|
Acknowledgements |
5 |
|
|
Contents |
6 |
|
|
Authors |
8 |
|
|
Part I: Opportunities of Data Mining and Profiling |
18 |
|
|
Data Dilemmas in the Information Society: |
Data Dilemmas in the Information Society: |
|
|
19 |
19 |
|
|
The Information Society |
19 |
|
|
What This Book Is About |
20 |
|
|
Responsible Innovation |
21 |
|
|
Data Mining and Profiling |
23 |
|
|
Data Mining: A Step in the KDD-Process |
23 |
|
|
From Data to Knowledge |
26 |
|
|
Profiles of Individuals and of Groups |
28 |
|
|
Why We Need These Tools |
29 |
|
|
Discrimination, Privacy and Other Issues |
31 |
|
|
Any News? |
31 |
|
|
Problems and Solutions |
33 |
|
|
Structure of This Book |
34 |
|
|
Part I: Opportunities of Data Mining and Profiling |
34 |
|
|
Part II: Possible Discrimination and Privacy Issues |
35 |
|
|
Part III: Practical Applications |
36 |
|
|
Part IV: Solutions in Code |
37 |
|
|
Part V: Solutions in Law, Norms and the Market |
38 |
|
|
Part VI: Concise Conclusions |
39 |
|
|
References |
40 |
|
|
What Is Data Mining and How Does It Work? |
43 |
|
|
Introduction |
43 |
|
|
Data Mining and Related Research Areas |
44 |
|
|
Database Terminology |
45 |
|
|
Basic Techniques |
47 |
|
|
Classification |
48 |
|
|
Clustering |
50 |
|
|
Pattern Mining |
52 |
|
|
Supporting Techniques |
54 |
|
|
Pre-processing Techniques |
54 |
|
|
Database Coupling |
55 |
|
|
Conclusion |
57 |
|
|
References |
58 |
|
|
Why Unbiased Computational Processes Can |
Why Unbiased Computational Processes Can |
|
|
59 |
59 |
|
|
Introduction |
59 |
|
|
Characterization of the Computational Modeling Process |
61 |
|
|
Modeling Assumptions |
62 |
|
|
Origins of Training Data |
63 |
|
|
Types of Problems |
64 |
|
|
Accuracy and Discrimination |
65 |
|
|
Scenario 1: Incorrect Labels |
66 |
|
|
Scenario 2: Sampling Bias |
67 |
|
|
Scenario 3: Incomplete Data |
68 |
|
|
Potential Solutions for Discrimination Free Computation |
69 |
|
|
Basic Techniques That Do Not Solve the Problem |
69 |
|
|
Computational Modeling for Discrimination Free Decision Making |
71 |
|
|
Conclusion and Open Problems |
71 |
|
|
References |
72 |
|
|
Part II: Possible Discrimination and |
Part II: Possible Discrimination and |
|
|
74 |
74 |
|
|
A Comparative Analysis of Anti-Discrimination |
A Comparative Analysis of Anti-Discrimination |
|
|
75 |
75 |
|
|
The $Huber$ |
The $Huber$ |
|
|
75 |
75 |
|
|
Place of the Two Rights in the EU Legal Order |
77 |
|
|
Discrimination, a Concept in Search of Unity |
78 |
|
|
Differences in the Scope of EU Data Protection and Anti-Discrimination Legislation |
81 |
|
|
A Legal Regime Comprising Both an Administrative Structure and a Bundle of Subjective Rights |
82 |
|
|
Data Protection and Anti-Discrimination: Two Regulatory Human Rights |
87 |
|
|
Overlaps: At the Crossroad between Data Protection and Anti-Discrimination |
90 |
|
|
Conclusions: Articulating the Two Rights |
95 |
|
|
References |
97 |
|
|
The Discovery of Discrimination |
104 |
|
|
Introduction |
104 |
|
|
Classification Rules for Discrimination Discovery |
108 |
|
|
Classification Rules |
108 |
|
|
Measures of Discrimination |
109 |
|
|
Direct Discrimination Discovery |
113 |
|
|
Indirect Discrimination Discovery |
114 |
|
|
Argumentation |
116 |
|
|
Affirmative Actions |
117 |
|
|
The DCUBE Tool |
118 |
|
|
Conclusions |
119 |
|
|
References |
119 |
|
|
Discrimination Data Analysis: |
Discrimination Data Analysis: |
|
|
122 |
122 |
|
|
Introduction |
122 |
|
|
Sociological and Legal Perspectives |
123 |
|
|
Labour Economic Perspective |
125 |
|
|
(Quasi-)Experimental Perspective |
128 |
|
|
Auditing |
128 |
|
|
Controlled Experiments |
129 |
|
|
Profiling Perspective |
130 |
|
|
Racial Profiling |
131 |
|
|
Credit Markets |
132 |
|
|
Knowledge Discovery Perspective |
133 |
|
|
Conclusions |
135 |
|
|
References |
135 |
|
|
Risks of Profiling and the Limits of Data |
Risks of Profiling and the Limits of Data |
|
|
149 |
149 |
|
|
Introduction |
149 |
|
|
Risks Associated with Profiling |
150 |
|
|
Discrimination |
150 |
|
|
