Database Reference
In-Depth Information
technological measures and transparency about their design may prevent data
mining results that may easily lead to discrimination or privacy infringements.
The use of discrimination-free and privacy preserving data mining techniques
may prevent many problems, but may not be sufficient. Transparency regarding
the use of these techniques is required to create more awareness and understanding
among data subjects on how their data is used. 22 In the end, more transparency
may create more trust, provided that the data mining and profiling methods used
are not discriminating or violating privacy. Even if they are, disclosing these facts
will bring these issues to the forefront of the political discussion. Thus, political
forces might assure that the data mining processes carried out by the state or
private firms are acceptable by the broader public.
Apart from more transparency and generating greater trust, accountability is a
key element in a call for greater transparency in data mining and profiling. To
enforce proper use of personal data, it is crucial that cases of discrimination and
infringements of privacy are easily and quickly detected, even internally. For this,
technology may be useful again. 23 With proper detection systems, a rapid and
adequate response can be given to situations where discrimination or privacy
violations take place. The next section will discuss this in more detail.
19.2 Further Research
In today's society we are continuously profiled, the profiles used may be
intrinsically discriminatory and our privacy may be violated. Discovering
discrimination and privacy violations, however, is difficult, since they can be
hidden in very specific niches. We can use data mining, i.e., the use of automated
data analysis techniques to uncover previously undetected relationships among
data items, for discovering hidden discriminative contexts. Data mining does not
only come to our help, however: more and more data mining is becoming a crucial
tool when designing decision procedures. Many decision procedures are, at least
partially, being automated and in this automation, unfortunately, often little
attention is paid to anti-discrimination and privacy preservation. We will argue
that in many circumstances the use of data mining will lead to the construction of
discriminating models, which is particularly dangerous as these techniques offer a
false comfort of providing unbiased solutions based upon solid statistical
evidence, not affected by subjective human interpretation. Ethical and legal
implications regarding anti-discrimination legislation have been underestimated
and neglected for far too long by the data mining and machine learning
communities. In order to breach this gap between legislation and technology the
following directions need to be explored in future:
Using existing and newly developed data mining tools for automatically
detecting and assessing discriminative profiles used in a decision process.
22 See Chapter 3 and Chapter 17.
23 See Chapter 5.
Search WWH ::

Custom Search