Reifying Ethics: Codes of Conduct

Loading...

En provenance du cours de University of Washington

Communicating Data Science Results

124 notes

University of Washington

Communicating Data Science Results

124 notes

Cours 3 sur 4 dans Specialization Data Science at Scale

Important note: The second assignment in this course covers the topic of Graph Analysis in the Cloud, in which you will use Elastic MapReduce and the Pig language to perform graph analysis over a moderately large dataset, about 600GB. In order to complete this assignment, you will need to make use of Amazon Web Services (AWS). Amazon has generously offered to provide up to $50 in free AWS credit to each learner in this course to allow you to complete the assignment. Further details regarding the process of receiving this credit are available in the welcome message for the course, as well as in the assignment itself. Please note that Amazon, University of Washington, and Coursera cannot reimburse you for any charges if you exhaust your credit. While we believe that this assignment contributes an excellent learning experience in this course, we understand that some learners may be unable or unwilling to use AWS. We are unable to issue Course Certificates for learners who do not complete the assignment that requires use of AWS. As such, you should not pay for a Course Certificate in Communicating Data Results if you are unable or unwilling to use AWS, as you will not be able to successfully complete the course without doing so. Making predictions is not enough! Effective data scientists know how to explain and interpret their results, and communicate findings accurately to stakeholders to inform business decisions. Visualization is the field of research in computer science that studies effective communication of quantitative results by linking perception, cognition, and algorithms to exploit the enormous bandwidth of the human visual cortex. In this course you will learn to recognize, design, and use effective visualizations. Just because you can make a prediction and convince others to act on it doesn’t mean you should. In this course you will explore the ethical considerations around big data and how these considerations are beginning to influence policy and practice. You will learn the foundational limitations of using technology to protect privacy and the codes of conduct emerging to guide the behavior of data scientists. You will also learn the importance of reproducibility in data science and how the commercial cloud can help support reproducible research even for experiments involving massive datasets, complex computational infrastructures, or both. Learning Goals: After completing this course, you will be able to: 1. Design and critique visualizations 2. Explain the state-of-the-art in privacy, ethics, governance around big data and data science 3. Use cloud computing to analyze large datasets in a reproducible way.

À partir de la leçon

Privacy and Ethics

Big Data has become closely linked to issues of privacy and ethics: As the limits on what we *can* do with data continue to evaporate, the question of what we *should* do with data becomes paramount. Motivated in the context of case studies, you will learn the core principles of codes of conduct for data science and statistical analysis. You will learn the limits of current theory on protecting privacy while still permitting useful statistical analysis.

Motivation: Barrow Alcohol Study5:57

Barrow Study Problems4:59

Reifying Ethics: Codes of Conduct6:50

ASA Code of Conduct: Responsibilities to Stakeholders4:43

Other Codes of Conduct6:51

Examples of Codified Rules: HIPAA3:30

Rencontrer les enseignants

Bill Howe
Director of Research
Scalable Data Analytics

[MUSIC]

So one discussion that you see in ethics in general is whether

your talking more about principles or more about rules.

So principles are these overarching guidelines,

but stop short of really trying to specify specific behavior.

Other folks believe that at the goal that the right thing to do is to actually write

down a firm set of rules that should not be violated and

adheres to these rules, constitutes ethical behavior.

So I think that I come down fairly strong on the side of principles,

because once you get into rules, you're really, I push,

let's see, just from the previous slide, if you're want to talk about rules then I

want to talk about formal guarantees, mathematical proofs about things.

And we'll talk about some of the limits of what you can do with that.

And if you're gonna give up on those, it's not clear to me that

legislative rules have any extra power.

So declaring and

agreeing on the principles we should agree to and writing them

down perhaps in the form of these codes of conduct I think is the right way to go.

What you're doing is appealing to people's

beliefs about these topics as opposed to trying to sort of control their behavior.

Okay, so I think and in the Barrow example I think demonstrates this where

really the ethical principles that were violated.

They essentially followed the rules that were applied at a time.

And really many of the rules that have been developed since then

in terms of how to handle private data we're not necessarily violating.

But at several steps along the way there were ethical principles violated, putting

their own needs and their own interests ahead of those of the research subjects,

the client, and society.

And recommendations against doing that do appear in these codes of conduct that

we'll look at briefly.

