ETHICAL QUESTIONS ON THE USE OF BIG DATA

ETHICAL QUESTIONS ON THE USE OF BIG DATA IN OFFICIAL STATISTICS ESS Big Data Workshop 2016, 13 -14 OCTOBER 2016, LJUBLJANA Alma Rutkauskiene@sogeti. lu

Objective • To identify possible ethical issues • To formulate ethical questions • To propose ethical approach • Possible impact on Co. P

• Ethics - a social, religious, or civil code of behaviour considered correct, especially that of a particular group, profession, or individual • An ethical issue is a problem or situation that requires a person or organization to choose between alternatives that must be evaluated as right (ethical) or wrong (unethical). • Ethical question - asks people to choose which solution(s) or solutions can be considered "right" or ethical

Official statistics - Professional ethics • Declaration on professional ethics, adopted by the International Statistical Institute (ISI) Council in 2010. • UN Fundamental principles of official statistics • The European Statistics Code of Practice (Co. P)

Ethics of Big Data • Privacy - People should have the ability to manage the flow of their private information across massive, third-party analytical systems, to control access; private data have to be kept confidential • Identity - The identification of individuals and organizations should not be done without participation or agreement. • Ownership - For big data to work in ethical terms, the data owners (subjects) need to have a transparent view of how their data is being used • Reputation - For big data to work in ethical terms, the data owners (subjects) need to have a transparent view of how their data is being used

Stakeholders • The data subjects (individuals and companies). • Businesses that collect and hold Big Data • Users of statistical data (researches, politicians etc. ) • The general public

Possible ethical issues • Situation • Ethical questions • Professional values and principles • Possible ethical approach

1. Access to Big Data • Big Data in most cases collected by private companies for their business purposes. • Statistical authorities – to find the way to access the data. What can be given to private companies in return? • Professional independence of official statistics and impartial treatment of the data providers • Partnerships with the data holders; Tailored data; • Changes in the law provisions

2. Purchases of the data • Core of data holders’ business model → Commercial value → provided to the clients at substantial price • Official statistics and commercial relationship with data providers ? Professional independence of official statistics and impartial treatment of the data providers • To select the data providers on the basis of professional statistical considerations. • Purchases - one for all.

3. Trust in Big Data providers • Big Data holders not always accountable companies. Consumers are unaware how the data is used. Data brokers. • Legitimacy and privacy policy of the Big Data holders? • Responsibility. Transparency. Reputation of official statistics • Statistical authorities - to be informed about the privacy policy of the data holders

4. Data subjects’ consent • Sensitive personal information. Risk of disclosure commercially valuable information. • Should statistical authorities be bound with the requirement of clear data subjects’ consent? • Respect the privacy of data subjects • Big Data holders inform the customers through “Terms and Conditions” of their services

5. Predictive models and data linking • Predictive models. Possibility to create new information that is not known in advance (mosaic effect). Risk disclosure personal and commercially valuable information. • What would be a right (ethical) way to use predictive models and data linking in official statistics? • Confidentiality of individual data • Clear and strong rules of data security and data confidentiality protection and communication of them to public. It concerns the directly or indirectly identifiable data.

6. Continuity and stability of Big Data sources • Big Data are not designed for statistical purposes. For some of the sources the data holders may decide to collect different data or entirely stop their data collection. • • What would be a right way of use such data sources for official statistics ? • Statistical authorities strive to collect and analyse data of the highest quality possible • NSIs collaborate with the data holders to obtain knowledge on the origin of Big Data sources; with academia to develop new methodologies;

7. Manipulation • Some of the Big Data sources can be easily accessed by different users. Risk of being manipulated. • What would be a right way of use such data sources in official statistics ? • Quality of the data sources. Transparency. • Data Quality checks • If manipulation cannot be detected and corrected – not be usable for official statistics

8. New skills – outsourcing analytics • Big data analysis require special skills that at present are not common at statistical agencies. • What would be a right way of using Big Data analytics of private companies (that possess the appropriate skills) in producing official statistics ? • Professionalism. Responsibility. • Partnerships with companies - the role of NSI as guarantor of quality and impartiality

9. Complexity of Big Data analysis • Methods of producing statistical output might be so complex that in practice cannot be verified. • What would be a right way to disseminate the statistical output if the techniques to obtain it cannot be easily explained to the users? • Transparency of the sources and methods used. • Methods and their parameters have to be published

10. Competition or cooperation • Other players in the information market of producing statistics using Big Data. Competition – duplication of work. • What would be a right way for statistical agencies to avoid competition in producing statistics? • Statistical systems should collaborate rather then compete with the private sector operators • Forming partnerships with private entities under the leading role of the statistical agencies

Possible impact on the European Statistics Code of Practice and indicators of its implementation

What has to be emphasised • Duty to explore new data sources (Big Data) in order to • • adapt to the digital world Impartial treatment of the Big Data providers Trust in Big Data providers Personal data treatment policy Knowledge on the origin/provenance of the Big Data sources Clear rules on the use of Big Data analytics and predictive models Forming partnerships with Big Data providers Active collaboration with academia