7 ms·

I’m a trans woman – here’s why algorithms scare me

8 points by amyjess 8y ago

maceurt 8y ago
An algorithm that trys and detect a male or female is not bad or unethical period. Gender != sex, males and females have different bodies and faces based on different genetic makeup between XX chromosomes and XY chromosomes.
- SketchySeaBeast 8y ago
  > If the researchers had consulted any trans people prior to beginning the project, they would know that many of us transition because we don’t want to be linked to our past name or appearance. We want to live our truth in the present and define our own future – not be algorithmically chained to false identities we were forced to wear in the past. The concern doesn't seem so much to about the existence of the algorithm, but instead it's uses.
dlivingston 8y ago
> At the University of North Carolina at Wilmington, researchers created a system that attempts to identify transgender people before and after they medically transition with hormone replacement therapy. To do this, researchers scoured YouTube for ‘transition timeline’ videos, which typically involve a series of photographs charting a person’s face changing over time. These videos are extremely personal, and a source of empowerment for the trans community... This isn't a trans issue, this is a 'digital consent' issue. As noted in the linked Verge article, "[t]he MegaFace dataset compiled by the University of Washington, for example, contains 4.7 million images of roughly 627,000 individuals — all taken from Flickr users." It is fair to say that many of those photos are deeply personal. So then, this raises a much larger question - not about digital ethics as applied to trans people, but digital ethics as applied to people: where are the lines drawn in fair-use for public content? Do you get to dictate what happens to your data that you upload to someone else's server publicly?
- crooked-v 8y ago
  I find it entirely reasonable to hold mass aggregation of content to different standards than use that's either individual or low-context.
cwkoss 8y ago
I think that the best solution to this problem is to ban gender-based discrimination by advertisers and data-aggregators. These algorithms only exist because marketers want to buy impressions from "25-34 year old men" If that was not a legal option, these algorithms would not have an economic incentive to exist. Surely "people who like power tools" as a demographic is much more targeted, both excluding "men who don't care about power tools" false positives and including "non-men who do" false negatives. A system of advertising that allows gender based targeting inherently reinforces gender stereotypes.
- dlivingston 8y ago
  Targeted ad placement is much more valuable for advertisers and those who profit from ad placement. Surely there are use-cases that warrant a sex-based discrimination: sports bras, as an example, or hair pomade, or gendered magazines (i.e. Men's Fitness; Seventeen).
  - cwkoss 8y ago
    You're certainly right that advertisers and advertising platforms have little to no financial incentive to mitigate the externalities of reinforcing stereotypes by segmenting demographics on the basis of protected class distinctions. However, I think as a society, we should discuss whether these externalities are worth the value that advertising provides to society.
natchiketa 8y ago
> These false ideas about gender have been thoroughly debunked by science. As much as I wish that were true, it's not. And not only does claiming it to be solid science not change anything, but also it weakens the standing of the person saying it. And by extension, the rest of us who live as we feel we need to live, but not at the expense of truth and intellectual honesty.
- eyelidlessness 8y ago
  Can you clarify which ideas are scientifically disputed?
  - belorn 8y ago
    I would guess they refer to the suggestion that gender is determined by the genitals that a person is born with, which the parent comment think is correct. Just like the article however, I agree that science has debunked that theory. Androgen insensitivity syndrome for example prevent the masculinization of male genitalia in the developing fetus with male phenotype. They are described by science as XY female, but does not have ovaries or a uterus. There is also studies done on neuroscience of sex differences, which found that transsexual individuals tend to have the brain characteristics and structures that do not match the genitals that the person was born with. Again a rather strong indication that gender and genitalia do not always match. There is quite a lot of research that indicate transgender as caused by biology, where individuals get a mix of both male and female traits, and genitals in those cases is simply one of many traits.
    - eyelidlessness 8y ago
      I gave you an upvote because this is all important information and likely helpful for other readers, but unless I've misunderstood the parent commenter I'm not sure this is what they meant. All of this information is consistent with the article author's claims, which the parent commenter disputed.
      - belorn 8y ago

8 points by amyjess 8y ago

An algorithm that trys and detect a male or female is not bad or unethical period. Gender != sex, males and females have different bodies and faces based on different genetic makeup between XX chromosomes and XY chromosomes.

