L5: Technology and morale in banking Flashcards
what are hard information and soft information
hard information can be accurately quantified and efficiently transmitted. soft information is hard to quantify and transmit efficiently.
Role of Fintech in helping spread information
Fintech made it easier to quantify and transmit information once considered soft. Undermining ‘;soft information monopolies’ intensifying competition in the industry.
What is credit discrimination
information processing for credit decisions causing unjust treatment of different categories of people.
Types of discrimination
Access discrimination, price discrimination and algorithmic discrimination
what causes algorithmic discrimination
dataset issues, algorithmic issues, outcome issues.
How do algorithmic issues happen?
Even when the factor not included in input, proxies for gender, race etc such as where you shop, where you live can create biases in algorithms
what are outcome issues
computer implements algorithm itself or human implements results without understanding algorithmic steps or biases.
Apple Card case study for technology and discrimination.
Partnership between Apple and Goldman Sachs in 2019. Couple found that husband had 20x the credit limit than wife who had better credit score. Noone at Apple fully understood the algorithm, Goldman insisted no gender bias as it does not use gender as an input. DFS investigatged and performed flip test (swap genders), but often in financial services auditors are not allowed this sensitive information which can cover discrimination. Laws and regulators not up to date with technology.
algorithm fairness concepts
veil of ignorance, equality of opportunities, equality of odds, demographic parity.
what is the veil of ignorance algorithm fairness concept
the model does not account for X so it does not discriminate against X
