by Bill Uzgalis

Bill Uzgalis (BU): Tell us a bit about yourself. Where did you get your philosophical training and what were your interests before you became involved with projects which connect philosophy and computers?

Tony Beavers (TB): I earned my M.A. in philosophy from Trinity College in Hartford, Connecticut in 1987 and my Ph.D. from Marquette University in 1990. While working on my M.A. I worked as a computer programmer designing database systems for small businesses. My philosophical training was largely in classical phenomenology. When the World-Wide Web was born in the mid-1990’s, I discovered that much of what I had learned from phenomenology was applicable to the database work that I had done earlier and to understanding the nature of hyperlinks and cyberspace. The rest of my work in computers and philosophy follows from that discovery.

BU: Can you explain some of the connections you see between classical phenomenology and databases, hyperlinks and so on?

TB: Phenomenology is a method used to describe consciousness, its proper "field", the various "objects" within it and the way all these elements relate. The relationships between consciousness and its objects are governed by intentionality, which is analyzable in terms of a correlation between Noesis and noema. Cyberspace — not the physical machines on the Internet, but the arrangement of information, both in and out of electronic media — is structured in such a way as to lend itself to phenomenological analysis. Hyperlinks are intentional, in a way. Their meaning resides in pointing to another document (or other information source), and their direction is steered according to the wishes of an intelligent author. Consequently, hyperlinks are not random and arbitrary. A final causality guides them that allows us to operate on the assumption that objects linked in cyberspace are linked for a reason. Already, we see that, at least, two kinds of intentionality govern the hyperlink. The actual connection to a document and the reason why this particular connection exists. Together, these intentionalities combine to form a series of complex relations among the various objects on the Internet.

The intentional character of the hyperlink allows us to draw horizontal lines around sections of the Internet, in effect separating domains of discourse. If that sounds vague, you might imagine a hypertext document that contains, say, twelve links to other documents. Because of the intentionalities that steer hyperlinks and relate objects on the Internet, we can characterize this document plus the twelve that it links to as belonging in a community. Together, these thirteen documents constitute a field that is hypertextually separable for the sake of analysis from the rest of the Internet.

Understanding these intentionalities (and others) and quantifying them allows us to construct algorithms for searching and indexing so that we can use computers to synthesize a meaningful [section of] cyberspace, a region of the Internet isolated for a particular purpose.

Following a similar type of phenomenological analysis (phenomenology applied to cyberspace, instead of consciousness), we should also be able to build indices that both auto-organize and regulate the quality of information. These tasks have become the recent focus of our work in the Internet Applications Laboratory.

In 1995, foremost on my mind was the possibilities that the Internet provided for building large-scale collections of easily accessible information. A central hub on the Internet dedicated to particular topics could be used to screen the input going into a search engine, making it act as a filter between an interested user and a sub-section of the broader Internet. Understanding this possibility led to the development of the Argos algorithm.

BU: Search engines seem to be at the heart of your efforts to connect philosophy and computing. I think I heard of you originally in connection with the Hippias project. But I gather Argos was an even earlier effort in this direction. Could you describe them and how Hippias developed out of Argos.

TB: Argos was a search engine I designed in 1996 which was dedicated to ancient and medieval studies. Hippias is built on the Argos model.

The procedures that limited the range of Argos to a particular area while adding a dimension of quality control are quite straightforward. To make Argos work, we wrote a web crawler that searches 12 carefully selected Internet indices in ancient and medieval studies that are known for their selectivity. Argos was designed to crawl these 12 sites plus everything to which they link. This procedure, in turn, passes editorial control over the contents of Argos to the editors of the 12 "associate sites" just mentioned. When an editor adds a link to their site, Argos picks it up. When they delete a link, Argos drops it. In this way, the decisions of the trained professional that administrate the associate sites determine the contents of Argos.

Having developed the software for Argos, building another search engine in a different area was as simple as selecting a different group of associate sites. So, in 1997, I contacted Peter Suber asking him to be general editor of an Argos-style search engine in philosophy. I would serve as managing editor. We built Hippias in a weekend, and it has been running ever since.

BU: I gather that the Noesis project was an outgrowth of the Hippias project. Why don’t you tell us how this came about. What, exactly is the difference between the two projects?

TB: While the Argos/Hippias model does limit the range of search requests to pre-defined topics, it does not permit users to browse the dataset. In addition, it cannot control the formatting of titles, the sectioning of document types, and searching across only subsections of the database.

