- From: Lynn Andrea Stein <las@olin.edu>
- Date: Wed, 13 Jun 2007 10:14:19 -0400
- To: "John Black" <JohnBlack@kashori.com>
- Cc: Pat Hayes <phayes@ihmc.us>, Sandro Hawke <sandro@w3.org>, Ian Davis <lists@iandavis.com>, <semantic-web@w3.org> <semantic-web@w3.org>
There is a gap here between * those who argue that URIs must be used in a particular way (no punning or homonyms, etc.) because that is *how* *they* *work* and * those who argue that no such set of rules can ever accomplish the task of securely pinning a particular URI on the peach I am about to eat or the planet Venus or the Roman goddess of the same name. Strangely, I think you're both at least partly right, but for different reasons. Pat's argument that logic (and statements about the URIs) will never make the URI stick is dead on; that's not what logic is *for*. But those who say that usage needs to conform to certain rules are, I think, simply trying to articulate how actual meanings and bindings work in the non-web world. The problem comes when, like Tim, they want to locate the right to make those assignments of meaning and bindings to the "owner" of the URI; this only works in some cases. I have come to believe that we can't understand meaning on the web -- including the references of URIs -- without a better understanding of meaning in the world. It turns out that John Searle, of all people, has a fairly articulate description of the kind of meaning-making that seems to be involved here. I'm appending a part of a draft paper I wrote to try to explain this connection. For those without the patience to read the whole thing, here's the punch line: In the world, there are many things that *mean* only because we collectively agree that they should have that meaning. For example, an IOU, the Crown of England, and playing a trump card in a friendly game of bridge each carry significance only because of a collective attribution of that significance (among at least the giver and receiver of the IOU, the majority of the English population, and the players at the game of bridge). This meaning, far from being tenuous, can be so strong that entire institutions -- legal systems, banking systems, international relations, nuclear deterrence, environmental treaties -- are built upon it. The semantic web (and, in my thinking, the web itself) carries similarly socially constituted meaning. However, the constitutive rules are somewhat less clear here -- we're still evolving them -- and the stability of the structure is less clear in some places than in others. What we're arguing about in this thread is largely the nature of what those rules ought to be. While we do need to reach some consensus, we don't necessarily need a single agreement: meanings can vary from one group to another and from one context to another, though interoperability becomes more difficult. (Witness the current US uncertainty regarding the meaning of marriage....) It would, however, be a mistake to believe that we can control all possible social meaning by dictating a set of rules. There will always be groups of punky teenage anarchists (or SWeb visionaries) determined to create and live within their own social realities and, depending on how little interoperability they need, they will be well within their rights to do so. Lynn ---------------- = The Social Reality of the Web = [Searle, J. R. 1999. The Construction of Social Reality. New York: The Free Press.] describes the way that certain facts in the world are built out of mutual (social) recognition. This accounts for phenomena like money, marriage, and games, all situations in which reality as we know it and the actual functioning of systems come into being through mutual human consent. For example, an athlete's winning of a major sports event is not an intrinsic fact of physical reality but rather a social agreement as to what constitute the rules of the game, the proper winning conditions, etc. (Witness the overturning of certain victories when it is retrospectively determined that these rules may have been violated.) Searle refers to this as "the construction of social reality" and to the facts of these situations as "institutional facts." Searle refers to statements like "Mount Everest has snow and ice near the summit" as "brute facts", i.e., facts that are true independent of social practices, beliefs, or any other human institutions.[1] In contrast, statements like "the piece of paper in my pocket is a five dollar bill" or "the Red Sox won the 2004 World Series" are inherently dependent on social constructions such as monetary systems and major league baseball. Searle's book gives an account of the necessary preconditions and practices surrounding such "social reality" and "institutional facts". I will demonstrate that Searle's framework can also ground the notion of meaning in the World Wide Web. == Constituting Social Reality (Institutional Facts) == Searle builds his notion of social reality out of three constituents. Each of Searle's requirements for social reality is also met by the web. A. The assignment of function. By this, Searle means that (some) things have purposes within goal-based or value-based systems. Social reality will rely on objects with normative function, specifically, the ability to represent or stand for something else. Searle's assignment of function comes easily to web statements and terms. Virtually all web constructs have purposes such as promising, committing, advertising, purchasing, relating, and above all, representing. Semantic web terms and URIs are generally used for just this representational ability. B. Collective intentionality, wherein a group mutually intend something. This involves public - or at least mutual - recognition and joint, rather than simply parallel, commitment. Collective intentionality on the web is inherited from