Return to Gender 101

[seebs]

Originally Posted by chunksmediocrites

I'd like you to go down to the duck pond where you see thousands of ducks. I want you to count the number of female ducks you see during an 8- hour period.

But that number doesn't mean anything, because you haven't counted the male ducks.

Did I get that right?

First off, not quite comparable simply because (most) ducks can be categorized as male or female. So we can sort of assume that from a population P, P = P(f) + P(m). This doesn't work for Bechdel(f) and Bechdel(m) populations of movies, where a given movie could in theory be in both categories or neither, as well as being in one or the other.

But.

Let's say I go down to the duck pond, and I count female ducks. And at the end of the day, I've counted 500. What do we know?

Nothing. Because we don't know how many total ducks I actually looked at enough to have counted them. If I only saw 750 ducks total, two thirds of what I saw were female. If I saw 3000 ducks total, one sixth were. Without that information, you have no idea how many ducks there were, or how many were female; all you know is that at least 500 female ducks were present, so presumably there were at least 500 ducks. Probably more.

If I count both, you can derive information about at least the relative frequency. It becomes possible to make interesting claims.

[beyelzu]

Originally Posted by chunksmediocrites

I'd like you to go down to the duck pond where you see thousands of ducks. I want you to count the number of female ducks you see during an 8- hour period.

But that number doesn't mean anything, because you haven't counted the male ducks.

Did I get that right?

No.

[chunksmediocrites]

The presenter in the videos discussed already the limited use of the Bechtel Test, and what it was and wasn't useful for. But I guess we could go further in really getting down to the root of the limitations of the Bechtel Test as not meeting the rigors of a true disciplined scientific examination of the problem. As ancillary non-issues go, it is quite a beaut.

[chunksmediocrites]

Originally Posted by seebs

First off, not quite comparable simply because (most) ducks can be categorized as male or female. So we can sort of assume that from a population P, P = P(f) + P(m). This doesn't work for Bechdel(f) and Bechdel(m) populations of movies, where a given movie could in theory be in both categories or neither, as well as being in one or the other.

But.

Let's say I go down to the duck pond, and I count female ducks. And at the end of the day, I've counted 500. What do we know?

Nothing. Because we don't know how many total ducks I actually looked at enough to have counted them. If I only saw 750 ducks total, two thirds of what I saw were female. If I saw 3000 ducks total, one sixth were. Without that information, you have no idea how many ducks there were, or how many were female; all you know is that at least 500 female ducks were present, so presumably there were at least 500 ducks. Probably more.

If I count both, you can derive information about at least the relative frequency. It becomes possible to make interesting claims.

Weird, here I was thinking we could determine the number of female ducks one saw in an 8-hour period. Especially if that was what the assignment was. But hey- no one checked the pH of the water- what the fuck? I can't tell a thing from your results.
ETA:

Originally Posted by seebs

it's just an intuitive appeal to emotion until it's got some kind of control group or other way of putting the results on a scale.

Extra credit on the intuitive emotion angle there. Why the test is nearly histrionic! Just the kind of thing one comes to expect from some genders!

[beyelzu]

Are you really this obtuse? If you purport that female ducks are underrepresented then the number of ducks in the population is not some sort of inane factoid.

It would in fact be an important piece of information.

[Qingdai]

At some point, usually in estimates of cultural valuation of some behavior, scientific or quantitative methods become untenable.
Then you have to rely on qualitative methods, the Bedchel test is a reasonable measure of "meaningful portrayal of women in mainstream cinema" using qualitative methods. A control for qualitative methods is usually triangulating your results. For example, maybe women's portrayal in cinema isn't culturally relevant to the status of women in general, so you could also look at the participation of women in other fields, say how many women are CEOS, then perhaps look at wages for doing the same work and compare.

Not all research methods can be the double blind, control group standard of scientific enquiry. There are, fortunately, other ways to study a problem.

[seebs]

Originally Posted by chunksmediocrites

Extra credit on the intuitive emotion angle there. Why the test is nearly histrionic! Just the kind of thing one comes to expect from some genders!

No, it's the kind of thing you expect from humans.

If you present humans with an argument in favor of a true conclusion, they consistently think the argument looks persuasive, whether or not the argument is actually persuasive.

To put it another way:

What do the results of the Bechdel test tell us? If the results were different, would it tell us something else?

[seebs]

Originally Posted by Qingdai

At some point, usually in estimates of cultural valuation of some behavior, scientific or quantitative methods become untenable.
Then you have to rely on qualitative methods, the Bedchel test is a reasonable measure of "meaningful portrayal of women in mainstream cinema" using qualitative methods. A control for qualitative methods is usually triangulating your results. For example, maybe women's portrayal in cinema isn't culturally relevant to the status of women in general, so you could also look at the participation of women in other fields, say how many women are CEOS, then perhaps look at wages for doing the same work and compare.

