Page 14 - FIGI - Big data, machine learning, consumer protection and privacy

and framing key issues for consideration when devel- to, and responsibility for the things firms may do with
oping regulatory frameworks (including potentially personal data, including in relation to accuracy in
self-regulatory frameworks). machine learning models (section 6.1), bias and dis-
23
Section 4 introduces the key concepts in play, criminatory treatment (section 6.2), data breach and
starting with the technology and market trends re-identification (section 6.3), and transfer of data to
of big data and machine learning (section 4.1), the third parties (section 1.1).
kinds of data that are used (section 4.2), what profil- Section 7 turns to the post-engagement phase,
ing and automated decisions use such data (section and the consumer’s means of holding big data and
4.3). It then turns to explain and then the regulatory machine learning operators accountable for viola-
dimensions that these raise throughout the paper: tions of consumer protection and data privacy laws.
consumer protection (section 4.4) and data privacy It looks at consumers’ rights to access personal data
(section 4.5). about themselves, rectifying errors in it and request-
The paper then proceeds to consider consumer ing that it be erased (section 7.1), transparency dif-
protection and data privacy in three broad phases ficulties with obtaining explanations for complex
of the consumer’s encounter with service providers machine learning model outputs (section 7.2), the
that rely on big data and machine learning: right to contest decisions and obtain human inter-
Section 5 discusses the pre-engagement phase, vention (section 7.3), and the challenge of showing
which primarily concerns what disclosures and notifi- harm (section 7.4).
cations are required to be made to consumers about The paper discusses in section 8 some practical
how and for what purpose their personal data will steps firms may take to reduce risk in face of the
be collected, used and transferred to third parties, legal and regulatory uncertainties. It closes in section
and requirements for obtaining consumer consent to 0 with a short list of areas for further development
legitimize use of personal data. in this field, whether in the development of ethics,
Section 1 discusses the engagement phase, which standards or procedures.
relates to the restrictions on, requirements relating

2 UNDERSTANDING BIG DATA, CONSUMER PROTECTION AND DATA PRIVACY

2�1 What are big data and machine learning? Figure 1 – Machine learning, xkcd.com
Artificial intelligence involves techniques that seek
to approximate aspects of human or animal cogni-
tion using computers. Machine learning, a form of
artificial intelligence, refers to the ability of a system
to improve its performance, often by recognising
patterns in large datasets, doing so at multiple layers
of analysis (often referred to as deep learning). 24
Machine learning algorithms build a model from
training data, i.e., historical examples, in order to
make predictions or decisions rather than follow-
ing only pre-programmed logic. Neural networks
analyze data through many layers of hardware and
software. Each layer produces its own representa-
25
tion of the data and shares what it “learned” with the
next layer. Machine learning learns by example, using
the training data to train the model to behave in a
certain way. Machine learning is not new, but as a
26
result of big data, it is suddenly being deployed in
numerous practical ways.
27
Big data relies upon and is typically defined by,
computer processing involving high volumes and
varieties of types of linked up data processed at high

12 Big data, machine learning, consumer protection and privacy