collective intentionality off the web. For example, web users and designers - as well as the stockholders of BrazilNuts, Inc. - collectively intend that it should be possible to buy kitchen appliances via their e-commerce site. The owner of a URI (according to TBL) or its users (where a group use the URI consistently for their own internal purposes) intend something by its use. C. Constitutive rules and institutional facts. These are rules that describe what makes something so. For example, it is constitutive of tic tac toe that two players take turns placing distinct marks into a 3x3 grid. This isn't a fact about tic tac toe; it is a fact that establishes the game of tic tac toe (or constitutes it). Entities built of constitutive rules (possibly also with more mundane components) are institutional facts. Being a physician is an institutional fact: it consists in being licensed by the state, although it may also require adequate training, etc. Simply practicing medicine, however, does not constitute the institutional fact of being a physician. The web, too, has constitutive rules and institutional facts; indeed, essentially all web facts are institutional facts. For example, the model theory of RDF [Hayes 2004] indicates how RDF is to be interpreted. Nothing prevents it from being interpreted differently, but if a program does interpret some (nominal) RDF triples contrary to the constitutive, that program is not treating the triples as RDF at all and has no recourse when objecting to what it has gleaned - unsupported - from the data. == How this Adds Up == Searle's social reality depends on a specific form of constitutive rule: X counts as Y in (context) C Specifically, a social reality like marriage comes into existence when a community collectively intends something like "standing in front of an authorized person and declaring 'til death do us part' constitutes becoming married" (and the persons who engage in these actions become spouses). Performing this action in this context creates the institutional fact of a marriage -- i.e., it gives reality to the marriage -- because collectively we take the actions to constitute performing a marriage and the consequence to be that the two individuals so joined now stand in a certain relationship to one another. Web activity creates institutional reality similarly. On the web, pressing the accept button counts as making a purchase from BrazilNuts. If you doubt this, try refusing to pay for it and you will shortly discover the collective intentionality behind these institutional facts. The intense work going on to establish the semantics of web reasoning, the appropriate query language and response, the representation of data on the web are all reflective of the desire to create constitutive rules for semantic web technologies. These rules will establish what statements count as asserting what truths or commitments; what actions count as making offers; what conclusions count as licensed and by whom; and how a particular URI can come to stand for ("be stuck on") a particular real world object. Are the rules of the (semantic) web constitutive? Absolutely. They specify how terms are intended to be interpreted; if you interpret them otherwise, you will have no recourse when the world refuses to uphold your conclusions. The constitutive rules of how the web is intended to work, together with the social processes of standards bodies, legal systems, and de facto practice, create the institutional facts by which the web actually works. == Consequences of Institutional Facts == Searle identifies several other properties of social realities/institutional facts that bear mention here. Again, social reality on the web shares many properties with Searle's institutional facts: a. Institutional facts can be built out of other institutional facts. If X counts as Y in C, Z may later be made to count as X or form a part of C. The "authorized person" before whom one becomes married is an example of a status created by other institutional facts. Institutional facts on the web are also cascaded. Constitutive rules for OWL rely on constitutive rules for RDF which in turn relies on institutional facts about URIs. The social realities created by these technologies will be used to build applications such as the BrazilNuts e-commerce site. b. An institutional fact generally reflects a collective intention to treat X as Y, i.e., to assign a social function to X. This collective intention confers Y-status and Y-power and Y-function, i.e., a Y-role, on X. So, for example, becoming married accords the participants the social status of spouses (including, e.g., certain legal rights). Power, rights, functions, etc. all accord to the roles created by web institutional facts as well. By clicking the join button on a social networking site, I acquire the rights and privileges of a member of that site. c. Institutional facts such as these are often established by means of an explicit performance of an action, such as a marriage ceremony or formal licensure. They are also often marked by a status indicator such as a wedding band or professional license. Confirmation buttons, reconfirmation emails, etc., are all actions that are explicitly required in the establishment of institutional facts. These take on legal meaning as well as pragmatic utility. Performance becomes a marker of status change. "Print this page" confirmations are status indicators much as wedding bands are; so are "Welcome, Chris" messages at the entry page to your favorite web sites. d. Institutional facts persist only so long as there is collective intention to sustain them. Once the community decides not to honor certain currency or licensure, the prior institutional facts cease to hold. Like their non-web counterparts, institutional facts on the web persist only so long as