Notes on "Information and Architecture" by John Perry and David Israel
David Israel and John Perry build upon the idea of pure and incremental
content (developed in previous work) to introduce their notion of an
information system of signal structures, describing three general kinds
of information architectures: coincident architectures, combinative
architectures, and flow architectures.
Incremental information content is the content indicated by a signal or indicating fact given a connecting fact and relative to a constraint. For example, the fact that my tongue was severely burned(the indicating fact or signal) indicates that the coffee was very hot (the indicated incremental content). The carrier of the signal is the tongue, and the connecting fact is that I sipped the coffee. The constraint is that if someone sips very hot coffee (e.g. ≥ 66° C) they will burn their tongue. The pure informational content might be characterized as the person who sipped the hot coffee was burned on the tongue.
The authors also introduce the notion of reflexive information which they contrast with incremental information content. Reflexive information is characterized by having the carrier of the signal occur in the indicated content of an informational statement. Reflexive information need not be mediated by a connecting fact because the carrier of the signal may constitute a mode of presentation picking out the individual or subject of the information. Retaining the constraint, but putting aside the connecting fact, we can transform the previous example into an example of reflexive information: the fact that x's tongue was burned indicates that the coffee that came into contact with x's tongue was very hot.
Information systems are systems of signal structures. A signal structure is composed of a signal, the carrier, and any relevant constraints, connecting facts, and content. Israel and Perry consider how the architectural relations between signal carriers influence content structure by considering three types of architectures:
In a coincident architecture, carriers of information signals stand in architectural relation that induces a relation between their respective indicated contents. Such architectures rely on connecting fact like that found in simple incremental content, but the connecting fact plays a role similar to that of the carrier in reflexive information: in particular the carrier of one signal picks out the individual the information about which depends on another signal.This may be best understood using an example. Consider a portrait photograph of the kind used in travel documents as it relates to hair and eye color. It is in the nature of a (un-doctored) photograph (the signal carrier) that when one uses it to determine hair color and eye color, that these determinations of hair color share the relation of being descriptions of the same person. We might say, for example, that the fact that the color of the eyes in the photograph is brown carries the information that the person in the photograph whose hair is black has brown eyes. The relationship of identity between the person with brown eyes and black hair is induced by the photographic process and the general circumstances in which such photographs are produced.
Another example, the one the authors develop in their article is that of an apparatus which simultaneously measures the height and weight of the individual who steps on it. The apparatus is constructed in such a way that any measurement event produces two measurements of the same individual. More generally, the relation need not be one of identity, however. The characterizing aspect of coincident architectures is that the relationship between the information contents (the color of the eyes and hair, the weight and height of the person) is induced or determined by the architectural relation between the carriers of the signals.
A combinative architecture does not induce some relationship between informational contents. Instead, the architecture between signal carriers reflects or is determined by some relationship already existing between those contents. In some contexts one might say that a combinative architecture is designed to make some relationship between contents apparent when it otherwise would not be. The authors illustrate the idea in the following way. A patient of a physician may have many different documents. In order to make the relationship between these documents (namely the fact that they are all documents about the same patient), all these documents are kept in a single folder marked with the name of the patient: the patient's file. The folder does not induce the relationship between contents- it already exists. Instead, the folder acts as a signal to make that relationship apparent.
The authors distinguish combinative architectures from cases of what they term convergence in which a relationship may be discerned by an examination of the indicating properties of the signal or indicating fact. In contrast, in combinative architectures such relationships are discernible as a relation between signal carriers. We will illustrate this by two examples.
1. Suppose that the patients papers were taken out of the folder and scattered, and the folder lost. By careful examination of the documents scattered around the office, the physician and her staff is able to reconstruct the patient's file.
2. Consider an investigator who would like to determine whether an individual in one photograph is the same individual as that found in a second photograph. By careful examination, the investigator is able to confirm this by showing that a small but distinctive tattoo is found on the individual's left shoulder in both photographs.
Note that these determinations are made by an examination of the internal indicating properties of the signals, not by some architectural relationship between the carriers. This is a case of convergence, as Israel and Perry conceive it.
Having discovered that the photographs are indeed of the same person, the investigator labels both photos "Suspect #1". The matching labels on the two photographs reflect the relationship that exists between the indicated contents of each. We see that the labeling does not induce the relationship between them, in the sense that one does not make them photographs of the same person simply by labeling them as such, even if everyone believed and acted as if it were the case when in fact it was not. It is not beliefs induced by combinative information systems that are at issue.
Flow architectures facilitate the movement or flow of information from one carrier to another, obeying a variant of Dretske's Xerox principle. Dretske's principle states that:
If A carries the information that B, and B carries the information that C, then A carries the information that C (Dretske, 1981, 57)[as quoted by Israel and Perry].
Israel and Perry reformulate this principle to fit more naturally into their notion of an information system:
If (i) there are architectural constraints C and architectural connection C such that s carries the architectural information that b is F relative to C given C, and (ii) there are constraints C′ and connecting facts C′ such that the fact that b is F carries the information that Q relative to C′ and C′, then there are contraints C″ and connecting facts C″ such that s carries the information that Q relative to C′ given C′. (Israel and Perry, 1991, 157).
Israel, David, and John Perry. Information and Architecture. In Proceedings of the Second Conference on Situation Theory and Its Applications, edited by Jon Barwise, Jean Mark Gawron, Gordon Plotkin, and Syun Tutiya, 147-160. Vol. 2. Stanford, CA: Center for the Study of Language (CSLI), 1992.