Combining and Analyzing Judicial Databases - Discrimination and Privacy in the Information Society - page 192

With respect to the relationship manager of the monitor, it should be clear that
the data in the monitor are closely related. The output of one organization in the
criminal law chain often serves as the possible input of another organization. For
instance, the input of cases into the prosecutorial level largely depends on the
number of suspects handed over by the police. Therefore, a plausible rule in the
manager would be: number of suspects handled by the police ≥ number of case
handled by the prosecution service; meaning that the police usually do not send
all cases to the prosecution service. Using such rules the plausibility and consis-
tency of the data is maintained. Similar rules can be formulated in order to handle
variables coming from different sources. Take, for instance, the number of com-
munity services imposed by the Public Prosecutor. This information comes from
two organizations in the chain: the prosecution service and the organization re-
sponsible for executing sentences. As a rule, the number of community services
registered by the executing organization is lower than the number of community
services registered by the prosecution service, as suspects may die or 'disappear'
before the sentence is executed. Such rules are typically based on historical data
and domain knowledge and can be extended with error values to allow for incom-
plete or tentative data (see Choenni et al., 2001).
The primary goal of the monitor is to alert users when there are large differ-
ences between input and output, or between the actual input or output and the fo-
recasted input or output. In this way, policymakers are able to indentify potential
capacity problems at an early stage. Therefore, rules are added to the relationship
manager that detect large deviations. Based on these rules, three types of alerts
are provided to the users:
1. large deviations in the proportion of organization X's output to its own input;
2. large deviations in the proportion of organization X's input to organization Y's
output;
3. large forecasting errors.
The user interface presents the user with an overview of the input and output data
and the corresponding alerts in either table or graph format, depending on the us-
er's preference. In this way a quick overview of the irregularities in the data is
provided. In these views, the user is able to zoom in on specific parts of the crimi-
nal law chain by selecting a subset of data categories. More specifically, the user
can subdivide the data into various categories including the age and gender of the
suspect, the region in which the crime was committed, and the type of crime com-
mitted by the suspect. Thus, the user can, for instance, choose to only show the
input and output of male suspects who are older than 18 years or the input and
output in a specific region. Additionally, the monitor periodically produces written
reports through a printing on demand module as described above.
In this section it was shown how data from various judicial databases may be
combined and integrated using two different approaches: a data warehouse and a
dataspace. In the first approach, data is linked explicitly on an individual level. In
the second approach, more dynamic relations or rules are established to link data
and maintain data quality. Thus, a dataspace differs from a data warehouse in the