Not all research methods can be the double blind, control group standard of scientific enquiry. There are, fortunately, other ways to study a problem.

There's a whole lot of middle ground between totally objective double-blind tests and not enough information to justify any sort of conclusion.

One way to improve the value of results, even highly subjective or qualitative reuslts, would be to have more than one so that they could be compared. Do the same thing with men talking to each other about something other than a woman, and suddenly you have some kind of framework for expressing the idea that there is a disproportionate relationship. Do the same thing, but group results by decade, and you can show whether things are getting better or worse.

But with a single massed set of results, it's hard to say anything other than "intuitively, this feels like it's a bad result". Which is an okay starting point, except that human intuition about results is crap.

[Qingdai]

Yes, that's why I said triangulating. Do you read beyond the first sentence? You quoted the whole thing.

[Kael]

Originally Posted by seebs

We don't have any way of evaluating what the results mean.

That's the thing about this somewhat ridiculous debate regarding a hypothetical Reverse-Bechdel Test - We're not trying to derive and support some nebulous meaning for the results of applying the test to Academy-Award-nominated movies (for which we would probably need more information). We are saying it is the results themselves that are the problem, not what they might mean.

To rephrase, no one is saying that few films pass the Bechdel test for meaningful interaction between female roles and therefore blah blah blah. What is being said is that few films pass the Bechdel test for meaningful interaction between female roles. That's it. Full stop. Point made, discussion closed. Certainly there are interesting questions to ask from there, and interesting tests and data that might apply to those questions, but whatever they might be they are not and have never been what the Bechdel test is about.

[seebs]

Originally Posted by Kael

That's the thing about this somewhat ridiculous debate regarding a hypothetical Reverse-Bechdel Test - We're not trying to derive and support some nebulous meaning for the results of applying the test to Academy-Award-nominated movies (for which we would probably need more information). We are saying it is the results themselves that are the problem, not what they might mean.

That doesn't make sense, and I honestly don't believe it to be in any way true, or even a thing which could possibly be true.

The results can't be a problem unless they tell you something about the world that is bad. Otherwise, you would be no more interested in the number of AA-nominated movies which pass the test than the number of eighteens on a roulette wheel on a given day. It matters in some way. And that way has to do with the meaning.

To rephrase, no one is saying that few films pass the Bechdel test for meaningful interaction between female roles and therefore blah blah blah. What is being said is that few films pass the Bechdel test for meaningful interaction between female roles. That's it. Full stop. Point made, discussion closed. Certainly there are interesting questions to ask from there, and interesting tests and data that might apply to those questions, but whatever they might be they are not and have never been what the Bechdel test is about.

Except that's obviously not true, because if it were true, no one would be talking about it at all. There is no "point made" unless something has meaning. That you think a point has been made means you do think the result means something.

Thing is, concepts like "many" or "few" are nearly always relative. In particular, they are relative to expectations.

Imagine that you did a test exactly like the Bechdel test, only instead of "women" it was "MTF transwomen". No one would think this test was interesting or said anything. They wouldn't discuss it. That they discuss the Bechdel test tells us it does have meaning.

It seems to me that the meaning people are getting from it is something to the effect of "women are mostly supporting characters, with men in the more significant roles that drive the story". (This is certainly true, in any event, and seems to correlate with what the test is checking.)

And the thing is, it makes a point by appealing to our intuition that in general men and women should be reflected in reasonably comparable ways and numbers, and that they clearly aren't being so reflected. It's just that, when you get past the initial intuitive "wow, that looks wonky", it becomes useful to have multiple numbers so you can compare them to each other.

This is like seeing a survey of how many women are CEOs, and coming back with "well, we found three thousand", and stopping there without even asking the question "and how many CEOs are there, total?"

[lisarea]

I think the disconnect is twofold.

First, the Bechdel test is not supposed to be proof of the phenomenon. It's a pithy observation from a comic strip that people use to illustrate the severity of a problem that most people know exists. It's not meant as proof, but an illustration of a) how prevalent and severe the issue is, and b) how acclimated we are to it that we probably never noticed how marginalized women are in mainstream media.

