tripleC 7(2): 202-207, 2009
ISSN 1726-670X
http://www.triple-c.at
CC: Creative Commons License, 2009.
Analysis of Semantic Information via Information Reports
Julio Ostalé
Universidad de Salamanca (PhD Programme), Spain. E-mail: ostale@yahoo.es
Abstract: Semantic information is analysed by means of two consecutive approaches. Firstly, we consider semantic
information via ordinary-language reports of the form “X carries the information that Y”. Secondly, and partially based on the
previous analysis, we obtain a number of a priori conditions a physical system has to fulfill for semantic information to take
place in such a system.
Keywords: semantic information, information reports, a priori conditions, situation theory, channel theory, semantics,
objectivity problem
Acknowledgement: The author carries out his research within the project “Las lógicas de la red” (ref. HUM2006-12848-
C02-01), supported by the Spanish “Ministerio de Educación y Ciencia”. He is also indebted to Francisco Salto Alemany
and José María Díaz Nafría for encouraging him to take part into the “First International Meeting of Experts in Information
Thepries” held in León (Spain), during November 6
th
– 7
th
2008.
hat is semantic information? This is
after all the topic we are going to deal
with. Notice that, even before we propose
a definition of semantic information, our
choice of the very term “semantic information”
apparently involves two things: (i) because of
the English grammar, our choice should imply
that there is something we recognize as
information, of which semantic information is
but a variety; (ii) because of the well
established semiotic distinction syntax-
semantics-pragmatics, it should also imply
that there are two further varieties of
information, namely syntactic information and
pragmatic information. Now, I will neither
accept nor deny theses (i) and (ii). The reason
I have put the expression “semantic
information” in the title of this paper is that I
wish to focus on certain topic, indeed a very
old one, that in the last sixty years has been
discussed under the umbrella of that
expression. “Semantic information” is used as
a label for those inquiries that try to merge
ideas from Shannon's Statistical Theory of
Signal Transmission with some theory of
meaning.
1. Information Reports
A neutral formulation of our topic is given in
the major work of Channel Theory: “How is it
possible for one thing to carry information
about another?” (Barwise and Seligman,
1997: xi). Channel Theory, whose
philosophical background is consistent with
the ideas of this paper, is one of the most
recent descendants of Situation Theory, a
mathematical logic-oriented theory of
information first developed in Barwise and
Perry (1983). The general framework of
Situation Theory, including terminology and
basic hypotheses, was fixed by Devlin (1991).
Actually, the picture is much more
complicated once we observe some facts:
there are different versions of Situation
Theory in the literature; the standard Channel
Theory as developed in Barwise and
Seligman (1997) stems from an earlier version
that was also called Channel Theory, see
Moss and Seligman (1994); finally,
relationships between different versions of
Situation Theory and different versions of
Channel Theory are not completely clear. Our
considerations in this paper will run in parallel
tripleC 7(2): 202-207, 2009 203
CC: Creative Commons License, 2009.
with those ideas that are commonplaces to
every version of Situation Theory and every
version of Channel Theory.
What does it mean that some things carry
information about each another? We know
that fire causes smoke; this (causal) regularity
or constraint supports in some way the fact
that every piece of smoke carries the
information that there is a fire; however, and
this is the controversial point, the semantic
relation denoted by “carries the information
that” is usually thought of as different from the
nomic relation denoted by “causes”. In other
words, we figure out and speak of the world
as if there were information flow out there;
furthermore, this information flow is taken as
something that rests on regularities, although
it is not a regularity by itself. The main
problem is how to achieve a good explanation
for information on the basis of regularities.
We are going to work out an indirect notion
of information guided by the use of language.
In which sense? The nature of information is
not given by English language, of course, but
we can substitute the abstract notion of
information by a more concrete notion with the
help this language. The key is to identify
typical reports describing what we are
interested in, and then to define our topic as
the reference of those reports. Let us start. By
an “information transfer” I suggest to
understand hereafer anything whose typical
linguistic expression is
X carries the information that Y (1a)
or some other variant close enough to this
one. We will say in the sequel that X is the
“source” and Y is the “target” of the transfer.
At the linguistic level, statements like (1a) are
called “information reports” after Israel and
Perry (1990). Typical examples of information
transfers are: dark clouds in the sky bear the
information that it is going to rain nearby, the
open door of my house carries the information
that someone is in there, your smiling brings
the information that you are happy, a jacket in
my brother's wardrove means that my brother
owns the jacket, and so on. Of course, there
are also more complex information
phenomena; to put an example, the word
“open” on the door of a shop indicates that we
can enter the shop, though this permission is
not something we can easily recognize as a
state of affairs. For convenience, we consider
in this paper only those information transfers
in which both X and Y are state of affairs we
can easily describe in terms of “item s in of
the type T”. In the previous example, it is far
from clear whether the item is a concrete
person, any person, all the persons, the
situation itself, the relation of a person with
respect to the shop, etc.