So I want to just consider three and what's.

It's sort of interesting is that this is not really cherry picked.

This is kinda an attempt at finding all the lists, all the codes of conduct

that one can find on

put out there by organizations representing the data science community.

So I won't claim that actually is exhaustive.

I'm sure they're popping up all the time and

I'm sure there's many that I've missed.

But there's not a lot of them.

Nor is there one single that everyone's agreed and

that's the reason why there's not a lot of them, it's still quite early.

The other thing I'll point out is that this first one on this list,

the American Statistical Association is just that,

the American Statistical Association.

It's not a data science organization and so this code of ethics has been around

longer, and I think that you can tell just from reading, the maturity is higher.

So one question you can ask is whether we need a separate data science code

of conduct, different from that of statisticians.

Whether you think you should or not, it's useful, I think,

to be aware of at least these three and have read through them and

sort of consider what points they're bringing to the fore front and

what emphasis they're placing on things, and I think it was instructive for

me to compare some of these points and I think it will be instructive for

you to have seen them at least in this form in the lecture, okay.

Okay, so the first one we're going to look at is

the code of conduct from the American statistical Association.

It has the following top level sections.

One of them professionalism, and

then the rest of this is broken down by responsibilities to various parties.

And the first one is funders, clients, and employers, publications and testimony,

research subjects, research team colleagues, other statisticians or

statistical practitioners, the community at large in your field, and

regarding allegations of misconduct.

And then of employers themselves surrounded in your responsibility to

the employers, responsibility of an employer of said practitioners.

And I really like this breakdown because we'll see

a couple other ways just from the divided space and I think it gets a lot muddier.

I think thinking specifically about the parties that

have the stake holders involved.

When you are taking on these projects and thinking about their needs and considering

them all at every step along the way, is the right way to think about this.

Okay so I just want to show some, in some cases I paraphrased these,

in some cases I provided in verbatim and we're not going to initially read every

bullet on every page because you can go to the web and do that.

But one is I want to make sure that you have access them as you're

sort of watching through this video, you have an excuse to kind of look at them and

read through them if you wish.

And then the other thing I think I'll do is highlight a few of the ones that

strike me as relevant or arguably,

perhaps violated, in the barrow example we considered.

So the first section of the Code of Conduct, from the ASA, is just

professionalism corresponds pretty well to the codes of conduct for professional

engineering societies as well, where it's sort of how to do the job properly.

So, relevance and statistical analyses, remaining current in the practice,

bringing a certain level of subject matter expertise using only those methodologies

that are appropriate for the data.

The one I've highlighted in blue here is guard against the possibility that

a predisposition by investigators or

data providers might predetermine the analytic result.

I think that's unique to statistical and data science, as opposed to engineering.

And I think it was relevant perhaps in the Barrow example where it seemed like

the sampling procedure and it seemed like the way the results were presented and it

seemed like some of the methodologies were designed with a specific outcome in mind.

Namely, that the ethnic and cultural identity of the research subjects

would be linked to the certain behaviors of the use of alcohol.

And so that's a pretty dangerous idea to come in to the study with.

Now I'm not making a lean claim that we have proof of that one way or another, but

I think that a strong argument could be developed in in favor of that.

Okay, so another thing' here is respect the contribution of intellectual

property and disclosed conflicts of interest and so on.

The last one here is sort of interesting too, which is expert testimony should be

given with the same level of vigorous peer view research.

So this is you know, when you're adding something to the scientific literature,

you jump through all sorts of hoops but if you're talking to your employer just for

giving them results and say a power point presentation.

Maybe it could be a little looser with the rigor, and the pointers know you shouldn't

do that and these things have a lot of ethical topics.

Things are sort of obvious but I think crafting the language around,

holding it up as a document that people can sort of read internalize so

that they start using that language too.

It's important even when the basic idea is somewhat obvious.

[MUSIC]

Explorer notre catalogue

Rejoignez-nous gratuitement et obtenez des recommendations, des mises à jour et des offres personnalisées.

Coursera propose un accès universel à la meilleure formation au monde, en partenariat avec des universités et des organisations du plus haut niveau, pour proposer des cours en ligne.

© 2019 Coursera Inc. Tous droits réservés.

Télécharger dans l'App Store

Disponible sur Google Play