bigdata ethics presentation, Slides of Computer Science

Download bigdata ethics presentation and more Slides Computer Science in PDF only on Docsity!

Big Data Analytics

Big Data and Ethics

Jesse Eickholt, PhD

Ethics

“An ethical analysis [of … ] would be

well served by starting with the

concepts of ethics. However …”

Timmermans et al., The Ethics of Cloud Computing.

…this would require a description of a

philosophical foundation from which

concepts of morality can be

developed.

A large financial institution wants a “big

data” approach to determining a who would

be a “risky” customer (i.e., who should be

charged a higher interest rate).

A bright, up-and-coming employee decides

to scrape social media posts for all current

employees. This information is combined

with the companies historical payment

data.

Big Data AnalyticsBig Data Analytics

Example

Maybe it works great! Identifies a “risky” customer 80% of the time. o (^) Facebook posts o (^) Tweets o (^) Comments posted on websites Extract common words (e.g., positive and negative), common phrases used (e.g., my boss is a ….), movies and music tastes.. Payment history (e.g., on time, late, etc.) Machine learning Classify as: Risky / Safe Use to make predictio ns on new, potential customer s.

A health insurance company wants to have a better idea of the health of its customers. Clients with additional “risk factors” will be charged more or possibly denied coverage. The company monitors social media posts and looks for mentions of restaurants visited and looks a purchases on Amazon. Wait! How did they get that data?

A crude heuristic is developed to look

for certain combinations of fast food

restaurants and certain sizes of clothing

(e.g., many fast food restaurants and

large clothing sizes indicates a risk).

Again, what is wrong with this

approach?

The department store would also like

to know when a customer visits the

store but does not make a purchase.

The store offers free WiFi and

customers can connect via their phone

which contains a unique identifier

(MAC address).

What? No way! How? Ok. But how to connect to a customer’s name.

MAC addresses are tied to a customer name

by two routes…

i) cross checking credit card/reward

purchases with the presences of a particular

MAC address [if they both coincide more

than 5 times could it be a coincidence?]

ii) customers can download and use an in

store app which provides coupons through

discounts.

To collect even more data the

department store logs what websites

(or IP addresses) are accessed through

the store’s WiFi.

To summarize … the department store

knows…

• what products a customer buys
• when the customer visits the store (even if

nothing is bought)

• where the customer goes in the store and

which stores the customer has visited

• what is viewed through the store’s WiFi
• details that the store’s app was able to pull

from the phone

Why do this?

If you ask a company what data they

are collecting, they may or may not

tell you and they may or may not give

it to you.

How might this type of data be used

against you? Who might want to buy

it?

What is the Driving Force?

Not to be cynical, but profitability is

key here!

There is value in that data!