there is collective intention to sustain them. Abandoned standards, deprecated formats, and outdated URLs all mark cases where institutional facts have been abandoned by collective intentionality. = Social Meaning on the Web == Searle does not explicitly discuss the web, but his theory maps beautifully onto it. According to Searle's theory of social reality, certain actions on the web count as commitments, making transactions possible. We assign web statements functions because we collectively intend for them to be used/treated that way, and we set up constitutive rules so that a sequence of web code plus (human) action constitutes an actual meaningful real world action (albeit likely an institutional-fact kind of action) like a purchase or an offer to sell. We create speech acts ("confirm?") and status indicators (MyShoppingCart MyOrders). Once the standards - the constitutive rules - become sufficiently clear and widespread - collectively embraced - these institutional facts are backed by the full weight and force of the legal system gets behind this. Unlike the classical semantics approach - which provides an excellent basis for constitutive rules but little means for reaching outside the web to embrace intended meaning - Searle's theory of social reality and institutional facts provides precisely an account of how things on the web can come to stand for things in the world, including abstract things such as commitments and contracts. By creating constitutive rules that establish how X comes to count as Y, the web provides an extension of social reality into an automated framework. The same constitutive rules that govern human interactions across the web can, in some cases, be automated, allowing computers to participate in our collectively intended construction of institutional facts. There have been numerous arguments among the semantic web community concerning "social meaning". While this term is rarely well defined, it generally calls out something about meaning that is omitted from the classical semantics view and that seems to find its origin in communities of practice. Recasting web meaning in terms of social reality allows us to begin to unpack this discussion. Importantly in a system as distributed as the web, collective intentionality needn't be universal. So if the FOAF folks say that apple means computer and the Beetles say it means music, they can each construct their own social reality, possibly using the same web stuff, and be perfectly happy about it so long as they don't try to share their constitutive rules. When they do, lawyers get involved. A group of people who share an ontology can use that ontology to build inter-compatible theories. According to the model theoretic or axiomatic approaches to semantics, these theories are consistent and can share models (or consequence sets). Now consider a person who operates at the edge of this community but also in another (adjacent) community. This person may adopt some of the first community's terminology, but may not buy fully into the complete ontology and theories of the first community. In other words, the peripheral person may adopt only part of the terminology and logical infrastructure of the community and may further adopt terminology and logical infrastructure of his/her own or even of a second community's. The world according to this peripheral person may be completely consistent, but that world may not be consistent with the omitted parts of the original community.[2] Meaningful interaction essentially requires this ability to adopt some but not all of the infrastructure of a community. This means that not all models (in the classical sense) for the original community are models for the peripheral person and vice versa. Similarly, not all models for the peripheral person need be models for the second community. We can continue this way, chaining communities via persons who overlap them, until we get to the point where we may have actual inconsistency between the original community and the final community in the chain. This means that no model-theoretic or axiomatic semantics can ever hope to provide meaning for the full set of communities (though it can possibly do so for individual communities). Mini-example: I believe that it is British usage that to table something (in a discussion) means to open it for discussion. I know that American usage (at least according to Robert's Rules of Order) has tabling an issue meaning essentially removing it from discussion. It wouldn't do for someone used to one usage to be completely unable to engage in discussion with someone used to the other. Enter social reality. Meaning on the web emerges from our taking web statements to count as assertions, commitments, etc. There is no requirement for global consistency. Instead, towards the periphery of a community an institutional fact may carry less weight, require more affirmation, be less credible on its own. (Paper money is of less value in places where the endorsing government has less credibility. Status markers may need documentary backup or demonstration of proficiency to be fully accepted in distant jurisdictions.) Because social reality finds meaning in social practice, gradual transitions from one set of meanings to another are not merely possible but everyday occurrences. [1] Searle explicitly distinguishes the linguistic description of the brute fact about Mount Everest from the brute fact itself. I am less sanguine that this is an entirely reasonable move, but the question is really outside the scope of this paper. [2] If complete consistency is required, then there is no possibility of ever merging any aspects of two independently developed communities unless they happened to luck into being completely consistent and compatible.)
Received on Wednesday, 13 June 2007 14:14:36 UTC