> for example, how many of them have been sexually harassed by a man? Some of them have never had that experience, others have had it dozens of times.
How does knowing that 80% of women will be sexually harassed in their life substitute for having had that experience? Speaking for myself, a man, I've been sexually harassed before (not in the workplace), and it's absolutely not an enjoyable experience, but my qualia, and the qualia a woman who is sexually harassed has are still going to be different.
And yes not every woman has the same experience. That's obvious. But who is more likely to have qualia that are more representative of woman's experience? A man who studies women, or a random woman? There is indeed the place for expertise (and similarly: data), but neither expertise nor diversity is a substitute for the other.
> I'm not saying everything can be gotten vicariously through research
Qualia cannot be. If you entirely discount the value that qualia have, that may not matter to you, but there are good reasons to believe that that's a bad idea.
> I've seen a decent number of accounts of people who belong to some group, complaining that progressives assume they speak for everyone in that group
Indeed, but this isn't unique to progressives. It's just a problem that minorities have to deal with. And that's absolutely not a good thing (and awareness of this is good!). But again, it's not a progressive attribute (or one caused by diversity initiatives) to assume that the "other", whatever group it is, is cohesive in ways that the groups you are familiar with aren't. (e.g. When a member of a minority group commits a crime, you have reactions and narratives that minority crime is on the rise, or that this specific minority are criminals etc.)
A great way to fix this, by the way, is to interact with various people who are members of the other group, and see them disagree and debate. This can only really happen if you have a diverse enough group that you have multiple people of whatever minority to interact with. And this is good for everyone! It broadens majority perspectives and reduces microaggressions.
> And then the question becomes, which specific things do you want, and what's the best way to get them? I think the diversity discussion rarely gets that far—and if it did, strategies would end up very different.
I'm not really sure what you're saying here. Whose goals'? The business? The employees? The minority's? To explain why I'm dubious of the whole "data can solve the problem", someone has to drive the research to gather the data. By and large, if you want this to be done reasonably, you need to have input from the group under discussion. There's tons of hilarious-if-they-weren't-horrible examples of this if you're willing to look. The ML community is full of them (Gender Shades is perhaps the seminal example).
That is, with relatively few exceptions, for research to be done correctly, or in many cases, to be done at all, you need someone who is like the people being researched in important ways to be able to champion, contextualize and guide the research. Without that, a culturally unaware researcher is, if history is a guide, more likely than not to misconstrue cultural signals or make harmful and long lasting mistakes.
How does knowing that 80% of women will be sexually harassed in their life substitute for having had that experience? Speaking for myself, a man, I've been sexually harassed before (not in the workplace), and it's absolutely not an enjoyable experience, but my qualia, and the qualia a woman who is sexually harassed has are still going to be different.
And yes not every woman has the same experience. That's obvious. But who is more likely to have qualia that are more representative of woman's experience? A man who studies women, or a random woman? There is indeed the place for expertise (and similarly: data), but neither expertise nor diversity is a substitute for the other.
> I'm not saying everything can be gotten vicariously through research
Qualia cannot be. If you entirely discount the value that qualia have, that may not matter to you, but there are good reasons to believe that that's a bad idea.
> I've seen a decent number of accounts of people who belong to some group, complaining that progressives assume they speak for everyone in that group
Indeed, but this isn't unique to progressives. It's just a problem that minorities have to deal with. And that's absolutely not a good thing (and awareness of this is good!). But again, it's not a progressive attribute (or one caused by diversity initiatives) to assume that the "other", whatever group it is, is cohesive in ways that the groups you are familiar with aren't. (e.g. When a member of a minority group commits a crime, you have reactions and narratives that minority crime is on the rise, or that this specific minority are criminals etc.)
A great way to fix this, by the way, is to interact with various people who are members of the other group, and see them disagree and debate. This can only really happen if you have a diverse enough group that you have multiple people of whatever minority to interact with. And this is good for everyone! It broadens majority perspectives and reduces microaggressions.
> And then the question becomes, which specific things do you want, and what's the best way to get them? I think the diversity discussion rarely gets that far—and if it did, strategies would end up very different.
I'm not really sure what you're saying here. Whose goals'? The business? The employees? The minority's? To explain why I'm dubious of the whole "data can solve the problem", someone has to drive the research to gather the data. By and large, if you want this to be done reasonably, you need to have input from the group under discussion. There's tons of hilarious-if-they-weren't-horrible examples of this if you're willing to look. The ML community is full of them (Gender Shades is perhaps the seminal example).
That is, with relatively few exceptions, for research to be done correctly, or in many cases, to be done at all, you need someone who is like the people being researched in important ways to be able to champion, contextualize and guide the research. Without that, a culturally unaware researcher is, if history is a guide, more likely than not to misconstrue cultural signals or make harmful and long lasting mistakes.