Sometime in 1997, it occurred to me that shifting from (what we call) the associates model to a central database model would give us greater control over the dataset and allow better precision in quality-control. Noesis uses the database model. Each item in Noesis is hand-catalogued, a procedure that has been automated as far as possible so that it takes about a minute to catalogue a URL. This allows us to identify various types of documents. Noesis users can filter return sets to include primary texts or essays, or any combination of several different document types. In addition, it allows us to classify items by author or topic and allows users to search sub-sections of the database. This is fruitful because philosophical vocabulary often shifts its meaning depending on the area in which it is used: the term "metaphysics," for instance, means one thing in contemporary continental philosophy and quite another in Aristotle. By pre-selecting an area to search, users are, in effect, limiting the semantic range of vocabulary in advance of a search, and this, in turn, allows for search return sets that are more suited to a user’s needs.

Another important difference is that Hippias is dedicated to limiting the range of search requests to philosophy. It uses a type of quality-control, but not a refined one. Hippias is primarily based on relevance, not quality. Noesis, on the other hand, is dedicated to philosophical content only. It does not index many elements that Hippias does, such as the pages of professional associations and course syllabi. Soon, Noesis will pick up these other elements, but in a more refined way, allowing the same sub-set searching and type screening that it uses for other elements.

Incidentally, Noesis, too, was based on an earlier project called "Exploring Plato’s Dialogues" (see the review of this website by Daniel W. Graham in this issue-ed.) that uses a Noesis-style search engine to supplement the study of Plato. This engine was a test case to determine the viability of expansion to a large-scale project capable of serving the profession of philosophy globally. We have always had to deal with questions of size and domain. How should the topics be limited? Should a single site serve philosophy only or all of the humanities? What about all of academia? These are difficult questions to answer, since each model offers its own set of benefits and problems.

We have designed Noesis in such a way that it can expand naturally into other areas. We are uncertain of where to head next, but it makes some sense to expand into the other humanities before moving elsewhere.

BU: There seem to be certain central ideas that I connect with the Noesis project. One is that material is decentralized. Another is that the domain of the search is limited to what has been determined to be relevant to genuine philosophical inquiry. And I think Noesis represents this in its most general form—that is I can apply to the entire field of philosophy. What you like to talk about these aspects.

TB: Noesis has no content of its own. Instead, it is a synthesizing hub that draws resources from around the world into a central place. The benefit of this is that someone can, for instance, compare a translation of Descartes’ 3rd Meditation that is located in Australia with another that is located in France. This means that we can "virtually" gather resources no matter where they originate. A single Internet page can serve the world, since it is copied each time it is needed and sent via the web to the person requesting it.

I discovered that much of what I had learned from phenomenology was applicable to the database work that I had done earlier and to understanding the nature of hyperlinks and cyberspace. The rest of my work in computers and philosophy follows from that discovery.

Another consequence of this fact is that we don’t have to manage the actual contents of resources. Each author can do this himself, or the journal editor, in the case of an on-line journal. Because the work is spread across so many people, Noesis is efficient. It is not a collection of resources, but a collection of links. However, these links are intentional; they point to resources. When clicking a link brings a particular resource to a user, even a rudimentary listing of links is transformed into a collection of resources. For this reason, I don’t like the names "search engine" and "index". They both obscure the significance of what projects like Noesis can do. A large and organized Internet index is a library of resources in which its contents may physically reside all over the Internet. Even the word "library" has connotations that do not belong here. We have yet to develop an adequate language to describe what an index like Noesis is. Consequently, the true power of such centralizing hubs is not readily apparent.

We also have to deal with a different set of psychological expectations. If all of the items linked to by Noesis were formatted the same way—and I’m not suggesting that they should be—Noesis would appear to be a massive effort on the part of many philosophers to disseminate their research to the public quickly and efficiently. The absence of a common format makes Noesis look like a mere search engine, rather than an encyclopedic collection of philosophy. Hubs like Noesis can do most if not all of the organizing work to help foster a profession on-line, even though the various authors do not formally work on the project. With Noesis and other efforts like it, a global network of computers can take care of dissemination, leaving the authors to do their research and write their papers.

BU: "I noticed on the Noesis site that the current version is 2.0 and that it is a "beta" release. When will it come out of beta and what is in store for Noesis 3.0?"