I think it is reasonable to have Agent as an entity. It is logical to include that superclass in the model. That does not mean it has to be visibly present in a specific implementation of the model. But the superclass does provide a useful handle for categorizing and generalizing all entities which can create Works. A specific implementation might not have to mention Agents, depending on how it is written.
My main point was that we may not want to so precisely define what can or cannot be an Agent, or a Person. I can understand Gordon's argument that many (I would not say all) interested parties outside the bibliographic universe would agree that Works can only be created by humans. But I do not think that will always be true, and there are some who insist it is not even true today. Our understanding of creativity is changing; developments in computer science, cybernetics, genetics, and cognitive science are opening new possibilities in machines and animals; and the very definition of human may be questioned in the future. I think that the scope note for LRM-E6 and the definition for LRM-E7 are unnecessarily restrictive by limiting to human beings. As I indicated before, that definition probably could work for the next couple decades, but may become problematic over a longer term. We might want to be a bit more flexible to accommodate potential changes in what the outside universe accepts as creativity and personhood. But I'm not adamant on this point-I think we do have time. By the time such changes happen in the world, we may be developing more models.
This is separate from any argument about fictional or non-living characters attributed as creators. I am somewhat fond of the idea of fictional creators, but I would be quite satisfied if it is possible to note attribution to fictional characters. I've decided that not claiming fictional characters as real creators is probably better for exchange outside the bibliographic universe, but we must be able to represent the attribution because that is what we find inside the bibliographic universe. FRBR-LRM does not directly model such attribution, but it is possible to implement attribution within the model. It would work, but not as well as I would like. I'm still trying to think of changes to FRBR-LRM that would make that smoother.
I hope that clarifies my current (and fluid :) ) position. The discussion has been helpful in directing my own thinking on the matter, and hopefully other people too.
Steve McDonald
[log in to unmask]<mailto:[log in to unmask]>
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of Ted P Gemberling
Sent: Wednesday, April 06, 2016 6:14 PM
To: [log in to unmask]
Subject: Re: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
Stephen wrote: "So, one argument would be that librarians should not be the ones (the Agents?) who define who or what can or cannot produce Works."
So is that an argument that we need Agent as an entity? Are you saying we need Agent because Person might change definition but Agent won't?
As catalogers we try to figure out who the primary author of something is. One of our traditions seems to be to enter a work under title if we can't do that. I wonder if it will ever be necessary to determine the primary agent (author) of all works.
Thanks, Ted Gemberling
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of McDonald, Stephen
Sent: Wednesday, April 06, 2016 3:16 PM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
How about a different tack on the argument then. I don't want to bring the discussion further into the realm of science fiction, but the future is all but certain to someday produce something which challenges old definitions of "person". To avoid further digression from the point, I won't speculate on the possibilities. So, one argument would be that librarians should not be the ones (the Agents?) who define who or what can or cannot produce Works.
I would say that the current definition could probably work for a few decades. But I would be surprised if there isn't _something_ that challenges the definition of "person" in some way within fifty years. Is there truly a need to restrict our definition of Agent so absolutely?
Steve McDonald
[log in to unmask]<mailto:[log in to unmask]>
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of Ted P Gemberling
Sent: Wednesday, April 06, 2016 3:48 PM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
Benjamin,
I agree with FRBR-LRM that computers are just tools, not agents. A machine is not a machine unless it does what you design it to do. Now, it could be that we will design machines badly or delegate things to them that we shouldn't delegate, and that will put us in danger. Maybe that would make them agents, if we really lost control of them. But I still don't think they would be "persons."
I would not use the possible agency of computers as an argument for an Agent "entity."
Just my two cents.
Ted Gemberling
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of Benjamin A Abrahamse
Sent: Wednesday, April 06, 2016 2:30 PM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
Another thought regarding authorship under FRBR-LRM: there is a professor here at MIT who studies and creates computer-generated literature. He has published at least one book that he didn't so much write as code . He used a program to generate literary tropes and string them together into sentences and even chapters (see OCLC #881293033). When the book was "translated" into Polish, the code itself (or some portion thereof, I don't know the exact details) was translated, and a new book was generated (see OCLC #889766423).
Right now this is just a corner case that raises a number of interesting questions about the FRBR model. The one I'm focusing on is the question of whether the professor is indeed the "author" of the work. (Indeed, in this case we have treated him as the author and he is represented as such typographically on the title page.) And the fact that a computer wrote the text itself is pretty obvious once you read it.
But it's not a great leap of the imagination to think that in the future there may in fact be books "authored" by non-human agents, i.e., artificial intelligences. And they almost certainly will be much more sophisticated than the example I just mentioned--bordering on or even passing a literary equivalent of a Turing test. And it's only slightly science-fictional to conceive of these entities having little or no relationship to actual humans. They might live on the Web, for example, as disembodied neural networks churning out endless Harlequins on-demand. I would think in these cases considering the author of the program itself would not be all that satisfactory (particularly when "authorship" of code is itself a knotty cataloging problem.)
