By John Kleefeld
[social-bio] In my previous post, I characterized the subject categories in the Requested articles page as idiosyncratic and mused that they might be better based on the Library of Congress Classification system. As it happens, Wikipedia does map some of its articles (pages) into the LCC system, and also provides several other methods of organizing knowledge. Some of these are well known, some less so. I want to discuss them because I think that instructors and students alike should be familiar with ways of finding knowledge beyond today’s default method of keyword searching. First, though, I want to talk about two approaches to knowing or learning, which philosophers Gilles Deleuze and Félix Guattari termed arborescent and rhizomatic in their 1980 book, A Thousand Plateaus. I will then consider how these approaches are converging in web-based retrieval systems, including Wikipedia.
The Arbor and the Rhizome
The word arborescent is Latinate for treelike. Tree diagrams, also called dendograms (Greek: dendro, tree + gramma, drawing) have depicted genealogical relationships (“family trees”) since medieval times; the “tree of life” or variants on it have been used at least since Carl Linnaeus classified relationships among organisms; and the ubiquitous “org chart” is a kind of inverted tree. Arborescent ways of knowing are said to emphasize totalizing principles (the notion that universal facts can be discovered and classified), binary opposition (e.g., male-female; predator-prey), and relationships characterized by discrete branching hierarchies rather than horizontal interconnections.
The rhizome is also a botanical metaphor. A rhizome (Greek: “mass of roots”) is a stemlike root lying on or just below the soil surface and having the ability to send roots and shoots from its nodes. Think of ginger, ginseng, and many grasses. The concept allows for multiple, non-hierarchical entry and exit points, with one thing potentially networked with many other things, sometimes in unpredictable ways. Ecosystems, power grids, and Web 2.0 are quintessentially rhizomatic systems. And Wikipedia is a quintessentially a Web 2.0 phenomenon, in allowing users to collaboratively create knowledge with multiple jumping-in and jumping-off points based on hyperlinked text and images.
Having identified the differences, I don’t want to convey the impression that rhizomatic ways of knowing are either entirely new or necessarily superior to arborescent ways.
As to newness, consider the dictionary, which, even before the 1755 publication of Samuel Johnston’s opus, came to be organized alphabetically rather than topically. This form of organization, which must at first have seemed arbitrary compared to its forebears, turns out to be surprisingly rhizomatic: I can open a dictionary at any page, read a word’s definition, and be led to another word and its definition through an italicized or bolded word in the definition. The idea is so powerful that it stuck: with today’s dictionary apps, you jump to the new word by clicking on it instead of turning pages. Consider also the polymath—a distinctly rhizomatous type of scholar. An early example was Aristotle, whose knowledge spanned physics, metaphysics, poetry, theatre, music, logic, rhetoric, politics, ethics, and biology. In a similar vein, Hildegard of Bingen was an abbess, writer, composer, philosopher, musician, mystic and medical writer. Leonardo da Vinci not only created the world’s most famous painting but is also credited with inventing the parachute, helicopter and tank. Erasmus Darwin, physician, poet and propounder of the evolutionary theory more rigorously developed by his grandson, penned a 4,400-line treatise that straddled all branches of science and technology and used a whimsical rhyming couplet structure footnoted with his scientific observations. Though polymaths are at risk of becoming an endangered species, recent interest in interdisciplinary studies may yet save them and their rhizomatic tendencies.
As to superiority, rhizomatic approaches let people access material in ways that are most intuitive for them and explore the links and pathways that interest them most. But while this can lead to more engaged learning, it can also be inefficient and leave knowledge gaps that more structured approaches would avoid. In other words, it can be both over- and under-inclusive.
I’ll illustrate this with an example from my own discipline, law—specifically, the “implied undertaking” rule in litigation. Litigating parties have “discovery rights,” an aspect of which is that they can get copies of each other’s documents that are relevant to the litigation. In Canada, such rights are generally accompanied by an implied undertaking not to use the documents for a purpose outside the litigation (also framed as “collateral or ulterior” to the litigation). Suppose, for example, a dismissed employee sues her employer, a company, for wrongful dismissal. Through the discovery process, she obtains a report that one of the company’s officials made to a government ministry, with comments that criticize the former employee’s ethical conduct. She then starts a new lawsuit against the company official for defamation. If the lawsuit is based solely on the report obtained in the wrongful dismissal action, it will likely constitute a breach of the implied undertaking rule and be unsuccessful. I say “likely” because the rule’s scope is subject to interpretation, may be overlaid with other court rules specific to a particular province, and may be subject to exceptions or court procedures in which a party can seek to be relieved of the undertaking.
Suppose I want to learn more about the rule. One way is to enter “implied undertaking” into a search tool like CanLII’s public-domain search engine or Westlaw’s proprietary product (available free with a U of S library account). This approach generates useful information—in the form of hundreds of case reports decided by judges. On skimming the cases, it becomes apparent that they are weighted according to relevance, which, in a mechanistic attempt at modelling human reasoning, means that they are ordered by the number of times the term “implied undertaking” comes up. This is neither a reliable nor an efficient way to learn about the scope of the rule or its exceptions, and even if I apply filters (for example, isolating the search to cases only in a certain province, or only at appeal-level courts), I am still likely to have to sort through dozens—perhaps hundreds—of cases. Even then, I will likely miss a lot of material because other terms are also used to refer to the concept, such as “deemed undertaking” or “undertaking as to confidentiality.” Isn’t there a better way to get an overview of the subject? Yes, and it’s arborescent.
