Privacy-Sensitive Data Analysis: Addressing Failures of Existing Statistical Programming Languages

PLEASE SUBSCRIBE IF YOU LIKE THIS VIDEO

This talk was delivered to the Quantitative Methods Network (QMNET) at the University of Melbourne on Junw 12th, 2020 by Professor Simon Dennis, Complex Human Data Hub & Melbourne School of Psychological Sciences, University of Melbourne. The slides can be downloaded here:

TITLE:
Privacy-sensitive data analysis: Addressing the failures of existing statistical programming languages

ABSTRACT:
Popular statistical programming languages like R, python and MatLab make no attempt to protect the privacy of participants. This situation is unsustainable as progress in the behavioural and health sciences increasing relies on sensitive data collected at scale. In this talk, I will describe Private, a new language based on python and BUGS/JAGS/Stan that builds privacy in from the outset. All results in Private are tested for their sensitivity to the data from any given individual before they are released. In addition, Private provides a declarative, but incremental shell that prevents some errors that occur when using procedural languages and automatically distributed computation across a cluster without programmer intervention.

BIO:
Professor Simon Dennis is the Director of the Complex Human Data Hub (CHDH) in the Melbourne School of Psychological Sciences. The primary aim of the CHDH is to develop Computational Behavioural Science – bringing computational methods to bear on big data to better understand psychological processes. Simon’s research utilises large scale real world data, experimental paradigms and computational modelling techniques to investigate the cognitive architecture underlying memory and language. Much of his research uses experience sampling technologies to study psychological processes. He has created an extensive data collection, retrieval, visualization and analysis ecosystem provided by Unforgettable Research Services Pty Ltd of which he is the CEO. Simon also has an interest in privacy and the concept of participant owned data.