CompOnt Tenet 3: Human Ontologies are Built out of Symbols

Sym­bols — from core sym­bols like the Vir­gin of Guadalupe to abstract ones like the white­ness of Melville’s whale — fix and gen­er­ate onto­log­i­cal cat­e­gories. How and why this hap­pens is a ques­tion of deep inter­est to me, but that it is true seems obvi­ous and well estab­lished. Human beings cre­ate sym­bols like plants pro­duce oxy­gen, and sym­bol for­ma­tion is inex­tri­ca­bly bound to a defin­ing trait of human beings — rich, dis­cur­sive, and always already metacog­ni­tive lan­guage (human beings have always talked about talk­ing, a prac­tice that must be regarded as intrin­sic to human lan­guage). Lin­guis­tics up to now, wed­ded as it has been to a Chom­skyian Carte­sian­ism, has missed this role, although philoso­phers have not lost sight of it. As Ricoeur wrote, “the sym­bol gives rise to thought.”

The rela­tion­ship between lan­guage and sym­bol­ism is com­plex and (still) not well under­stood. My own view is close to that of the (admit­tedly dis­cred­ited in its orig­i­nal form) gen­er­a­tive seman­tics school, asso­ci­ated with George Lakoff. I believe that cat­e­gories and rules are in some way gen­er­ated by the trans­duc­tion of mean­ing that takes place between neural rep­re­sen­ta­tions of con­crete objects. Dis­cur­sive lan­guage — not “deep gram­mar” — tries to fix these mean­ings in propo­si­tional form, but the sym­bolic sub­strate has a dynamic qual­ity, in no small mea­sure do to its adap­tive nature in response to what Merleu-Ponty called the “pri­macy of perception.”

With writ­ing and then print­ing, and the monop­o­liza­tion of explicit knowl­edge, in the form of writ­ten records, ref­er­ence works, etc., by gov­ern­ments, uni­ver­si­ties, etc., the rela­tion­ship between dis­cur­sive fix­a­tion and embod­ied sym­bols becomes ten­u­ous and con­tested, result­ing in a mind/body prob­lem unfa­mil­iar to rit­ual societies.

In any case, a num­ber of prac­ti­cal obser­va­tions fol­low from this tenet, which I will quickly enu­mer­ate, and hope­fully take up later:

1. Human ontolo­gies are not plans.
2. Human ontolo­gies are overde­ter­mined. That is, there is always more than one way to express an ontol­ogy The fix­ing of mean­ings will always fail if the goal is to cre­ate non-overlapping, non-redundant descriptions.
3. Human ontolo­gies are rhi­zomic. In their nat­ural form, ontolo­gies are not hier­ar­chi­cal. Rather, the hier­ar­chi­cal rep­re­sen­ta­tion is one form of seri­al­iza­tion that works well because of its anal­ogy to kin­ship (see Durkheim and Mauss, Prim­i­tive Clas­si­fi­ca­tion).
4. Human ontolo­gies are local and sit­u­ated.
5. Human ontolo­gies evolve.

Posted in Comparative Ontology | 3 Comments »

CompOnt Tenet 2: Social Life Makes the Categories

Just as insti­tu­tions require shared ontolo­gies to func­tion, so are insti­tu­tions involved in cre­at­ing the cat­e­gories that nake up ontolo­gies. Admit­tedly, assign­ing agency to insti­tu­tions poses a num­ber of ques­tions that need to be answered; I won’t attempt that here. Essen­tially, I fol­low Mary Dou­glas (see video below), espe­cially her How Insti­tu­tions Think. The cat­e­gories and rules that com­prise human ontolo­gies fol­low and enable a prac­ti­cal logic that in turn enables sus­tain­able human interaction.

A corol­lary idea to this is that ontolo­gies exist (in large part) to medi­ate social action. They are the result of human beings’ mutual cal­i­bra­tion of indi­vid­ual cog­ni­tion through col­lec­tive inter­ac­tion. Typ­cially, this cal­i­bra­tion takes place through rit­ual. But media (old and new) — which grow out of but even­tu­ally dis­place rit­ual — also take on this role. (McLuhan’s fre­quent ref­er­ence to rit­ual to describe the effects of new media is telling.)