De-individualisation |
150 |
|
|
Stereotyping |
151 |
|
|
Information Asymmetries |
151 |
|
|
Inaccuracy |
152 |
|
|
Abuse |
152 |
|
|
Privacy and Data Protection in Light of Profiling |
152 |
|
|
Data Protection Law |
154 |
|
|
Drawbacks to the Current Approach to Data Protection in the Context of Profiling |
156 |
|
|
The ‘Binary’ Nature of Data Protection Law |
157 |
|
|
The Procedural Nature of Data Protection Law |
158 |
|
|
Inflation of the Personal Sphere |
158 |
|
|
Data Minimisation |
159 |
|
|
Is Data Protection Law an Adequate Solution? |
159 |
|
|
Shifting the Focus in Data Protection Law |
160 |
|
|
Differentiation in Data Protection: Data Centric Approach |
160 |
|
|
Focus on the ‘Why’ Instead of the ‘What’: Goal Oriented Approach |
161 |
|
|
Revisiting the Moral Reasons for Data Protection |
161 |
|
|
From ‘Privacy by Design’ to ‘Ethics by Design’ |
162 |
|
|
Conclusion |
162 |
|
|
References |
163 |
|
|
Part III: Practical Applications |
165 |
|
|
Explainable and Non-explainable |
Explainable and Non-explainable |
|
|
166 |
166 |
|
|
Introduction |
166 |
|
|
Explainable and Non-explainable Discrimination |
168 |
|
|
Discussion of the Legal Aspects |
168 |
|
|
Motivation for the Explainable Discrimination |
168 |
|
|
Discrimination in Decision Making |
169 |
|
|
Conditional Non-discrimination in Decision Making |
170 |
|
|
An Example on University Admission |
170 |
|
|
Measuring Discrimination |
171 |
|
|
Illustration of the Redlining Effect |
173 |
|
|
Illustration of the Reverse Discrimination |
174 |
|
|
Removing the Illegal Discrimination When Training a Classifier |
175 |
|
|
Techniques |
175 |
|
|
Computational Experiments |
177 |
|
|
Conclusion |
179 |
|
|
References |
180 |
|
|
Knowledge-Based Policing: |
Knowledge-Based Policing: |
|
|
182 |
182 |
|
|
Introduction |
182 |
|
|
Intelligence-Led Policing |
184 |
|
|
Origin and Epistemological Basis |
184 |
|
|
Reorienting Data, Information, Knowledge, and Intelligence |
185 |
|
|
Knowledge-Based Policing |
187 |
|
|
The Need for a New Foundation |
187 |
|
|
The Role of Boundary Objects in Augmented Reality |
188 |
|
|
Realizing the Augmented Reality Potential |
188 |
|
|
Discussion |
191 |
|
|
Databesity: The Ever Present Hunger for Larger Databases |
191 |
|
|
Augmented Reality: Real-Time Processing of Data-Streams |
192 |
|
|
Developing an Ubiquitous Sensor-Network |
193 |
|
|
Dealing with Privacy Invasion |
194 |
|
|
Dealing with Discrimination |
195 |
|
|
Dealing with Group-Think |
195 |
|
|
Sustaining Trust |
196 |
|
|
Consequences for Legislation |
196 |
|
|
Conclusion |
197 |
|
|
Abbreviations |
198 |
|
|
References |
198 |
|
|
Combining and Analyzing Judicial Databases |
201 |
|
|
Introduction |
201 |
|
|
Databases in the Dutch Criminal Justice System |
203 |
|
|
Collecting and Combining Judicial Data |
204 |
|
|
A Data Warehouse Approach to Combining Judicial Data |
205 |
|
|
A Dataspace Approach to Combining Judicial Data |
207 |
|
|
Challenges in Combining Judicial Data |
210 |
|
|
Protecting Privacy When Combining Judicial Data |
212 |
|
|
Risks of Analyzing Judicial Data |
213 |
|
|
Concluding Remarks |
215 |
|
|
References |
215 |
|
|
Part IV: Solutions in Code |
217 |
|
|
Privacy-Preserving Data Mining Techniques: |
Privacy-Preserving Data Mining Techniques: |
|
|
218 |
218 |
|
|
Introduction |
218 |
|
|
Identity Disclosure |
220 |
|
|
Attribute Disclosure |
222 |
|
|
Privacy of Decentralized Data |
224 |
|
|
New Challenges for Data Privacy |
226 |
|
|
Conclusion |
228 |
|
|
References |
228 |
|
|
Techniques for Discrimination-Free Predictive |
Techniques for Discrimination-Free Predictive |
|
|
231 |
231 |
|
|
Introduction |
232 |
|
|
Problem Statement: Discrimination-Aware Classification |
234 |
|
|
Motivation: Links to Legislation |
235 |
|
|
Measuring Discrimination |
236 |
|
|
Techniques for Discrimination-Free Classification |
237 |
|
|
Pre-processing Techniques |
237 |
|
|
Changing the Learning Algorithms |
240 |
|
|
Post-Processing the Induced Models |
242 |
|
|
Experiments |