Second, social phenomena are extremely difficult to quantify. In most cases, you can't quantify them accurately using readily available data. This is not the same as calling it intuitive or magical, BTW. Social phenomena are real things, and they could be quantified through analysis with infinite time and resources, which nobody has. As such, analyzing them does require a fair amount of fuzzy logic and guesswork. People who are uncomfortable with abstractions or speculation like to characterize fields that require it as being somehow fundamentally magical or absurd because of that, but the fact is just that they're simply too complex to be readily quantified and articulated. Complex systems are just like that.

Back to the Bechdel test specifically, all it really is or is intended to be is an illustration of the extent of an issue that most people are already aware of. You could, and some people have, applied a similar 'reverse Bechdel' to various media, and it should be easy (if tedious) enough to apply that to a specific subset of media, such as the movies nominated for best picture. I'm going to go out on a limb and say that if any of the nominated pictures would fail the reverse Bechdel, maybe The Help. Maybe not. I'd bet money that the rest of them pass it easily. But I'm not going to do the test because I don't want to watch all those movies, and I don't think it's important or useful. And the people who do go to the trouble to do it don't find it important or useful, either, because it's fairly obvious that the vast majority of general audience mainstream movies have two or more male characters who talk to each other about something other than a woman; and people asking them to do so seem to be making demands on their time to satisfy their own agendas and distract from their point.

[seebs]

Originally Posted by Qingdai

Yes, that's why I said triangulating. Do you read beyond the first sentence? You quoted the whole thing.

Yes, I did. But it's not triangulating to take a non-value and combine it with several results. This would be like "triangulating" where we take one directional reading from one location, and then we move to another location and report that we sighted to our target and it totally existed, and trying to use that to come up with a location.

I can't tell what's going on here. You guys are in general rational people whose opinions I respect. But you're trying to use a result when it cannot possibly contain information without some kind of comparison.

Either you've all suddenly gone insane, or you're using an inferred context for comparison which hasn't been made explicit. I'm guessing the latter, since that is pretty much always in place when people evalute things as "many" or "few".

In which case, just identify the context, make it explicit, and BOOM, you have useful information. It's still qualitative and all, but it's gone from "dimensionless nunber" to "actual data we could use in triangulating to try to find out whether there is a trend".

I feel like I'm being told "we weighed this, and it weighs sixty. That's heavy!" And then told that it is silly to ask for units, or compare to the weights of other similar things, because there's no need for that kind of scientific rigor when all we need to know is that sixty is heavy.

[Kael]

Okay, sorry, you're right there is a point being made, and I shouldn't have oversimplified, but that point is simply that so many films failing such a basic test for meaningful female interaction is a bad thing. That point can be drawn and supported without any further tests or comparisons with male interaction in the same body of films.

[seebs]

Originally Posted by lisarea

I think the disconnect is twofold.

First, the Bechdel test is not supposed to be proof of the phenomenon. It's a pithy observation from a comic strip that people use to illustrate the severity of a problem that most people know exists. It's not meant as proof, but an illustration of a) how prevalent and severe the issue is, and b) how acclimated we are to it that we probably never noticed how marginalized women are in mainstream media.

Ahh!

Thank you, that explains a lot.

As an illustration of a thing already known, it's really very effective.

Second, social phenomena are extremely difficult to quantify. In most cases, you can't quantify them accurately using readily available data. This is not the same as calling it intuitive or magical, BTW. Social phenomena are real things, and they could be quantified through analysis with infinite time and resources, which nobody has. As such, analyzing them does require a fair amount of fuzzy logic and guesswork. People who are uncomfortable with abstractions or speculation like to characterize fields that require it as being somehow fundamentally magical or absurd because of that, but the fact is just that they're simply too complex to be readily quantified and articulated. Complex systems are just like that.

Yes. I still dislike the existence of fuzzy things which are not kitties, but I grudgingly admit that there are no real alternatives on offer.

Back to the Bechdel test specifically, all it really is or is intended to be is an illustration of the extent of an issue that most people are already aware of.

See, the thing is.

To me, extents are a kind of a quantified thing. If I knew that 95% of movies passed a reverse Bechdel, and 10% of movies pass the original Bechdel, that would give me a feeling of extent. On the other hand, if it were 15% and 10%, it would also give me a feeling of extent, but I'd think the problem was a lot more minor.

If I only have one of those two numbers, though, I can't get any sense of extent. I don't have a frame.

Maybe we need a Baychdel test, which is in how many Michael Bay movies two men discuss something which is neither a woman nor an explosion.

And the people who do go to the trouble to do it don't find it important or useful, either, because it's fairly obvious that the vast majority of general audience mainstream movies have two or more male characters who talk to each other about something other than a woman; and people asking them to do so seem to be making demands on their time to satisfy their own agendas and distract from their point.