In ordinary language, while keeping the
term “information” playing the role of a direct
object, we can distinguish two groups of
expressions reductible to (1a). Firstly, there
are those expressions whose main verb is
static, in the sense that doesn't refer to
change, and indicates possession: X conveys
the information that Y, X supports the
information that Y, X has the information that
Y, X gives you the information that Y, etc.
Secondly, there are those expressions whose
main verb is dynamic and indicates
transportation: X bears the information that Y,
X brings the information that Y, X transports
the information that Y, etc. This distinction is
admittedly not very sharp in everyday
conversations; nevertheless, it will become
quite useful for philosophical analysis.
Further, there are expressions akin to (1a)
without the term “information”: X informs that
Y, X reports that Y, X means that Y, X
involves Y, or X implies Y.
There is the habit of assuming X and Y to
be of different nature. It is frequently accepted
that X stands for something physical, the so-
called “signal”, whereas Y stands for
something either mental or abstract, the so-
called “information content of the signal”. For
some reason, schema (1a) is assimilated to
the schema of propositional attitudes: A
believes that p, A sees (that) p, A truthfully
asserts that p, A remembers that p, A wonder
wether p, and so on. This is the case for
instance in Israel and Perry (1990), a brief
article whose influence can be traced until
important monographs like Pérez (2000) or
Floridi (2004). But is this the only way for an
understanding of the above schema? We
would like to show that the similarity between
(1a) and propositional attitudes is misleading,
hence a potential source of pseudo-problems.
By saying or writing first X, then a verb, and
then Y preceded by the word “information”, it
204 Julio Ostalé
CC: Creative Commons License, 2009.
seems to many authors that Y is qualified in a
deep sense that X is not. To complicate
matters, this linguistic distinction between a
subject whose form is “X” and a direct object
whose form is “the information that Y” meets
to quickly a far from clear intuition involving
the double nature of information, namely the
intuition by which information involves
physical ingredients (signals, documents, TV
screens, pictures...) as well as ingredients that
are not physical (meanings, contents,
propositions...). In abstract thinking, the
encounter of two dichotomies (A versus B)
and (M versus N) too frequently yields to a
new dichotomy (A or M) versus (B or N). In
our case, the above linguistic distinction
together with the above intuition behave like
that. By means of a one-to-one implicit
attribution, many authors come to think that X
stands for something physical whereas Y
stands for something non-physical. They
interpret (1a) by saying that the signal X
conveys the informational content Y. Yet this
assimilation is too rude, even if we accept the
token-type dimension of every information
transfer. If you rephrase (1a) by the
synonymous
X informs that Y (1b)
then you begin to have reasons for being
suspicious about the analogy between (1a)
and propositional attitudes. With (1b) it is not
so clear that Y has to be a proposition.
Actually, the original meaning of the verb, by
which informing boils down to imposing a form
over an object (see Capurro, 1978), suggests
a relation between two entities of the same
sort. And more importantly, (1a) and (1b) can
sometimes be reversed into.
Y carries the information that X (1c)
This transformation does not always make
sense. An example for the possibility of such
a transformation: snow on the roof of your
house bears the information that it is snowing,
and the fact that it is snowing informs that
there must be snow on the roof of your house.
A counterexample: the door of my house
being open informs that someone is in there,
but someone being at home does not usually
carry the information that the door is open.
Now, the important thing is not under which
conditions (1b) can be transformed into (1c).
The point here is that the mere possibility of
such a transformation falsifies the analogy
between (1a) and propositional attitudes,
since there is no way of reversing for instance
A believes that p into p believes that A.
This is plenty of consequences. Consider
e.g. that an important discussion when
interpreting Y in (1b) as a proposition
concerns veridicality. Some authors defend
that the information content of a signal is a
true proposition, as in the case of p in A
knows that p. Others maintain that the
information content of a signal is a proposition
that can be either true or false, as in the case
of p in A believes that p. However, both thesis
share that the information content of a signal
is a proposition. If we interpret Y as something
distinct of a proposition, the debate on
veridicality calls for a new perspective. Let us
consider an argument by analogy. I consider
that (1b) is closer to (2b) than to (2a) in these
examples:
John is 90 kg weight.
