Foreword 1. EXECUTIVE SUMMARY

1. Executive Summary

Information society – knowledge society 

In the eyes of many, one of the most challenging problems of the information society is the fact that we are faced with an expanding mass of information that both increases and changes as we move along. Selection of the relevant bits of information seems to become more important than the retrieval of all these data: the information is all out there, but what it means and how we should act on it may be one of the big questions of the 21st century. If an information society is a society with an exponential proliferation of data, a knowledge society must be the one that has learned how to cope with this.

Profiles as automated knowledge construction 

One of the attempts to deal with the information explosion is the automated construction of ‘knowledge’ by the data mining of large data bases. With the use of mathematical, or rather statistical techniques, it becomes possible to search massive quantities of data for patterns of correlations that produce a new type of knowledge. This knowledge, consisting of linear or nonlinear correlations between data, is what profiles are all about. Rather than just a collection of unrelated data, a profile is a set of correlated data. Whereas data is information, profiles – when interpreted – are knowledge constructs. 

Interdisciplinar perspective of the main text 

In this FIDIS deliverable 7.2 researchers from 11 FIDIS partners have put together their expertise and experience in the field of profiling. The aim of this deliverable is first of all to provide a comprehensive descriptive analysis of the process of profiling. It hopes to have created some common ground between the social, technological, mathematical and computer science perspectives that inform the process of profiling. As a consequence, apart from chapter 4, this deliverable does not yet elaborate extensively on issues of privacy, security and equality – even if present around every corner. Also the legal implications of profiling are not dealt with at this point. Deliverable 7.2 studies the techniques, technologies and practices of profiling, including its purposes and effects. It should function as a building block for subsequent deliverables that elaborate the legal aspect and the relationship with Ambient Intelligence (FIDIS deliverable 7.3); the implications for Europe as a constitutional democracy (FIDIS deliverable 7.4) and the relationship between RFID, AmI and profiling technologies (FIDIS deliverable 7.7). 

In chapter 2 the concept of profiling is analysed by pointing out important distinctions, of which the difference between group profiling and personalised profiling is the most salient one (others include automated and hand made profiles, profiling of on-line and off-line behaviour, construction and application of profiles). Some working definitions are provided that incorporate these distinctions and point to the difference between a collection of data and a set of correlated data that describe a particular data subject (whether a person, a thing or an event). The purpose of profiling practices should be taken into account, as this determines both the adequacy of the construction of profiles and their impact on our world. The purpose of profiling is taken to be an assessment of risks and opportunities of the data subject, enabling private and public service providers to tune their services to the targeted group or person and enabling public authorities to anticipate threats to security, public health or any other relevant public good.

Chapter 3, the central chapter of this report, describes both group profiling and personalised profiling in some detail, to get a good grip on the difference between the two and a clear idea of the techniques involved. For group profiling the process of KDD (knowledge discovery in data bases) is analysed. KDD can be described in 6 steps: (1) collection and recording of data; (2) aggregating or preparing of the data in databases; (3) data mining or running algorithms through the database; (4) interpretation of the emerging patterns; (5) decision taking on the basis of the resulting profiles, and (6) proliferation of these profiles in the relevant social contexts. Special attention is given to data mining techniques, to get some interdisciplinary awareness of the nature of this type of knowledge. At the same time the importance is stressed of practical knowledge during the whole process of profiling: a combination of relevant experience, healthy scepticism and trained intuition is needed to collect the relevant data; to store them in a way that allows retrieval of interesting correlations; to find the algorithms that will locate interesting patterns; to have an adequate understanding of the possible meaning of the emerging patterns and to choose the optimum action based on this knowledge, considering also the wider implications in the real world. Group profiling produces group profiles that can be understood as a kind of prototype or abstract person. A group profile does not represent any particular person, but rather categorises persons as belonging to a certain group of category. Personalised profiling, on the other hand, focuses on the attributes of one individual, identifying and representing her on the basis of biometric features or past behaviour. The section on personalised profiling discusses user modelling in relation to adaptive applications and biometric profiling of individual data subjects. Personalised profiling can identify preferences and risks the profiled person may not be aware of, which gives the data controller an advantage in terms of knowledge over the data subject. One of the conclusions of this chapter is that the type of knowledge produced by profiling technologies is in several ways different from the findings of traditional social science research.

Chapter 4 deals with some of the wider implications of this new type of knowledge. After a brief exploration of the role of profiling in the emergence of data surveillance (dataveillance), the implications of personalised profiling are analysed, after which a more abstract mapping is constructed of the anticipated consequences of profiling. This is done in terms of purposes and effects.

Chapter 5 explores several fields of application, ranging from marketing, employment, financial institutions and e-learning to forensics. The different contributions aim to illustrate the state of the art in different fields, without pretending to provide a complete inventory. The idea is that these examples give the reader a sense of the actual functioning of profiling technologies in different contexts.

Chapter 6 locates issues for further clarification that should be dealt with in subsequent deliverables, such as the commodification of (personal) data and issues of privacy, security, trust, usability and equality.  

Multidisciplinary Appendix 

The multidisciplinary nature of the FIDIS consortium has as a consequence that the partners do not always speak each others language. The objective of this deliverable has been to focus on an interdisciplinary main text, that creates common ground for all those involved in the project. In the Appendix those involved in the different disciplines had the opportunity to address a more specialist audience, thus also allowing a more profound analysis of central elements of the analysis. This concerns forensic profiling in the field of RFID and biometrics; legal grounds for customer loyalty programs in Germany; physical and behavioural biometric profiling; profiling of web users; mathematical tools for datamining; the nature and use of algorithms; user modelling in virtual communities and a description of the profiling game, played by PhD-students at the first PhD training event at Riezlern.