That way is an encyclopedia article, in this case, from the Canadian Encyclopedic Digest (CED), another Westlaw product. One of the ways of accessing its material is through an alphabetically ordered, branching, hyperlinked table of contents. This requires some thinking about how the subject matter might be organized. After a bit of trial and error, the title “Discovery” seems an option. It has two headings, one for Ontario and one for the Western provinces. Scanning through the topics under either of these headings leads to the subheading for the article on the rule, as can be seen below (omitted subjects marked with an ellipsis).
Canadian Encyclopedic Digest
… D …
VII — Use of Discovery Evidence
1 — General
Clicking on the subheading pulls up an 11-paragraph article that succinctly outlines the rule and how it works, footnoted with a couple of dozen cases for those who want to drill down to get more information. This still leaves me with a fair bit of reading to do, but not nearly as much as by following the purely rhizomatic approach.
In fact, the story I have just told is oversimplified, because it doesn’t explain how in modern knowledge systems, rhizomatic and arborescent approaches are evolving towards convergence. For example, the “Implied or Deemed Undertaking” article in the CED also has a paragraph that links to headings in the Canadian Abridgment Digests, which cover cases on the rule from across Canada in digest (condensed) form. Further, even if I had only used my simple keyword search, “implied undertaking,” I could have arrived at the digest list through the first case on my search results by clicking on a sidebar headed “Related Resources.” Unfortunately, this list does not link back to the CED article on the topic, indicating that convergence is still a work in progress. But armed with the new information of a subject entitled “implied or deemed undertaking,” I can use that as a phrase in the CED search engine to find the encyclopedic article of the same name.
Other systems work in similar ways, with varying degrees of convergence. An example of a highly convergent system—one that allows for multiple searching approaches that are cross-linked to each other—is the Library of Congress Online Catalog. Its main web page allows you to choose from three searching methods: Browse, Advanced Search, and Keyword Search. Let’s suppose I want to look for books about collaboration in Wikipedia. I might start with something as simple as a Browse for subjects containing “Wikipedia” (see below).
This takes me to a page that lets me select from several subjects, including one just called “Wikipedia” with 24 records. In that list, I find two titles that seem to relate to my topic: Good Faith Collaboration: The Culture of Wikipedia and Global Wikipedia: International and Cross-Cultural Issues in Online Collaboration. I click on them to see their records:
Those titles may be enough to satisfy me, but I won’t have done a very thorough job of researching if I don’t also look at the information below the book title. The first thing I notice is that two books have different Library of Congress (LC) classification numbers.
The first starts with AE100. LC Class A is “General Works,” and subclass AE is “Encyclopedias.” The second starts with ZA4482. LC Class Z is “Bibliography, Library Science” and subclass ZA is Information resources/materials. Thus similar-sounding material may be indexed under different numbers—in the one case, the focus is more on encyclopedias and in the other more on user-generated knowledge—but the LC system allows me to search on both those numbers to expand the search pie. I can do that simply by clicking on the number beside “LC classification (partial)” or by searching for it as a call number in the Browse feature (see below).
If I do that, about 15 records come up for AE100 (followed by higher-numbered records; as the name implies, this feature is similar to browsing books on a shelf). If I do the same thing for ZA4482, about 20 records come up. I have now expanded my search results from 24 to 35 potentially relevant records.
If searching by call number seems too geeky to you, a different option presents itself in the form of related subject headings. There are six of these for the Good Faith book and five for the Global Wikipedia book, with an overlap in the first subject heading, “Wikipedia.” All of these are hyperlinked and take me to the subject listings in the LC catalog. For instance, if I click on the subject heading “Authorship–Collaboration–Case studies” under the record for the Good Faith book, I get three further entries, including one German book that falls in the ZA4482 class and a book on textual curation that falls in the AE1 class.
In my next post, I will show how Wikipedia uses the LCC system as well as several other organizational systems that allow for the convergence of both arborescent and rhizomatic ways of knowing.
John Kleefeld is an associate professor at the College of Law and a 2017 teaching fellow at the Gwenna Moss Centre for Teaching Effectiveness, where he is coordinating a campus-wide project on integrating Wikipedia assignments into course materials. Portions of this blog series are from an article that he and a former law student wrote about using a Wikipedia assignment for class credit. See J. Kleefeld and K. Rattray, 2016. “Write a Wikipedia Article for Law School Credit—Really?” Journal of Legal Education, 65:3, 597-621.
 I am indebted to a colleague, Heather Jensen, for the following example.
 See Goodman v Rossi (1995), 24 OR (3d) 359 (CA), 1995 CanLII 1888.
 My first search result was Jutte v Jutte, 2007 ABQB 191, reflecting the large number of times the term “implied undertaking” appeared in the case. Under “Related Resources” is the hyperlinked heading “Abridgment digests and classifications for all levels of this case.” Clicking on that leads to the following branch of the Canadian Abridgment Digests (the numbers of cases digested for each heading are shown in parentheses):
CIV Civil practice and procedure
CIV.XII.1 Introductory (413)
CIV.XII.1.d Deemed or implied undertaking (153).
Clicking on the last of these headings leads me to the digests, sorted in reverse date order.