Obvious isn't the same thing as true. It seems likely, but part of the point of the test is that people who think intuitively that mainstream movies have lots of strong and independant roles for women may in fact be completely wrong.

I am not trying to distract from their point; I am trying to point out that they could go from "here is an intuitive illustration of a thing" to "here is a large gap which illustrates the extent of the problem" in an afternoon, probably. It seems to me it'd be easy just because, I suspect, most of these movies have enough easy examples to come to mind that you wouldn't have to watch them to have an example handy. Thinking about it, the only movie I've seen in ages that I couldn't say confidently passes a hypothetical reverse-bechdel is How Stella Got Her Groove Back, and that's because I can only think of one male character in it.

Heck, I'ma go make the list, just because I suspect there are other people whose trained response to context-free numbers is to disregard them, and I suspect having a number handy would mitigate that.

[seebs]

Originally Posted by Kael

Okay, sorry, you're right there is a point being made, and I shouldn't have oversimplified, but that point is simply that so many films failing such a basic test for meaningful female interaction is a bad thing.

That seems true.

That point can be drawn and supported without any further tests or comparisons with male interaction in the same body of films.

I guess. To me, it's not support until I have a comparison. Numbers-in-a-vaccum are not support for claims.

Consider the huge difference in impact between "275 children have drowned in buckets since 1984" and "275 children have drowned in buckets since 1984, and 93 people die in traffic accidents every day". The former sounds like a problem; the latter makes the former problem sound utterly insignificant (which, statistically, it is).

[seebs]

Okay, quick check: Out of the first 22 movies in my Netflix ratings history, 20 were definite passes of the reverse-Bechdel, and two I couldn't remember for sure but I think they were.

So I would say that, pending someone having better data to look at, the Bechdel test is showing something highly statistically significant about relative frequencies, and can't just be handwaved away with "but men would fail that test too".

http://www.seebs.net/log/articles/63...e-bechdel-test <-- the list, in case anyone wants to dispute it.

[Qingdai]

Originally Posted by seebs

Yes, I did. But it's not triangulating to take a non-value and combine it with several results. This would be like "triangulating" where we take one directional reading from one location, and then we move to another location and report that we sighted to our target and it totally existed, and trying to use that to come up with a location.

I can't tell what's going on here. You guys are in general rational people whose opinions I respect. But you're trying to use a result when it cannot possibly contain information without some kind of comparison.

Either you've all suddenly gone insane, or you're using an inferred context for comparison which hasn't been made explicit. I'm guessing the latter, since that is pretty much always in place when people evalute things as "many" or "few".

In which case, just identify the context, make it explicit, and BOOM, you have useful information. It's still qualitative and all, but it's gone from "dimensionless nunber" to "actual data we could use in triangulating to try to find out whether there is a trend".

I feel like I'm being told "we weighed this, and it weighs sixty. That's heavy!" And then told that it is silly to ask for units, or compare to the weights of other similar things, because there's no need for that kind of scientific rigor when all we need to know is that sixty is heavy.

Apparently you're using some specific engineering style definition of triangulating.

http://en.wikipedia.org/wiki/Triangu...social_science)

[lisarea]

Originally Posted by seebs

Yes. I still dislike the existence of fuzzy things which are not kitties, but I grudgingly admit that there are no real alternatives on offer.

The things themselves are not unquantifiable or fuzzy, though. Just our understanding of them. That's a pretty important distinction.

Pretty much any reasonably complex natural system is subject to and even dependent on a certain amount of guesswork and interpretation. Obviously, social sciences are, but so is medicine. So is artificial intelligence. Natural language processing, for example, relies very heavily on intuition because of the limitations of the articulable information we have about how language works. There are rules, but we don't know enough about them to explain it accurately, and the first and several of the intermediary steps to articulating them involve relying on your intuition about how they work. (Intuition isn't magic, either, it's actually a type of knowledge or understanding that you can't consciously articulate.) Just like there are rules to the way drugs interact with our bodies, but researchers discover medicinal uses and negative effects by observation, rather than by running calculations that accurately predict outcomes somehow. Because these systems are all too complex for us to articulate or predict at this point. In fact, I can't think of a major area of intellectual pursuit that doesn't, at some point, get weird and difficult to quantify.

People like to draw these big black lines between soft and hard sciences, as though there's some kind of fundamental difference between the topics themselves. The real difference--and even this one isn't super clear--is between things that we've figured out how to quantify consistently and those we haven't.

