Page 11 - FIGI - Big data, machine learning, consumer protection and privacy

Big data, machine learning,

consumer protection and privacy

1 INTRODUCTION

Big data, artificial intelligence and machine learning sons who have borrowed and repaid or defaulted on
are dominating the public discourse, whether from debts in the past.
excitement at new capabilities or fears of lost jobs Digital financial service providers can not only
and biased automated decisions. The issues are generate commercial profit but, with informa-
not entirely new. However, public awareness of the tion about and analysis of consumers’ background
1
potential of powerful computing systems applying and interests, can also add substantial public value
complex algorithms to huge volumes of data has through improved access to financial services.
grown with stories of computers beating humans Artificial intelligence is increasingly used to ana-
at games and as people increasingly enjoy services lyze a wide range of data sources to create a coher-
produced by such systems. 2 ent assessment of consumers’ creditworthiness and
Personal identifiable data is widely collected, make lending decisions. Instead of relying merely on
shared and available on commercial data markets. the borrower’s representation of income and existing
Such data may include an individual’s internet and debts in the loan application, or an interview by the
transaction history, registration with public and pri- local bank manager, or checking a credit reporting
vate organizations, and use of social media. Firms agency’s score (e.g., FICO), the combination of artifi-
and governments routinely collect, process and cial intelligence and big data allows firms to analyze
share such data with third parties, often without the an individual’s digital footprint to predict the proba-
user’s knowledge or consent. bility of default. This enables access to services that
The beneficial opportunity data presents for devel- may otherwise have been unavailable.
opment is widely recognized, particularly for the pro- Big data analytics may be used to enhance tradi-
vision of digital financial services. Many financial ser- tional means of credit assessments. Credit reference
3
vices depend on risk assessment and management. bureaus, such as Equifax, have claimed to have made
For example, a loan’s value is in large part based on significant improvements in the predictive ability of
the borrower’s creditworthiness, as well as the collat- their models by using big data analytics. This can be
eral that may secure the loan. The more data there is particularly useful in assessing individuals who lack
about the borrower, the better the lender can assess a traditional credit history, thus giving them access
their creditworthiness. Big data enables inferences to credit services. This opportunity extends beyond
about creditworthiness to be drawn from a borrow- enhancing traditional means of credit assessment
er’s membership of one or more categories of per- to entirely new models. For example, Upstart uses
4

Big data, machine learning, consumer protection and privacy 9