“Social Life Makes the Cat­e­gories“
The late Mary Dou­glas being inter­viewed in 2006
about her book Purity and Dan­ger.
See the full inter­view and films at Sci­enceS­tage.

Posted in Comparative Ontology | No Comments »

Tenets of Comparative Ontology 1: People already have ontologies.

This is the first in a series of posts in which I define some of the tenets of com­par­a­tive ontol­ogy, in order try to flesh out its sig­nif­i­cance to the work of mak­ing and using RDF vocabularies.

The over­whelm­ing con­clu­sion to be drawn from the ethno­graphic record is that human beings are sur­pris­ingly struc­tured in their think­ing and behav­ior, even when that behav­ior seems to be ran­dom and non-linear. Although it is a com­mon­place to observe that social life is inher­ently messy, unpre­dictable, and resis­tant to cap­ture by physics-like laws (post-Einsteinian included) it remains true that pat­terns of cul­ture are remark­ably wide­spread and per­sis­tent. Lan­guages, mar­riage prac­tices, cal­en­dar sys­tems, gift exchange sys­tems, mar­kets, etc. — essen­tially, any func­tional human insti­tu­tion — all rely on shared cat­e­gories and rules to oper­ate, and these are dis­cov­er­able and describ­able. The mis­take of the struc­tural­ists was to con­ceive of these cat­e­gories and rules as log­i­cal in a for­mal, almost scholas­tic sense, like a plan that agents fol­low strictly. Instead, it is more likely that they exist as dis­po­si­tions that con­strain behav­ior and encour­age improviza­tion, as Bour­dieu describes in his idea of the habi­tus (which he got from Mauss, by the way).

To the extent that for­mal RDF ontolo­gies are meant to medi­ate human-computer inter­ac­tion (and not sim­ply allow com­put­ers to share infor­ma­tion), ontolo­gies should be designed to inter­dig­i­tate with the cat­e­gories of their human par­tic­i­pants. Machine ontolo­gies should be inter­op­er­a­ble with human ontolo­gies. They should be designed to encour­age the sym­bi­otic devel­op­ment and evo­lu­tion of human col­lec­tive rep­re­sen­ta­tions (to use Durkheim’s expres­sion), given the role of the net­worked com­puter and com­puter net­work as an insti­tu­tion in its own right.

Posted in Comparative Ontology | No Comments »

Two More Thoughts on Comparative Ontology

Fol­low­ing up on my last two posts (here and here), I thought I’d flesh out a lit­tle more what I mean by Com­par­a­tive Ontol­ogy — and, at this point, maybe I need a han­dle for the con­cept, like CO or “comp-ont.”

First thought: I imag­ine that a comp-ont vocab­u­lary could be accom­plished in two ways: (1) by just using exist­ing vocab­u­lary def­i­n­i­tion lan­guages (RDFS and OWL) to cre­ate a new vocab­u­lary, or (2) by cre­at­ing a new vocab­u­lary def­i­n­i­tion lan­guage using, say, Rod­ney Needham’s cod­i­fi­ca­tion of Eng­lish Struc­tural­ism in Sym­bolic Clas­si­fi­ca­tion.

As an exam­ple of type 1, one might deploy the implicit ontol­ogy of Kautilya’s Arthashas­tra (with maybe a lit­tle of Machiavelli’s Prince thrown in) using OWL to cre­ate an “Enemy of an Enemy” (EOAE) vocab­u­lary. This would be a use­ful vocab­u­lary for describ­ing polit­i­cal agents, link­ing them through the neg­a­tive tran­si­tive logic of polit­i­cal alliance for­ma­tion, with the result that you could actu­ally pre­dict or sug­gest alliances that don’t yet exist. I can imag­ine this vocab­u­lary describ­ing peo­ple (politi­cians, pun­dits, etc.), ideas, insti­tu­tions, etc., and applied to daily news sources.