244 |
|
|
Discussion and Conclusion |
245 |
|
|
References |
246 |
|
|
Direct and Indirect Discrimination Prevention |
Direct and Indirect Discrimination Prevention |
|
|
248 |
248 |
|
|
Introduction |
248 |
|
|
Preliminaries |
251 |
|
|
Basic Notions |
251 |
|
|
Direct and Indirect Discriminatory Rules |
252 |
|
|
Taxonomy of Discrimination Prevention Methods |
253 |
|
|
Types of Pre-processing Discrimination Prevention Methods |
254 |
|
|
Direct Discrimination Prevention Methods |
255 |
|
|
Indirect Discrimination Prevention Methods |
257 |
|
|
Measuring Discrimination Removal |
258 |
|
|
Measuring Data Quality |
258 |
|
|
Experimental Results |
259 |
|
|
Conclusions and Future Work |
260 |
|
|
References |
261 |
|
|
Introducing Positive Discrimination in |
Introducing Positive Discrimination in |
|
|
262 |
262 |
|
|
Introduction |
262 |
|
|
The Naive Bayes Classifier |
265 |
|
|
The Problem of Discrimination in Data-Mining |
266 |
|
|
Discrimination-Free Naive Bayes Classifiers |
270 |
|
|
Using Different Decision Thresholds |
270 |
|
|
Two Naive Bayes Models |
271 |
|
|
A Latent Variable Model |
272 |
|
|
Comparing the Three Methods |
274 |
|
|
A Note on Positive Discrimination |
275 |
|
|
Concluding Remarks |
276 |
|
|
References |
277 |
|
|
Part V: Solutions in Law, |
Part V: Solutions in Law, |
|
|
278 |
278 |
|
|
From Data Minimization to Data Mini$mum$ |
From Data Minimization to Data Mini$mum$ |
|
|
279 |
279 |
|
|
Introduction |
279 |
|
|
Data Mining and Profiling Techniques |
281 |
|
|
Data Protection Legislation |
281 |
|
|
Anti-discrimination Legislation |
282 |
|
|
Data Minimization Principles |
283 |
|
|
Loss of Contextuality |
285 |
|
|
Data Minimummization |
288 |
|
|
Conclusion |
290 |
|
|
References |
292 |
|
|
Quality of Information, the Right to Oblivion |
Quality of Information, the Right to Oblivion |
|
|
294 |
294 |
|
|
Quality of Information |
294 |
|
|
The Quality of Information as an Instrument to Guarantee Certain Fundamental Rights |
298 |
|
|
A New Fundamental Right: The Digital Reputation |
298 |
|
|
Quality of Information and Automated Individual Decisions |
299 |
|
|
Quality of Information and Time and the Right to Oblivion |
300 |
|
|
Conclusions |
303 |
|
|
References |
304 |
|
|
Transparency in Data Mining: From Theory |
Transparency in Data Mining: From Theory |
|
|
305 |
305 |
|
|
Introduction: Transparency, Technology and Prediction |
306 |
|
|
Predictions, Data Mining, Personal Information and Information Flows |
307 |
|
|
Example: Data Mining and Security2 |
307 |
|
|
Prediction and Data Mining: Technology, Human Discretion and Policy Decisions |
308 |
|
|
The Nature of Transparency in Predictive Modeling: Working through the Information Flow |
310 |
|
|
Why Transparency? |
313 |
|
|
General |
313 |
|
|
Transparency – From Theory to Policy |
315 |
|
|
Bringing It All Together: Towards a Policy Blueprint for Transparency |
323 |
|
|
Coda: The Limits of Transparency |
326 |
|
|
References |
327 |
|
|
Data Mining as Search: Theoretical Insights and |
Data Mining as Search: Theoretical Insights and |
|
|
329 |
329 |
|
|
Introduction: Beyond the Visceral Response to Governmental Data Mining |
329 |
|
|
Governmental Data Mining: Definitions, Participants and Problems |
332 |
|
|
Governmental Data Mining and/as (Illegal) Searches? |
333 |
|
|
Finding a Theory |
333 |
|
|
Data Mining as “Searches”: Introducing Three Perspectives |
335 |
|
|
Conclusion: Novel Practices, Classic Concepts and Policy Proposals |
340 |
|
|
References |
342 |
|
|
Part VI: |
Part VI: |
|
|
343 |
343 |
|
|
The Way Forward |
344 |
|
|
Concise Conclusion: Shifting Paradigms |
345 |
|
|
The Failure of Access Controls |
346 |
|
|
The Failure of Anonymity |
348 |
|
|
The Failure of Purpose Specification |
349 |
|
|
Focus on Transparency and Accountability |
351 |
|
|
Further Research |
352 |
|
|
The Future of Discrimination |
354 |
|
|
The Future of Privacy and Data Protection |
357 |
|
|
References |
359 |
|
|
Author Index |
361 |
|
|
Subject Index |
362 |
|