From Data Minimization to Data Minimummization - Discrimination and Privacy in the Information Society - page 273

organisation of gathered material should be clear and metadata about the data-
base itself should be included, for example about who owns it, where it is lo-
cated, why and when it was build, when the data were included and when they
were updated.
3. Analysing data: when analysing data, the previous cluster of data and the meta-
data about the gathering of information, the database and the organisation and
categorization of the material should be preserved. Added should be metadata
about the process of analyses, the algorithms used, the databases harvested and
the methodology of mining. This may ensure that it can be assessed from hind-
sight whether patterns, profiles and rules distilled from the data are (indirectly)
discriminating or privacy violating. Finally, the context for which the data were
gathered, i.e. the purpose limitation, must be respected.
4. Using (aggregated) data: when using the patterns, profiles and rules obtained
through data mining, the metadata regarding the gathering of the data, the data-
base, the organisation and categorization of the material and the used analysing
techniques as well as the clustered set of data should be accessible. Finally, da-
ta must be gathered about in what context the patterns, profiles and rules will
be applied and used, so as to assess whether this may lead to privacy violations
or discriminatory practices. This may also help to assess whether a discrimina-
tory rule may lead to positive discrimination or is objectively justifiable.
As previously argued, the loss of context may lead to or aggravate privacy and
discrimination problems. Inherent to current data mining and profiling practices
seems a loss of contextuality, a loss which is not restored, but only aggravated by
the data minimization principle. The four data mini mum mization principles, on the
other hand, may be used and implemented to preserve the contextuality of data in
data mining and profiling practices. How this should be done is beyond the scope
of this chapter.
15.8 Conclusion
A common definition of autism is context blindness. 51 People suffering from au-
tism treat data, rules and knowledge as isolated facts, as absolute, and thereby dis-
regard the context in which they play a role. Thus, an autistic person may stop at
the middle of a zebra-crossing if the traffic light turns red. To him, 'red' signifies
'stop' and nothing else, independent of the given context, while for non-autistic
persons, a red traffic light when at the middle of a zebra-crossing signifies 'walk
faster', rather than 'stop'. Thus, a set of rules and facts beget a different meaning
in different contexts.
Data always signify a certain meaning in a specific context. If this context
changes, the information may lose its or beget another meaning. With regard to
indexical words such as 'I', 'You', 'Here', 'There', 'This', 'That', 'Now', 'To-
day', 'Yesterday' and 'Tomorrow', one needs to know where, when and by whom
a phrase was uttered to determine the meaning of the phrase. More generally, all
51 Vermeulen (2009).