As an exam­ple of type 2, my ques­tion would be whether con­structs like anal­ogy or metonymy could be con­structed on top of set the­ory (OWL) or if it would require a whole new sys­tem (and there­fore logic of inference.)

Sec­ond thought: It strikes me that one way to bring home the value of comp-ont is to view it in the con­text of knowl­edge man­age­ment (KM). Clearly there is a strong con­nec­tion between KM and ontol­ogy (and the seman­tic web and linked data). But within KM, it is clear that ontolo­gies can’t just be log­i­cal. Ontolo­gies must be prac­ti­cal, and intel­li­gi­ble to the peo­ple in an orga­ni­za­tion who engage in ontology-mediated knowl­edge trans­fer. They must serve the pur­pose of medi­at­ing between indi­vid­ual and col­lec­tive mem­ory. But this is pre­cisely what the great onto­log­i­cal sys­tems from the ethno­graphic record pro­vide — from Aus­tralian totemic sys­tems to Mayan cal­en­dars. Viewed from the per­spec­tive of social mem­ory, then, it makes a great deal of sense to build KM ontolo­gies on gen­eral prin­ci­ples adduced from anthro­po­log­i­cal exam­ples, and described in terms of struc­tural­ism.

Posted in Comparative Ontology | No Comments »

Comparative Ontology (ii)

So, I make the claim in my pre­vi­ous post on Com­par­a­tive Ontol­ogy that com­par­a­tivsim will help us make bet­ter ontolo­gies and tools for using them. I real­ize I did not actu­ally make that case. Let me try to do that now, although, to be hon­est, I view the claim as an inter­est­ing direc­tion to fol­low (i.e. a hunch), not an estab­lished fact.

My argu­ment is pretty sim­ple. The Seman­tic Web, as every­one knonws, has a big prob­lem — in order for it to suc­ceed, pro­duc­ers of web con­tent have to mark up texts using RDF, with its long URIs and arcane vocab­u­lar­ies. Most folks think this is a deal killer. In addi­tion, many agree with Shirky’s claim that ontol­ogy is over­rated in the first place. (The meme even makes an appear­ance in Michael Wesch’s viral “The Machine is Us/Using Us.”)

I hold out hope, though. I think tools can be cre­ated to make the data entry part a lot easier–withness Dru­pal 6. I also think con­tent cre­ators show a remark­able ten­dency to use tags, enter meta­data, and engage in oth­er­wise tedious SEO tech­niques if they know that they will get some­thing out of it. This ten­dency is espe­cially strong among dig­i­tal humani­tists, dig­i­tal librar­i­ans, and other cre­ators and cura­tors of dig­i­tal col­lec­tions. (Think of the time peo­ple have put into TEI!) Right now, they don’t see a lot a ROI from using RDF. But that is chang­ing, and faster than many think.

As is often remarked, it’s a clas­sic chicken and egg argu­ment. But such prob­lems are actu­ally lib­er­at­ing, since you can inter­vene at any point in the cycle and have an effect. One junc­ture in the cycle is in the design of vocabularies.

The prob­lem now is that vocab­u­lar­ies and ontolo­gies are just not sticky enough, and I think a big rea­son for this is that they are writ­ten from the per­spec­tive of sym­bolic logic and set the­ory. These frame­works, though com­pre­hen­sive, are not human (enough). On the other hand, there is a whole tra­di­tion of describ­ing cog­ni­tive sys­tems called struc­tural­ism, which has the advan­tage of being empir­i­cally grounded in the long­stand­ing com­par­a­tive study of cul­ture. It stands to rea­son that if we cre­ated vocab­u­lar­ies using more “human” constructs–such as anal­ogy, metonymy, etc.–and vocab­u­lar­ies that rec­og­nized the role of verbs in pred­i­cates, then we might be able to pro­duce more sticky ontologies.

It also opens a way to har­vest the latent ethno­graphic data con­tained in folksonomies.

But I’ll leave it at that for now.