Microsoft recently experimented with creating an AI-twitter. Unfortunately because its entire intellectual horizon consisted of what people post on Twitter it had little to discuss besides smoking pot and voting for Donald Trump, and was taken down. But it's early days and I'm sure they learned a lot from the experiment, and we'll see something of the sort again.
So not fictional, but still not human or "alive" in the normal sense. If we've hard-coded the notion of "being a living human" as necessary to be considered by a cataloger as a real author, how will we deal with this kind of literature?
Benjamin Abrahamse
Cataloging Coordinator
Acquisitions and Discovery Enhancement
MIT Libraries
617-253-7137
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of Ted P Gemberling
Sent: Wednesday, April 06, 2016 2:30 PM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
Kevin,
I agree that agents shouldn't have to be human. It is true, though, that things created by non-human entities are usually mediated by human persons: say, if a chimpanzee does a painting, someone like a scientist will tell us about it. But that's true of human persons, also. The primary artist is sometimes not entirely independent. I think there are autistic persons whose artwork has been published, who couldn't have done that on their own.
Philosophically, I think a person is any organism whose character develops. Peirce called personality "developmental teleology": persons have purposes that develop over time. But that might be getting more philosophical than we need to be for library work. Maybe a person can be any individual we give a name to. As Bob said, agency is not necessarily inherent in that. An infant or a vegetative adult is not an agent but is arguably still a person.
Ted Gemberling
UAB Lister Hill Library
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of Kevin M Randall
Sent: Tuesday, April 05, 2016 8:39 PM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
What I've been seeing, besides excessive simplification, is just a general lack of arguments supporting the restriction in the Agent class. Why must an agent be a human person or a collection of human persons? What purpose is served by this restriction? If adding more attributes pertaining to reality and species-and qualifications to relationships to distinguish them as actual or purported-is not satisfactory, why?
Kevin M. Randall
Principal Serials Cataloger
Northwestern University Libraries
Northwestern University
www.library.northwestern.edu<http://www.library.northwestern.edu>
[log in to unmask]<mailto:[log in to unmask]>
847.491.2939
Proudly wearing the sensible shoes since 1978!
From: Program for Cooperative Cataloging [mailto:[log in to unmask]] On Behalf Of Ted P Gemberling
Sent: Tuesday, April 05, 2016 5:39 PM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: [PCCLIST] Excessive simplification / was: FRBR-LRM: "agent" as an entity
Here are some belated comments on these issues. Sorry to be so late. I realize this may not be the best forum for posting comments at this late date.
I agree with Bob that there can be excessive simplification in conceptual models, and I think FRBR-LRM has several.
In the message below, Bob says "I'm not sure of the utility of introducing the idea of 'capable of' into an entity definition." Is an entity an "agent" before he/she/it is an agent? I don't think so. Later in the discussions, a commenter pointed out that FRBR-LRM says: "An arbitrary sequence of signs is not a nomen until it is assigned to be an appellation for something in some context" (p. 21). Why is the case different with "agents"? I question that anything is really gained by creating this super class besides the perceived elegance of the model.
I would go further and question res as an entity. FRSAD achieved some real insight by simplifying subjects into one entity Thema, and apparently the drafters of FRBR-LRM have decided to go one step further: since most authorized access points can be subjects, why not say that the superclass of all bibliographically relevant entities is the same as subjects? Thema then becomes superfluous. But it's not exactly true that all AAP's can be subjects. We know that at least in LC cataloging, Ceylon can't be a subject heading for Sri Lanka. I suppose one could argue that is only a particular "implementation" and that Ceylon could be a good subject heading in principle. But something is only a subject when it is a subject. I would retain Thema. It is a way to express the insight that most things (res) can be subjects but are not inherently subjects.
Heidrun and Thomas discussed the rather unclear statement on p. 49: "In general, the appellation relationship would be many-to-many; however, in the context of a particular library system, the intention is that each nomen is used in an unambiguous sense by being associated with a single res." I am guessing that they were only trying to say that libraries want to use controlled vocabularies so they can collocate resources by the same authors or on the same subjects. But the statement was overbroad: they probably didn't mean to include things like 245 or 246 titles, which are nomens in the broad sense. Nomens but not controlled nomens.
A big part of what the authors were talking about is the arbitrariness of language. Something can be a nomen for something if we agree it is. It's not inherently a nomen. I expounded on this in my recently published C&CQ article, "FRSAD, Semiotics, and FRBR-LRM" (v. 54, no. 2, 2016). One example I used there for "non-inherentness" if not arbitrariness is the mountain on Mars that looks something like a human face. Unless you subscribe to the view that aliens built the structure to send a message to us, the mountain is not inherently a sign for the human face, but it's capable of being one once we recognize the similarity. Nomens, or what C.S. Peirce called "symbols," take this a step further by being arbitrary: they don't have to look like the thing they represent or be in close proximity to it. Robert Galbraith can be a nomen for a writer of fiction but isn't until we make it one.
Another point I brought up in my article is that not all URI's are nomens. At least URL's, I think, are not. Rather than referring to something, they take you to it.
Just my thoughts. Feel free to post this on other lists if you want, and please let me how I can continue the discussion there.
Ted Gemberling
UAB Lister Hill Library
