密歇根大学

Data Science Ethics

25 评分

What are the ethical considerations regarding the privacy and control of consumer information and big data, especially in the aftermath of recent large-scale data breaches? This course provides a framework to analyze these concerns as you examine the ethical and privacy implications of collecting and managing big data. Explore the broader impact of the data science field on modern society and the principles of fairness, accountability and transparency as you gain a deeper understanding of the importance of a shared set of ethical values. You will examine the need for voluntary disclosure when leveraging metadata to inform basic algorithms and/or complex artificial intelligence systems while also learning best practices for responsible data management, understanding the significance of the Fair Information Practices Principles Act and the laws concerning the "right to be forgotten." This course will help you answer questions such as who owns data, how do we value privacy, how to receive informed consent and what it means to be fair. Data scientists and anyone beginning to use or expand their use of data will benefit from this course. No particular previous knowledge needed.

从本节课中

Societal Consequences

In Module 8, we consider societal consequences of Data Science that we should be concerned about even if there are no issues with fairness, validity, anonymity, privacy, ownership or human subjects research. These “systemic” concerns are often the hardest to address, yet just as important as other issues discussed before. For example, we consider ossification, or the tendency of algorithmic methods to learn and codify the current state of the world and thereby make it harder to change. Information asymmetry has long been exploited for the advantage of some, to the disadvantage of others. Information technology makes spread of information easier, and hence generally decreases asymmetry. However, Big Data sets and sophisticated analyses increase asymmetry in favor of those with ability to acquire/access.

Societal Impact16:26

Ossification7:24

Surveillance4:54

Case Study: Social Credit Scores7:59

Case Study: Predictive Policing8:49

与讲师见面

H.V. Jagadish
Bernard A Galler Collegiate Professor
Electrical Engineering and Computer Science

Let's talk for a moment about surveillance.

We are collecting more and more data about us,

and this is just going to keep increasing.

We have video cameras, of course,

but we have smart meters that are recording data about what's happening with

our utility consumption in ways that old style mechanical meters just didn't do.

We have things like Fitbits that are recording every activity that we participate in,

and every move we make literally.

As this continues, we're going to get a record of everything that we do,

every breath that we take.

And we do this because we get benefits,

as individuals, from the data that we collect about ourselves.

If you wear a Fitbit,

you do so because you want the health benefits from monitoring your exercise.

There are societal benefits too from the collected data.

For example, law enforcement may get

benefits from looking at some of these detailed records.

Medical science may get benefits from

analysis of large numbers of electronic medical records.

So, potentially, at one extreme,

we could decide, as a matter of policy,

that we want all this information record to be completely open,

to be accessible to everyone,

to live in a world where everybody knows everything,

and there's no deceptions.

The problem with that is the deceptions actually may matter.

We need deceptions.

We have an explicit notion of expunging past history for a fresh start.

People's records are cleared of

their prison time for many offenses after seven years in the US,

of driving offenses after three years.

We know that personal relationships are

hard and sometimes having a fresh start is important.

Just in terms of our social interactions,

we tell little white lies all the time.

We would need to adjust dramatically if we needed

to live in a world where these little white lies weren't possible,

and we had the stark glare of complete honesty.

There's another issue to remember.

The process of quantification loses meaning.

There's often a lot of richness in understanding the actual data.

But when it's reduced to a number,

that number may not reflect what we actually had in that rich data.

We should be wary of overreliance on such quantified numbers,

which is often the record that we create.

So these are a few reasons why deceptions matter.

There are more obvious and more serious things too,

where if you live in a country with an authoritarian government,

then small deceptions may actually be necessary even just to lead a normal life.

So for all of these reasons,

we've got to figure out,

as we create this valuable and enormous record of data,

how we're going to control who has access to what,

and there are difficult choices to make.

And it's particularly hard because of the anonymity conundrum,

which is that much value can come from the analysis of anonymized data sets,

but anonymization can be broken with sufficient effort in most cases.

And so, we have to worry about sharing of any sort

because unrelated patterns can provide identification.

As an example, Netflix released deidentified data of movies that people had rented,

and the scores that they gave to those movies.

Some of these people,

for some of these movies,

had also posted reviews on social media,

and correlating these was enough to re-identify.