Posted in Comparative Ontology | 1 Comment »

Comparative Ontology

The Seman­tic Web has pop­u­lar­ized the con­cept of ontol­ogy. The usual def­i­n­i­tion one sees for the term is Gruber’s “spec­i­fi­ca­tion of a con­cep­tu­al­iza­tion” (Gru­ber 1993). The def­i­n­i­tion could use some unpack­ing. Essen­tially, it refers to the idea that the terms used by AI agents should be defined by ref­er­ence to a shared schema which clas­si­fies those terms. The shared schema, or spec­i­fi­ca­tion, sim­ply defines how terms are related to each other in a graph. More than a sim­ple data model, where the rela­tion­ships between enti­ties or objects are not explic­itly defined, an ontol­ogy also names the pos­si­ble rela­tion­ships between terms. So, instead of just hav­ing “address” as a field of the entity “per­son,” an ontol­ogy will explic­ity assert that a per­son has an address. So verbs (which are often buried in “pred­i­cates”) are an impor­tant part of ontolo­gies. In fact, one way to think of an ontol­ogy is to call it a data model with verbs. (Sort of.)

Ontolo­gies are use­ful for pro­gram­ming agents, which are just pro­grams that con­sume and pro­duce infor­ma­tion on their own. (I call them infophages.) So, when you pro­gram an agent to respond to a term, you have the agent’s pro­gram refer to an ontol­ogy to dis­am­biguate the term among syn­onyms, and to make log­i­cal infer­ences about the term. For exam­ple, if the agent encoun­ters the term “Socrates,” it will be found to be spec­i­fied as a mem­ber of the class “Human” and since “Human” will have the prop­erty “is Mor­tal,” the agent can trans­fer the prop­erty to “Socrates” too. (Truth, or valid­ity, is just trace­abil­ity within a ref­er­ence graph.)

Or, if the agent encoun­ters the word “Madonna,” and that term is for­mally spec­i­fied as part of an ontol­ogy (say, by using Tim Berners-Lee’s URI method), the agent will be able to trace its class to either “Pop Singers” or “Reli­gious Fig­ures” or “Chris­t­ian Iconog­ra­phy” or whatever.

Sometimes–often–ontologies are described as if they pro­vide “mean­ing” for agents. Even the term Seman­tic Web implies this. When intro­duced to novices, the Seman­tic Web is often intro­duced as bring­ing “mean­ing” to the web. I find this expla­na­tion mis­lead­ing at best. Mean­ing is much to too impor­tant of a word to describe what is going on here, which is sim­ply a layer of clas­si­fi­ca­tion being added to what would other wise be a list of terms. Ontolo­gies are just sys­tems of clas­si­fi­ca­tion for terms, a second-order set of terms that (1) increases the prob­a­bil­ity that an agent can process a term in a way that users will expect, and (2) adds a layer of con­nec­tiv­ity, above the raw ver­biage on the web, that decre­ses the aver­age dis­tance, or degress of sep­a­ra­tion, between any two terms.

So, ontolo­gies have two prop­er­ties that make them use­ful and, as I want to argue, use­fully viewed from a com­par­a­tive per­spec­tive. First, they intro­duce verbs into the mix of data mod­el­ling. Sec­ond, they are just sys­tems of clas­si­fi­ca­tion–like Aus­tralian totemic sys­tems, Mayan cal­en­dars, and West­ern philo­soph­i­cal ontolo­gies such as Aristotle’s and Leibniz’s.

Com­par­a­tivism will help us cre­ate bet­ter ontolo­gies, and bet­ter sys­tems for using ontolo­gies, in two ways.