SketchySeaBeast 8y ago
> If the researchers had consulted any trans people prior to beginning the project, they would know that many of us transition because we don’t want to be linked to our past name or appearance. We want to live our truth in the present and define our own future – not be algorithmically chained to false identities we were forced to wear in the past. The concern doesn't seem so much to about the existence of the algorithm, but instead it's uses.

dlivingston 8y ago

> At the University of North Carolina at Wilmington, researchers created a system that attempts to identify transgender people before and after they medically transition with hormone replacement therapy. To do this, researchers scoured YouTube for ‘transition timeline’ videos, which typically involve a series of photographs charting a person’s face changing over time. These videos are extremely personal, and a source of empowerment for the trans community... This isn't a trans issue, this is a 'digital consent' issue. As noted in the linked Verge article, "[t]he MegaFace dataset compiled by the University of Washington, for example, contains 4.7 million images of roughly 627,000 individuals — all taken from Flickr users." It is fair to say that many of those photos are deeply personal. So then, this raises a much larger question - not about digital ethics as applied to trans people, but digital ethics as applied to people: where are the lines drawn in fair-use for public content? Do you get to dictate what happens to your data that you upload to someone else's server publicly?

crooked-v 8y ago
I find it entirely reasonable to hold mass aggregation of content to different standards than use that's either individual or low-context.

cwkoss 8y ago

I think that the best solution to this problem is to ban gender-based discrimination by advertisers and data-aggregators. These algorithms only exist because marketers want to buy impressions from "25-34 year old men" If that was not a legal option, these algorithms would not have an economic incentive to exist. Surely "people who like power tools" as a demographic is much more targeted, both excluding "men who don't care about power tools" false positives and including "non-men who do" false negatives. A system of advertising that allows gender based targeting inherently reinforces gender stereotypes.

dlivingston 8y ago
Targeted ad placement is much more valuable for advertisers and those who profit from ad placement. Surely there are use-cases that warrant a sex-based discrimination: sports bras, as an example, or hair pomade, or gendered magazines (i.e. Men's Fitness; Seventeen).
- cwkoss 8y ago
  You're certainly right that advertisers and advertising platforms have little to no financial incentive to mitigate the externalities of reinforcing stereotypes by segmenting demographics on the basis of protected class distinctions. However, I think as a society, we should discuss whether these externalities are worth the value that advertising provides to society.

natchiketa 8y ago

> These false ideas about gender have been thoroughly debunked by science. As much as I wish that were true, it's not. And not only does claiming it to be solid science not change anything, but also it weakens the standing of the person saying it. And by extension, the rest of us who live as we feel we need to live, but not at the expense of truth and intellectual honesty.

eyelidlessness 8y ago
Can you clarify which ideas are scientifically disputed?
- belorn 8y ago
  I would guess they refer to the suggestion that gender is determined by the genitals that a person is born with, which the parent comment think is correct. Just like the article however, I agree that science has debunked that theory. Androgen insensitivity syndrome for example prevent the masculinization of male genitalia in the developing fetus with male phenotype. They are described by science as XY female, but does not have ovaries or a uterus. There is also studies done on neuroscience of sex differences, which found that transsexual individuals tend to have the brain characteristics and structures that do not match the genitals that the person was born with. Again a rather strong indication that gender and genitalia do not always match. There is quite a lot of research that indicate transgender as caused by biology, where individuals get a mix of both male and female traits, and genitals in those cases is simply one of many traits.
  - eyelidlessness 8y ago
    I gave you an upvote because this is all important information and likely helpful for other readers, but unless I've misunderstood the parent commenter I'm not sure this is what they meant. All of this information is consistent with the article author's claims, which the parent commenter disputed.
    - belorn 8y ago