Sandra Matz

What do you think here kind of companies decide whether you might get a loan or not, whether you might get credit or not, what your insurance premium is or not? It's very similar. We try to understand how you might behave and then we shift the offerings that we have. We might try to sell you something that you don't need.

So I think this notion that it's not just about privacy, it's really about the second step of people then interfering with your ability to make your own choices. For me, that part is almost creepier.

So I think this notion that it's not just about privacy, it's really about the second step of people then interfering with your ability to make your own choices. For me, that part is almost creepier.

So I think this notion that it's not just about privacy, it's really about the second step of people then interfering with your ability to make your own choices. For me, that part is almost creepier.

And they might all be related to some of the protected categories, right? If we know that some of the behaviors that we show are related to you having low socioeconomic status or to your ethnicity or to your sexual orientation, then you don't need to capture that category because it's like somewhere embedded in the traces that you leave.

And they might all be related to some of the protected categories, right? If we know that some of the behaviors that we show are related to you having low socioeconomic status or to your ethnicity or to your sexual orientation, then you don't need to capture that category because it's like somewhere embedded in the traces that you leave.

And they might all be related to some of the protected categories, right? If we know that some of the behaviors that we show are related to you having low socioeconomic status or to your ethnicity or to your sexual orientation, then you don't need to capture that category because it's like somewhere embedded in the traces that you leave.

To some extent, I think on the global level, when we try and understand what are these models doing and are they potentially discriminating against people, I still think that there's something that we can actually do to probe. Oftentimes people say, well, we don't know what the models are doing because it's like these complicated neural nets and we just can't open a black box.

To some extent, I think on the global level, when we try and understand what are these models doing and are they potentially discriminating against people, I still think that there's something that we can actually do to probe. Oftentimes people say, well, we don't know what the models are doing because it's like these complicated neural nets and we just can't open a black box.

To some extent, I think on the global level, when we try and understand what are these models doing and are they potentially discriminating against people, I still think that there's something that we can actually do to probe. Oftentimes people say, well, we don't know what the models are doing because it's like these complicated neural nets and we just can't open a black box.

It can still look at the output. If you're thinking about, are we going to give people a loan or not? And you just see that none of the women are getting any loans and none of the women are getting hired into technical roles. Maybe then that's something that the model is picking up on. Right.

It can still look at the output. If you're thinking about, are we going to give people a loan or not? And you just see that none of the women are getting any loans and none of the women are getting hired into technical roles. Maybe then that's something that the model is picking up on. Right.

It can still look at the output. If you're thinking about, are we going to give people a loan or not? And you just see that none of the women are getting any loans and none of the women are getting hired into technical roles. Maybe then that's something that the model is picking up on. Right.

So even if you don't fully understand what it's doing, you can always look at the predictions and see, is there anything that we see among the categories or the social demographics that we want to protect? That seems to be off in terms of how often we do the thumbs up that the person gets the loan or gets the job.

So even if you don't fully understand what it's doing, you can always look at the predictions and see, is there anything that we see among the categories or the social demographics that we want to protect? That seems to be off in terms of how often we do the thumbs up that the person gets the loan or gets the job.

So even if you don't fully understand what it's doing, you can always look at the predictions and see, is there anything that we see among the categories or the social demographics that we want to protect? That seems to be off in terms of how often we do the thumbs up that the person gets the loan or gets the job.

Yeah, totally. And it identifies your different levels, right? So the example that you gave, like the three data points, that's coming from this notion of even if we anonymize data, but even if like I got all of the credit card spending from everybody in Manhattan and we say, but it's anonymized because we're not using any names, we're not using date of birth, we're not using an address.

Yeah, totally. And it identifies your different levels, right? So the example that you gave, like the three data points, that's coming from this notion of even if we anonymize data, but even if like I got all of the credit card spending from everybody in Manhattan and we say, but it's anonymized because we're not using any names, we're not using date of birth, we're not using an address.

Yeah, totally. And it identifies your different levels, right? So the example that you gave, like the three data points, that's coming from this notion of even if we anonymize data, but even if like I got all of the credit card spending from everybody in Manhattan and we say, but it's anonymized because we're not using any names, we're not using date of birth, we're not using an address.

Because your spending signature is so unique, right? Almost like a fingerprint. That it's very easy. If I know three things about you, I can just easily identify you in there. And then you're absolutely right. It's like if you think of identity at the next level, it's not just that I know, well, it's Jordan.