First, com­par­a­tivism can train our atten­tion on the great rab­bit war­ren of words and mean­ings that lies at the heart of the osten­si­bly neat world of triples–I refer to the pred­i­cates. Although the nouns that com­prise the vocab­u­lar­ies for sub­jects and objects in RDF can be neatly spec­i­fied as ana­lytic tax­onomies of terms, pred­i­cates are sub­ject to no such rules, and can con­tain within them­selves whole sen­tences. Unlike the nouns, pred­i­cates mask a great deal of what Kant would call “syn­thetic judge­ments.” I pro­pose a soci­olin­guis­tic approach to the use of verbs embed­ded in the pred­i­cate sys­tems of linked data vocab­u­lar­ies that will pro­vide a bet­ter basis for craft­ing predicates.

Sec­ond, com­par­a­tivsm can help us move beyond set theory–useful as it is–to con­sider other total­iz­ing schemes for orga­niz­ing ideas, such as those described by anthro­pol­o­gists. I men­tioned totemic sys­tems, cal­en­dars, and West­ern ontolo­gies. But there are many oth­ers to con­sider. What is more, anthro­pol­o­gists have done a great deal to describe, clas­sify, inter­pret, and even explain these. Such an approach would be grounded in the fol­low­ing works, all of which define or grow out of the struc­tural­ist tra­di­tion within anthropology:

1. Emile Durkheim and Mar­cel Mauss, 1903, Prim­i­tive Classification.
2. Louis Dumont, 1966, Homo Hier­ar­chi­cus.
3. Michel Fou­cault, 1966, The Order of Things.
4. Claude Lévi-Strauss, 1968, The Sav­age Mind.
5. Vic­tor Tut­ner, 1969, The Rit­ual Process: Struc­ture and Anti-Structure.
   
6. Mary Dou­glas, 1970, Nat­ural Sym­bols.
7. Pierre Bour­dieu, 1977 [1972], Out­line of a The­ory of Prac­tice.
8. Alfred Gell, 1975, Meta­mor­pho­sis of the Cas­sowaries.
9. Rod­ney Need­ham, 1975, Poly­thetic Classification.
10. Edmund Leach, 1976, Cul­ture and Communication.
11. Mar­shall Sahlins, 1978, Cul­ture and Prac­ti­cal Rea­son.
12. Jad­ran Mim­ica, 1988, Inti­ma­tions of Infin­ity.

And so forth.

Posted in Comparative Ontology | No Comments »

Of documents, data and “semantic energy”

Gard­ner Camp­bell has recently writ­ten a cou­ple of posts on the oppo­si­tion between “doc­u­ments” and “data.” I think this is actu­ally a very pro­found topic that gets at the heart of what is at stake in aca­d­e­mic tech­nol­ogy, both philo­soph­i­cally and tech­ni­cally, and the approach we take to Gardner’s con­cerns reveal a great deal about what we think the goals of aca­d­e­mic com­put­ing are. Regard­ing the tech­ni­cal issues, I’ve pre­sented a cou­ple of talks on the topic which you can see here, here and here. In this post I want to say some­thing about the philo­soph­i­cal dimension.

I think the document/data oppo­si­tion is really about the dif­fer­ence between the struc­ture of the struc­tural­ists and the dis­course of the hermeneu­ti­cists, and there­fore car­ries all the weight of that dis­cus­sion. As an anthro­pol­o­gist, I was steeped in the struc­tural­ism of Levi-Strauss (and the post-structuralism of Sahlins), but I dis­cov­ered on my own (and with the help of the incom­pa­ra­ble Larry Bouchard) the won­der­fully con­struc­tive crit­i­cisms of struc­tural­ism by Paul Ricoeur. I think Gard­ner echoes Ricoeur’s con­cerns when he expresses a wari­ness toward the RDF-world of the Seman­tic Web, ver­sus the more dis­cur­sive world of Web 2.0. But I also think Ricouer opens the way for think­ing about the rela­tion­ship between the two approaches, as two moments in an inter­pre­tive arc that tra­verses the “world” of the onto­log­i­cal and the always sit­u­ated domain of dis­course, the event within which mean­ing is not merely a repro­duc­tion of a prior struc­ture, but an achieved social real­ity. I think that is some­thing like what Gard­ner is get­ting at when he speaks of the “seman­tic ener­gies” of the two approaches.

Posted in Media Theory | No Comments »