On Roles of Models in Information Systems - Intentional Perspectives on Information Systems Engineering - page 33

where WAlbs1, WBstone3, WNkg2 are data types of positive real numbers.
The data types may now serve as names of different combinations of class con-
cepts and the quantitative concept weight W, and may be used to define data sets,
e.g., a table with the name WEIGHT-A and attribute WAlbs2 will denote a data col-
lection with one record for every American, WEIGHT-B can be defined similarly
for the British, and so on.
Unfortunately the usual situation is that the weight definitions and their relations
to the data types are either not made as explicit formal statements, or they are soon
lost in heaps of documentation, which are usually not well maintained. The weight
definitions must be seen as part of the definitions of the data types. So the rele-
vant data sets appear as undefined entities without explicit reference to what they
carry information about. The “memory” of the definitions is imprinted only in the
software processes where the variables of these data types are used.
4.4 Information Modeling
While data collections may contain several equal data items, information collections
are sets, where all individual members are different. Information sets are defined by
relating UoD concepts and data concepts. Langefors [ 16] introduced the concept of
elementary message (e,v,t), where e is an entity in the UoD, v is a measurement
value and t is the time of the measurement. We view information to be a relation-
ship between a quantitative concept, a domain concept and a data type. A message
is a relationship between an individual domain concept and a (quantitative concept,
data type) pair, an information set is a relationship between a domain class concept
(and relation class concept) and a (quantitative concept, data type) pair. The orig-
inal definition of elementary message may need to be expanded to include spatial
information in addition to time. We will not discuss this aspect further here.
In order for a data item to carry meaning it must be related to the appropri-
ate quantitative function, to the appropriate referent in the UoD, to the scale of
measurement s in the set of scales, and to the time and location, and to the data rep-
resentation of the observed value. The precision of the unit systemmust be specified,
e.g., whether the weight is represented with an accuracy of hundreds or thousands
on the chosen scale, be it kilogram, ton or pound. When represented in a computer
program, precision is defined by the data type of the appropriate numerical variable,
e.g., integer, real, double precision, decimal number, or some other user defined data
type, as shown in the previous section on data modeling.
Information sets must be defined relative to the conceptual model of the world
that is represented by the data items. For example, an information set that contains
the numerical values of the weights of all persons could be specified as a data col-
lection containing one element for each member of the extension of the appropriate
world concept. In order to qualify for an information set, each of the elements in
the set must be a pair consisting of a unique label for each individual of the domain
class concept and a value for the corresponding weight.
We illustrate this by continuing the example of weights of persons and
Norwegians. For Norwegians the situation is straight forward. All Norwegians