Trusted Smart Statistics A reflection on the future

Trusted Smart Statistics: A reflection on the future of (Official) Statistics Konstantinos Giannakouris, Eurostat, konstantinos. giannakouris@ec. europa. eu Fernando Reis, Eurostat, fernando. reis@ec. europa. eu Michail Skaliotis, Eurostat, michail. skaliotis@ec. europa. eu Albrecht Wirthmann, Eurostat, albrecht. wirthmann@ec. europa. eu Fabio Ricciato, Eurostat, fabio. ricciato@ec. europa. eu 27 June 2018 Session 27

Table of "questions" What do we expect in a world of Io. T and AI? Consequences for official statistics? What can we call smart and trusted statistics? Computational challenges? Implications in trust? Scope for algorithmic decision making in statistical operations?

If the future is becoming smarter in a hyper-connected environment … Capabilities of smart technologies embedded in smart devices/systems Sense monitor recognise Modify adapt interact Learn Conclude think reason anticipate Decide self-generate self-sustain

… and the Internet of Things prevails Smart sensors Smart building (building automation) Smart phone Smart cities, smart cars, smart traffic, smart parking Smart home Smart TV Smart grids Smart fridge Smart manufacturing Industry 4. 0 Smart farming

… what else could be the future of official statistics but smart? Statistics data layer within smart systems Embedded transformation of data Ambient intelligence of smart statistics

Trusted statistics might not be good enough Validity and accuracy of the outputs Privacy and confidentiality Institutional environment Privacy by design, end to end security Auditable data life-cycle, transparency Consent and purpose, Assessment boards Assessment, certification boards

Trusted Smart Statistics computational challenges Secure infrastructure Legal compliance Statistical Office Data input Trust in Input Data • • Data is veracious Data provision is continuous, stable processing Information output Trust in processing Trust in output Data are processed only Statistics described using quallity frameworks, correspond to perceived situation • • for agreed purpose by agreed method

Trusted Big Data Smart comes. Statistics into play Statistical Office collection Trusted Smart Statistics y cop Public Sector processing processing use Society Economy Environment processing Push out computation (partially) Private Sector New computational models Public Policy Media Research

It is happening…(algorithms ) Supermarket 1 Asian food Rice HICP Vegetables & potatoes Noodles Vegetables 50 000 global Trade Item Number (GTIN) 2 -3% change period Supermarket specific classification Machine Learning (for decision making) European Classification of Individual Consumption according to Purpose (ECOICOP) Country A

From handbooks to AI Prioritisation Data Processes embedded in aggregators, hubs aggregators, … hubs Data ta Da Da ta Data … Algorithmic processing Classification Association processing ta Da Filtering Data ta Da

Trusted smart statistics • (Trusted) Smart statistics conceptually encompass capabilities of smart systems • (Trusted) Smart Statistics can be seen as a technology embedded into smart systems aiming at transforming data to information that is presented in the form of statistics • Trusted Smart Statistics can be seen as a service provided by smart systems, • ensuring validity and accuracy of the outputs, and • respecting data subjects' privacy and protecting confidentiality

Trusted Smart Statistics: A reflection on the future of (Official) Statistics Thank you for your attention Konstantinos Giannakouris, Eurostat, konstantinos. giannakouris@ec. europa. eu Fernando Reis, Eurostat, fernando. reis@ec. europa. eu Michail Skaliotis, Eurostat, michail. skaliotis@ec. europa. eu Albrecht Wirthmann, Eurostat, albrecht. wirthmann@ec. europa. eu Fabio Ricciato, Eurostat, fabio. ricciato@ec. europa. eu