(2a)
Body a is attracted by the
gravitational field of body b.
(2b)
In (2a) we have a sentence whose subject
refers to a concrete entity, John, whereas the
object refers to an abstract entity, 90 kg
weight. In (2b) the difference between how the
subject and the object are expressed, putting
“gravitational field” only before the object,
does not constitute a big obstacle to our
understanding of gravitation as a kind of
relation between bodies. It is obvious that we
interpret (2b) as reporting a relation between
two bodies a and b, not as reporting a relation
between the body a and the gravitational field
of b. Likewise, I propose to understand the
subject and object of (1a) as entities of a
same species. Not necessarily spatio-
temporal entities, though.
To refuse that Y is a proposition, the so-
called “information content”, encoded by some
entity or event X, the so-called “signal”, is just
an argument for the plausibility of this claim: X
and Y are of the same nature. Well, but what
exactly are X and Y?
tripleC 7(2): 202-207, 2009 205
CC: Creative Commons License, 2009.
2. A Priori Conditions
It is time to ask for the internal structure of
both the source and the target of an
information transfer. And the first thing worth
saying is that, following Grice (1957) and
Situation Theory, entities carry information
about each other as long as they are in an
abstract state. It is not the bare cloud that
carries information whatsoever, but rather the
cloud-being-dark (as opposed to the cloud-
being-white) that carries the information of
rain. Just think of it by reductio ad absurdum:
something that cannot be but in one abstract
state remains indifferent to environmental
changes, hence its being in that state can
never be a source of information. This
interplay between concrete entities and
abstract states or types is of the biggest
importance in understanding the structure of
information transfers.
The distinction of tokens (concrete bearers
of information) and types (abstract states in
which tokens can be) is dependent of some
classification schema. In Channel Theory we
write {s
1
, s
2
...} and {T
1
, T
2
...}, respectively, for
tokens and types of certain classification
schema. Then we can write s : T to denote
that token s is of type T. Accordingly, two
classifications are two structures of the form
A = ( {a
1
, a
2
...}, {A
1
, A
2
...}, :
A
)
B = ( {b
1
, b
2
...}, {B
1
, B
2
...}, :
B
)
We omit superscripts in :
A
and :
B
, and we
take arbitrary a from {a
1
, a
2
...}, A from {A
1
,
A
2
...}, b from {b
1
, b
2
...}, and B from {B
1
, B
2
...}.
It makes now sense to say that it is a specific
token a as being of type A that can carry
information about another token b as being of
type B. In symbols, a : A carries information
about b : B. Actually, we should better say
that it is a specific token a as being of type A
that can carry the information that another
token b is of type B. Again in symbols, a : A
carries the information that b : B. The second
manner is better in that it avoids the notion of
information as something different from the
ordered pair b : B. Hence, schema (1b)
amounts to:
a : A informs that b : B
(1d)
Ordered pairs like a : B can be called
“information units”. From a different
perspective, Devlin (1991) obtains a schema
almost identical to (1d) for which he applies a
terminology that has become standard in
Situation Theory: a supports A, b supports B,
and a carries the information that B. (We can
accept to some extent this terminology,
although we also have to be conscious about
the fact that Situation Theory, as opposed to
Channel Theory, does not consider relations
between tokens; that's why the relation of
carrying information do not take into account
token b.) By the way, notice that Devlin's
distinction between two senses in which a
token relates to a type matches pretty well our
previous distinction between static and
dynamic verbs for information reports; to
support and to carry relate side by side to
static and dynamic verbs in information
reports.
By paying attention to the previous
analysis, we obtain two conditions a complex
system has to fulfill for information transfer to
take place between two of is parts (modelled
by A and B).
(Condition 1) For one token a to be the
source of some information transfer, it is
required that a can be classified by more than
one type. To be precise, in order to inform
about n different things it has to admit at least
n different types A
1
, ... A
n
.
(Condition 2) Given that a is the source of
an information transfer, for b to be the target
of the very same transfer, it is required that b
can be classified by more than one type. To
be precise, in order b to convey as much
information as a it has to admit at least n
different types B
1
, ... B
n
.
Lyre (2002) and Szaniawski (1984) have
similar observations on the a priori
foundations of information flow. The key idea,
up to now, is to settle those conditions a
system has to hold for information transfers to
occur (or not) within the system. Brackets are
important here. What we have distilled are just
a priori conditions that parts of a system,
modelled by A and B, candidates for being
the source and the target of an information
transfer, have to possess. But the potentiallity
206 Julio Ostalé
CC: Creative Commons License, 2009.
of information will not be actuallized unless
something links A to B.
