Hacker News Reader

Comment by jdp23 | Hacker News Reader

For feminist philosophers of science like Donna Haraway and Sandra Harding, it's not about rejecting objectivity, but rather about looking at how different perspectives influence what we mean by "objectivity".

Ssndra Harding discusses this in various places, such as her essay "What is Strong Objectivity".

For example, think about a medical study where subjects are primarily male. Do the results in fact represent "objective reality" for the general population?

Or, using your Bayesian example, consider fields that are disproportionately male, and there are gender differences in the observations o(X) that people make. Do the "facts" that are inferred represent "objective reality"?

View on HN

Your examples are not philosophical at all. Sampling error and parameterized observations are both well known problems in statistics.

At my company, we are putting various checks in to mitigate sampling error, e.g. flagging A/B tests as potentially erroneous when they terminate at non-integer numbers of weeks. Are we somehow doing feminist philosophy?

In your example of gender differences in observations, the correct solution is to parameterize the gender of the observer. I.e., you'd have o_m(X) and o_f(X). This is no different from parameterizing the time since a visitor hit a website in a conversion experiment - a visitor failing to convert after 3 seconds gives you different information than a visitor failing to convert after 3 days. (In this case you'd have o_t(X) instead of o(X).)

The facts that are inferred do, in fact, represent objective reality. They just may not provide a lot of information on it.

How many scientific papers parameterize the gender of the observer?

Does your own work do this?

My own work has never needed to, nor do most scientific papers. Most observations are made by machines. In my current projects they are made by javascript in the browser. In past experiments they were made by things like CCD cameras or MRIs.

I do parameterize other observer properties. For instance, the example of time parameterization is exactly what I'm working on today. Parameterizing space/time in an MRI is an area of active research and is vitally important in the big machines.

> Most observations are made by machines.

... by machines designed and calibrated by men, and interpreted by men.

> In my current projects they are made by javascript in the browser.

... made by code written by a man in a language designed by a man and implemented mostly by men on a platform designed by a man and implemented mostly by men.

So, if you don't acknowledge the possiblity that this could lead to observation biases, then no, it doesn't sound to me like you're doing feminist philosophy.

On the other hand the way you can simultaneously say "the correct solution is to parameterize the gender of the observer" and "my own work doesn't need to and neither do most scientific papers" certainly does seem to clash with the notion of a single "objective reality"!

... by machines designed and calibrated by men, and interpreted by men...made by code written by a man in a language designed by a man and implemented mostly by men on a platform designed by a man and implemented mostly by men.

Are you claiming that if my hit counter were written by a female, in a non-javascript language, then it would yield a different count? Or that if my CCD were designed by a woman, it would take different pictures?

I'm really trying to apply the principle of charity here but I'm failing horribly. So please - I'm shining a laser through a dielectric and imaging the output with a CCD. What bias does feminist philosophy think the gender of the experimentalist introduces? Do the pictures become dimmer? Is noise introduced?

To clarify my position: ordinary statistics has ways of handling observational bias (parameterizing it, averaging it out). I acknowledge that in some experiments this would affect things. I also believe that in the vast majority of experiments (e.g., shining lasers through materials, mixing chemicals together) that these effects are exceedingly implausible and can be safely ignored. Do you disagree with this claim?

> Are you claiming that if my hit counter were written by a female, in a non-javascript language, then it would yield a different count?

I'm claiming that if the experiment that involved a hit counter was designed and implemented by women, in a language designed and implemented by women, running on a platform designed and implemented by women, and the results were interpreted by a woman, it might well lead to a different conclusion about "objective reality".

And I'm suggesting that your unwillingness to acknowledge that possibilty is an indication that you're not doing feminist philosophy.

You seem to be moving the goalposts a lot. First you discussed sampling error and different observers, until I pointed out that these are the province of ordinary stats.

Now you are discussing whether social experiments in an alternate universe might come out different. Of course they might.

I don't claim to be doing feminist philosophy - given how rapidly the definition shifts, how could I possibly know if I am? I certainly do feminist philosophy as per your first definition (worrying about biased samples and observers), and I certainly don't do it as per your second (thinking about alternate universes).

You mean, I seem to you to be moving the goalposts a lot. Earlier in the thread you mentioned that you have a hard time understanding philosophy that isn't presented formally, so maybe some of what's going on here. So, get outside your comfort zone, and try taking a step back and rereading the discussion with the goal of trying to understand, rather than win a debate.

Are you suggesting women would calibrate CCDs a different way? Or that if Brandon Eich was a woman JS would be better or somehow "different" (given the same constraints)?

The idea that someones' chromosomes (or identified gender?) influence the outcome of machines seems like it'd need some rather massive evidence. Is there a short summary of this evidence? And why stop at XY vs XX? Why not include hair colour or other things?

Yes, I'm suggesting that languages designed by women -- and a software world where most languages, platforms, and tools -- would likely have different characteristics. Your assumption that it wouldn't seems to need just just as much evidence as the assumption that it would.

Reply on news.ycombinator.com