[erimir]

Originally Posted by seebs

To me, extents are a kind of a quantified thing. If I knew that 95% of movies passed a reverse Bechdel, and 10% of movies pass the original Bechdel, that would give me a feeling of extent. On the other hand, if it were 15% and 10%, it would also give me a feeling of extent, but I'd think the problem was a lot more minor.

This, pretty much.

I at least would certainly be interested in seeing how the numbers compare for some random selection of movies (perhaps every movie released in 2011, or some subset like movies released the first week of the month in 2011). I just think it would give more information. I don't dispute that men are much more represented, but I do like statistics.

[Ymir's blood]

Finally found/made time to watch the video. I'm glad that she's not obviously reading from cue cards or a teleprompter anymore.

I realized while watching it that Plan 9 almost certainly doesn't pass either. There are three female characters, but only two are named and none of them meet, IIRC.

[lisarea]

I kind of like this for calling out the kneejerk association so many people have that anything girls do is annoying and stupid. That vocal fry story was lousy with that.

Young Women Often Trendsetters in Vocal Patterns - NYTimes.com

[seebs]

Originally Posted by Qingdai

Apparently you're using some specific engineering style definition of triangulating.

http://en.wikipedia.org/wiki/Triangu...social_science)

Ahh, no. I was assuming it was being used informally.

It's just that... A data point without comparators does not actually give you information, and adding a whole bunch of data points with no comparators for them does not change that. Knowing the absolute number of female CEOs, without knowing how many CEOs there are in general, would tell you nothing. Adding multiple similar fragments of data points doesn't give you anything either. It's not until you have something to compare them to that they're data.

To put it another way:

Why does this test tell us something? Because, even if the comparison is implicit, we are comparing these results to something. Even if it's just an unconscious expectation.

The reason this tends to work out is that most of the time, unconscious expectations are basically reasonable enough that comparisons to them give some information... But if you want to avoid hilarious mishaps, it's a lot better to take the time to figure out what you're comparing with and make it explicit.

[seebs]

Originally Posted by lisarea

The things themselves are not unquantifiable or fuzzy, though. Just our understanding of them. That's a pretty important distinction.

Well, that's an interesting question: How do we know that these things are not actually unquantifiable or fuzzy?

It seems to me that it is entirely possible for the world to contain things which have fuzzy boundaries. Jumping back half a topic: Biological sex. We have nice clear categories for "male" and "female", and a couple-few million people who do not fit either of these categories exactly.

So if we have only unquantifiable understandings of a thing, I think it is worth considering the possibility that the thing itself is a little unquantifiable.

Pretty much any reasonably complex natural system is subject to and even dependent on a certain amount of guesswork and interpretation.

Certainly.

People like to draw these big black lines between soft and hard sciences, as though there's some kind of fundamental difference between the topics themselves. The real difference--and even this one isn't super clear--is between things that we've figured out how to quantify consistently and those we haven't.

Interesting. What's neat about this is that even if it turns out to be false, it's still an extremely useful way to think about the problem spaces.

But that leads me to think that it is probably often useful to see whether something can be firmed up and quantified a bit.

Consider, if you will, two paragraphs:

1. The Bechdel test is a test of how women are portrayed in media; it classifies movies as passing or failing on simple and easily-tested criteria. A majority of Hollywood movies fail the test. Obviously, this indicates a serious problem.

2. The Bechdel test is a test of how women are portrayed in media; it classifies movies as passing or failing on simple and easily-tested criteria. A majority of Hollywood movies fail the test. However, if you reverse the test and apply it to men, nearly all Hollywood movies pass the test. Obviously, this indicates a serious problem.

I would argue that 2 is unequivocally a stronger problem statement. It excludes a category of possible alternative explanations, and gives some framing for the results.

I would also argue that the only reason 1 is at all persuasive is that we intuitively recognize that 2 is probably true. (Conveniently, a quick test on a selection of movies not selected by an availability heuristic seems to confirm this.)

[seebs]

Originally Posted by erimir

This, pretty much.

I at least would certainly be interested in seeing how the numbers compare for some random selection of movies (perhaps every movie released in 2011, or some subset like movies released the first week of the month in 2011). I just think it would give more information. I don't dispute that men are much more represented, but I do like statistics.

I looked at the list of movies netflix thinks I've seen. Of the first 22, there were two where I'd have to watch them again to be sure, but I think the answer was yes, the other 20 were all easy and unambiguous "yes".

So the answer is that, whatever percentage of movies pass the Bechdel test (call it N%), it's an N%:100% comparison or very close to it. (I can't think of a second male character who has any non-trivial role in How Stella Got Her Groove Back. So there may exist a counterexample. That a counterexample leaps to mind as a movie that stood out suggests that we're in the 99%+ range.)