So we have to put in a broader context the
source and the target of any information
transfer. It is clear that information is neither
the property of a given source nor a relation
between an isolated source and an equally
isolated target. It is rather something that
occurs between two different information units
that in turn belong to two different
classifications. We get:
(Condition 3) For a token a in A to be
informative about a token b in B within a
system, it is required that at least one type of
a is regularly related to at least one type of b
by means of some information channel C
between A and B.
This condition is the most difficult one to
formalize. Indeed, last versions of Situation
Theory and the two existing versions of
Channel Theory do their best in order to
provide a formal account of regularities
between types (Situation Theory) or
classifications (Channel Theory). In either
case, we have to transform (1d) into
a : A informs that b : B with
respect to some channel C
(1e)
where the exact nature of channel C is to be
determined. It plays for information flow the
same role that a reference frame plays for the
study of movement. And we expect two
things: it should be an objetive subtitute of
notions like those of agent and knowledge,
and it should illuminate classical problems of
information flow.
Which problems are classical in the study
of information flow? Do other problems reduce
to classical problems? We have in mind two of
these classical problems. In the first place, we
have the informational phenomenon of
fallibility, by which sometimes X does not
inform that Y although it uses to. In the
second place, there is the problem of
relativity, by which X informs that Y
1
ot that Y
2
depending of the channel (to use our own
terminology) being adopted.
3. Conclusion
These pages can be seen as an attempt of
giving philosophical foundation to Channel
Theory. To be more concrete, we have
undertaken the task of analysing “the
grammar of information flow” (Barwise and
Seligman, 1997: 12-13) in order to provide
sound philosophical basis to the main
proposals of Channel Theory (Barwise and
Seligman, 1997: 35, 183)
By analyzing information reports we
concluded the symmetry (at ontological level,
though not at linguistic level) between their
respective sources and targets. That the
target of an information transfer is a
proposition has been discussed.
By searching the conditions of possibility of
information transfers, which are the intended
interpretations of information reports, we have
found that every source and every target
exhibits a token-type duality that rests upon
some classificatory schema. Furthermore,
semantic linkages between sources and
targets seem to depend on some higher-level
linkage C between classificatory schemas. As
an open problem, the phenomena of fallibility
and relativity of information should be faced
after defining C.
References
Barwise, J., Perry, J. (1983). Situations and Attitudes. Cambridge: The MIT Press / Bradford.
Barwise, J., Seligman, J. (1997). Information flow. Cambridge: Cambridge University Press.
Capurro, R. (1978). Information. Ein Beitrag zur etymologischen und ideengeschichtlichen Begründung des
Informationsbegriffs. München: Saur.
Devlin, K. (1991). Logic and Information. Cambridge: Cambridge University Press.
Floridi, L. (2004). “Information”. In L. FLORIDI (ed.). The Blackwell Guide to the Philosophy of Computing and Information,
pp. 40-61. Oxford: Blackwell.
tripleC 7(2): 202-207, 2009 207
CC: Creative Commons License, 2009.
Grice, P. (1957). Meaning. Philosophical Review, 66, 377-388.
Israel, D., Perry, J. (1990). What is information?. In P. HANSON (ed.). Information, Language and Cognition, pp. 1-19.
Vancouver: University of British Columbia Press.
Lyre, H. (2002). Informationstheorie. Eine philosopisch-naturwissenschaftliche Einführung. München: Wilhelm Fink.
Moss, L., Seligman, J. (1994). Classification domains and information links: A brief survey. In J. van Eijk, A. Visser (eds.).
Logic and Information Flow, pp. 112-124. Cambridge (MA): The MIT Press.
Pérez, M. (2000). El fenómeno de la information. Una aproximación conceptual al flujo informativo. Madrid: Trotta. [English
version: Pérez-Montoro, M. (2007). The Phenomenon of Information. Maryland: Scarecrow Press.]
Szaniawski, K. (1984). On defining information. Philosophia Naturalis, 3(4), 244-249.
About the Author
Julio Ostalé
Bachelor and Master in Philosophy (“Licenciado en Filosofía”, old-style 300 ECTS curriculum) by the University of
Salamanca. Main editor of the on-line journal Factótum (www.revistafactotum.com). In 2009 is about to finish both a MA in
Humanities by Universitat Oberta de Catalunya and a MSc in Logic by the ILLC of the Universiteit van Amsterdam. His field
of research is philosophical logic and the philosophy of information.