If you want data to understand whether the average woman who participate on Lesswrong are subject to substantial sexual harrasment the lesswrong data is okay. To the extend that we think about modifying how we talk about certain issues on Lesswrong that’s the demographic that we care about.
Having the question in the lesswrong data set also allows us to see whether the answer to the question correlates with other answers on the survey.
To the extend that we think about modifying how we talk about certain issues on Lesswrong that’s the demographic that we care about.
When we talk about these things, it’s most often in the context of potentially driving away demographics whose representatives might offer underrepresented insights or perspectives. Sampling from a set self-selected to not have been driven away yet isn’t going to give us the data we want.
When that’s not the context, we’re usually talking about issues depending on the general population, and the pitfalls of using LW data for that are obvious.
When that’s not the context, we’re usually talking about issues relevant to the general population, and the pitfalls of using LW data for that are obvious.
I don’t think we only care about the general population. We care about the people with whom we are interacting on a daily basis. We have a bunch of people in this community who want spend time with rational friends instead of spending time with an average member of society.
Even if we are not intending with rational people we are still unlikely to interact with the average person.
Most woman I meet, I meet during Salsa dancing. That activity selects for woman who are okay with strangers physically touching them during Salsa dancing.
Skrewing depends on the purpose of your data.
If you want data to understand whether the average woman who participate on Lesswrong are subject to substantial sexual harrasment the lesswrong data is okay. To the extend that we think about modifying how we talk about certain issues on Lesswrong that’s the demographic that we care about.
Having the question in the lesswrong data set also allows us to see whether the answer to the question correlates with other answers on the survey.
When we talk about these things, it’s most often in the context of potentially driving away demographics whose representatives might offer underrepresented insights or perspectives. Sampling from a set self-selected to not have been driven away yet isn’t going to give us the data we want.
When that’s not the context, we’re usually talking about issues depending on the general population, and the pitfalls of using LW data for that are obvious.
I don’t think we only care about the general population. We care about the people with whom we are interacting on a daily basis. We have a bunch of people in this community who want spend time with rational friends instead of spending time with an average member of society.
Even if we are not intending with rational people we are still unlikely to interact with the average person. Most woman I meet, I meet during Salsa dancing. That activity selects for woman who are okay with strangers physically touching them during Salsa dancing.