enormous" is as good as "Watertight and controllable".

Kropac: we should talk about abstraction. whatever text you have. Info is a result of processing not inherent in the text. Formalising a text is adding knowledge to a text, e.g. of a specialist, so that it can be read.

Burnard: ...

Tjemsland(?): we can do analysis of Swahili automatically making explicit what is implicit

Orlandi: The discussion must stop now, unfortunately. It's getting very interesting. but if we go aside, none will understand us. Someone must be able to teach these issues to students. This is now the main issue.

Burnard: Relation to traditional skills.

Buzzetti: Critical digital editions as object of teaching, not as an aid. Documentation is essential.

Kropak: Validated resources are needed; what pieces of methods are common to the humanities.

Orlandi: Can we leave this kind of problems to the computing scientists? There is something common in humanities. Do we change methods when we use computing, and if so, in the same way in different branches?

Burnard: Do humanities disciplines actually use computers in same way?

Bell: We need lots of minimally tagged texts and search tools, and training for students. also research methods. Some kind of conceptual searching.

Short: Suggestion -- only consider methods which (a) do old forms of scholarship in new ways; or (b) new forms of scholarship. The tools may be diffferent, but processes are similar -- forms of analysis, modelling. How it is done and how it is described. Analysis must be formally described, so that it can be clear, and transmitted. Tools and resources need to be described. This list as starting point.


In conclusion it was decided that Orlandi, together with Souillot, would arrange a meeting in Paris, Spring 1999:
(a) to define the range of problems pertaining to "Formal methods in the humanities"
(b) to prepare the chapter 6 of the European Handbook on Studies in Humanities Computing.

Also, the Members of the Steering Committee present at the workshop would ask the Committee to issue a statement concerning the desirability to teach the discipline Humanities Computing in the Faculty of Arts.