Social Republic.

This is a large federal company that is....

pretty common honoring dancers in 1990.

And Andrea O rebirth in existed just as he did in the 40s.

And probably be doing this

all day and hearing each day about your stages.

It guards the world today

and never goes away.

Find out how to get in these 2

to 3 minutes.

for Learning and Intelligence Systems, ELIS, where she is part of the Robust Machine Learning

Program and the Saarbrücken Artificial Intelligence and Machine Learning Unit.

Prior to this, she was an independent group leader at the MPI for Intelligence Systems

in Tübingen, Germany.

She has held a German Humboldt Postdoctoral Fellowship and a Minerva Fastrock Fellowship

from the Max Planck Society.

Professor Valera obtained her PhD in 2014 and Master of Science degrees in 2012 from

the University Carlos VIII in Madrid, Spain and worked as a postdoctoral researcher at

the MPI for Software Systems Germany and at the University of Cambridge in the United

Kingdom.

Today, it's a great pleasure to have you here and she will present on ethical machine learning

mind the assumptions.

Isabelle, I'm very glad to have you here and I'm very much looking forward to your presentation.

Thank you very much for inviting me to be here today.

Yes, and for the kind introduction and all my bio where I moved a lot.

I hope I have settled a bit for now.

And as we were talking today, we are going to talk about machine learning and some ethical aspects.

In particular, we will be talking about fairness and interpretability.

Rather than putting a lot of focus on methodological approaches,

I will mostly focus on assumptions and how technical assumptions

may actually trigger some problems if we don't make them carefully.

So just as a bit of motivation, the context where I will be mostly discussing

is machine learning to inform what I call consequential decisions.

And I call them consequential because I'm talking about settings in which the algorithm

is supplementing or even replacing human supervision in decision making processes

that may have important consequences in people's life or the individuals that they decide about.

So, for example, here we have some examples on pre-trial bail.

Some of the tools that are being used in the US are actually informed by machine learning.

So many of you might be familiar by now with the Compass dataset.

We are also considering settings like credit approval or hiring processes

where machine learning is becoming more and more ubiquitous around it.

And the questions that we are going to be handling is actually,

or the big question actually is how do we make sure that these algorithms

and the decisions that they make are aligned with human goals and values.

And of course, when I talk about human goals and values,

we are different individuals in society.

We have a different role every time that we are embedded in these kind of decisions.

And if you think of the owner of the algorithm,

maybe this individual is interested in making the algorithm precise,

to have a good performance and therefore to maximize the profit of the company