Page 12 - FIGI - Big data, machine learning, consumer protection and privacy
P. 12
machine learning to predict young adults’ creditwor- ing has triggered a vigorous policy debate about its
thiness drawing from data on their education, exam risks, and the need for coherent policy. The World
13
scores, academic field of study and job history data, Bank prepared a report for the 2018 G20 summit,
in an automated loan process. It offers loans direct- Use of Alternative Data to Enhance Credit Reporting
ly to consumers, as well as offering other lenders its to Enable Access to Digital Financial Services by Indi-
software as a service, i.e., a platform for their own viduals and SMEs operating in the Informal Econo-
lending services. my analysing key issues and making recommenda-
14
These business models are being increasingly tions to policy makers and regulators, on which this
deployed for financial inclusion not only in wealthy report builds.
nations but also in developing countries. Lenndo, a The Monetary Authority of Singapore recent-
5
fintech firm supporting credit evaluation with alter- ly published Principles to Promote Fairness, Ethics,
native data analysis, has partnered with the global Accountability and Transparency (FEAT) in the Use
credit agency FICO to make FICO score services of Artificial Intelligence and Data Analytics in Sin-
available in India. This service evaluates alternative gapore’s Financial Sector. These seek to apply FAT-
6
data from a consumer's digital footprint to produce style principles specifically to the context of AI and
a credit score for those who do not have sufficient machine learning in the financial sector, adding an
traditional data on file (“thin file” borrowers) with ethical dimension. The FEAT Principles are set out
one of the Indian credit bureaus for a traditional in Annex A (Monetary Authority of Singapore FEAT
loan approval. Branch.co and MyBucks are active in Principles). The Smart Campaign recently released
7
8
Africa and beyond, using identity proofing and auto- draft Digital Credit Standards, which include a num-
mated mobile app that uses credit-scoring engines ber of standards addressing use of data, profiling and
to generate credit scores from analysing a custom- automated decisions in digital financial services, and
er’s mobile phone bill, text messages, payment his- which are set out in Annex B (Smart Campaign Dig-
tory, bank account history (if the person has a bank ital Credit Standards). These will be referred to from
account), utility bills and geolocation data. time to time in this report to illustrate ways in which
Rapid access to large volumes of data is key to consumer protection issues might be approached.
the effectiveness of such technologies. For instance, The IEEE’s Global Initiative for Ethical Consid-
ZestFinance has a strategic agreement with its erations in Artificial Intelligence and Autonomous
investor Baidu, the Chinese internet search provider Systems has called for legislators to consider regu-
(equivalent of Google in China) that allows ZestFi- lation: 15
nance to access individuals’ search history, geoloca- Lawmakers on national, and in particular on inter-
tion and payment data to build credit scores in Chi- national, levels should be encouraged to consider
na, where around half of the population has no credit and carefully review a potential need to introduce
history. ZestFinance’s CEO famously said, “all data is new regulation where appropriate, including rules
9
credit data.” 10 subjecting the market launch of new AI/AS driven
Artificial intelligence is not only useful for credit technology to prior testing and approval by appro-
risk assessment. Any service involving risk assess- priate national and/or international agencies.
ment depends on information and analysis. The firm Longstanding laws and regulations that aim to
Progressive, for example, collects data on individuals’ protect consumers from adverse uses of personal
drivers’ driving performance through mobile applica- data are facing various challenges in terms of new
tions like Snapshot in order to predict risk of acci- data collection and analytical methodologies. Indeed,
dents and offer (or not) discounted insurance premi- some have ventured to say that even the most recent
ums. Artificial intelligence is being used in numerous of data protection and privacy laws, Europe’s Gen-
11
other applications in the field of insurance. Other eral Data Protection Regulation (GDPR), sometimes
12
areas where artificial intelligence is having a substan- referred to as the “gold standard” of data protection
tial impact on innovation and improvements to effi- and privacy law, is “incompatible” with the world of
ciency include personalization of savings products, big data. Similar concerns arise in relation to other
16
management of payment services, provision of virtu- global standards, such as the OECD Recommenda-
al assistance for customers (e.g., robo-advisory and tion Concerning Guidelines Governing the Protection
chatbots), and detection of fraud, money laundering of Privacy and Transborder Flows of Personal Data
and terrorism financing. (the OECD Privacy Guidelines) and the Convention
17
The rise in consumer use of products and services for the Protection of Individuals with regard to Auto-
relying on artificial intelligence and machine learn- matic Processing of Personal Data (ETS No. 108)
10 Big data, machine learning, consumer protection and privacy