On New Year's Day, I was brought back to this lesson for the first time since my college days. In the aftermath of my post about (non) resolutions, a friend shared with me one of his (non) resolutions for the coming year- 'read more books by women'. (2)
Interesting. My friend reads a lot of excellent books. His recommendations have led me to some of my favorite reads. At no point did it occur to me that his selections were biased. If anything, he seemed biased against rubbish literature. But bias against women?
And yet, there it was, clear as day, in the form of a (non) resolution, a suggestion about how to do better in the coming year. What the idea suggests is that, intended or otherwise, any selection process is prone to bias. And as John Snow proved in 1854, drawing from a contaminated source is the fastest way to an undesired outcome.
So (even though I knew I was a perfectly unbiased human being) I decided to check my own book selection process for any signs of bias. At the very minimum, by going through this entirely academic exercise I would have evidence in the future to wave in anyone's face- check this out, there's no 'i' in bias, at least as it applies to me...
My first step was to measure my own performance. Did anything in my reading history suggest bias in my book selections?
In 2016, I read one hundred and twenty-nine books. The breakdown was seventy-five men to fifty-eight women, a 56% male author proportion. Not proof of unbiased selection but I suppose it could have been worse (like, 57%). (3)
Next, I looked at my
Still, 'not bad' is sometimes a way of saying 'not good'. I was not quite at a male chauvinist level but there was an obvious imbalance. So, I considered another question. Could I explain away some or all of my unbalanced reading by identifying bias in my recommendation sources? (4)
Now, I was not out to accuse anyone who recommended a book to me of bias. And I was not out to find the evidence required to pompously make this accusation. But I considered the possibility because almost all my reading starts with a recommendation from one source or another. If I was pulling books off biased shelves, I would likely see these biases reflected in my end of year reading list. (5)
First, I looked at the two blogs I read most frequently. Lacking the energy to catalog each book mentioned in 2016, I instead opted for a proxy measure. Each blog published a 'books of the year' post similar to my own and I counted the authors they cited to represent the 'bias' shown by each blog.
Blog A did pretty well- ten women, six men. Blog B was a, umm, a tiny bit, well, just a little less, er, balanced- fourteen men, two women.
Next, I reviewed the websites I use to preview books. Again, I employed a similar technique to create a proxy measure.
Amazon split their books of the year lists into "editor's picks" and "reader's picks". The editors came out to a 50% split while the readers did a little worse, choosing titles with male authors for just over thirteen out of twenty books (I believe there were a couple of titles with two authors).
I browsed Goodreads but found that their lists drilled down into very specific genres. Given the small sizes of these lists, I opted to skip measuring this website.
Finally, I popped over to the Boston Public Library's main branch in Copley. Of their most borrowed books in 2016, six of the top ten were written by women.
SUMMARY
Blog A: 6 / 16
Blog B: 14 / 16
Amazon editors: 10 / 20
Amazon readers: 13 / 20
Boston Public Library: 4 / 10
Total: 47 / 82
Blog A: 37.5%
Blog B: 87.5%
Amazon editors: 50%
Amazon readers: 65%
Boston Public Library: 40%
Total: 57.3%I acknowledge that there are some glaring design problems with the above and some omissions on my part in terms of checking all my sources. Still, it was interesting to me that the proportion of my proxy measures worked out pretty close to my own observed rate- above parity. (6)
Those numbers don't lead to firm conclusions. I suppose a couple of the totals skew one way or the other but, as I learned in statistics classes, sample size is always relevant. Without gathering a lot more data, I don't really have the ability to make a conclusion.
So, my data brought me to a final question, which takes us back to my initial interest in my friend's (non) resolution. Are my friends and family biased in the way they recommend books to me?
Answering this question, in a way, mirrored the approach I took above. I did not ask anyone for their reading lists, opting instead to allow my intuition to guide me (my hunch- no bias). I then thought about potential sources they may have used to get their own initial notice about a possible read.
I found myself stuck at this point. The basic issue was figuring out how to aggregate the way all the unrelated individuals in my life first learn about the books they eventually read.
After some thought, I concluded that looking at the industry at the broadest possible level was the best (and perhaps safe) bet. Marketing is the starting point for almost everything in this consumerist society of ours. (7)
So, I did a little bit of research on the state of the publishing field. My most interesting finding came via The Stella Prize, an Australian organization that champions cultural change through its support of women's writing. This group found that women make up approximately two-thirds of the Australian author population.
So, in a world without bias, this implies that marketing efforts would allocate two-thirds of its resources to books authored by women. And yet, in a 2015 survey, this group found that twelve of thirteen publications it studied reviewed a higher proportion of male authors than it did female. This chart shows the five-year trend for nine of these publications. (8)
Now, proxy measures are always a problematic problem solving tool. In this case, I am using Australian publishing data to suggest that my American family and friends are exposed to book reviews championing the work of male writers over the work of female writers. This method would not hold up in the academic world, I'm guessing.
There are obvious problems beyond geography, as well. I have no clue whether a given book reviewer knows the gender of the author under review. And as always, who is to say if the imbalance is a product of pure chance?
On the other hand, it's not hard to find other observations of the same phenomenon. Take this Guardian article, for example, which cites a study by the US-based Vida confirming a similar general pattern in the US and UK book markets.
Again, I'm not making any firm conclusions with the data I've presented above. If at any point a blog I read, a website I browse, or a friend I trust has looked at a book and said 'you know what, I was just about to read this, but look at that first name, sounds girly...' I remain unaware of it (and ready to react with righteous shock were I ever informed of such an incident).
And though I'm open to examining my process and studying the flaws, it does seem like my 56% male author percentage from last year would rank pretty well, at least in Australia (assuming The Stella Group would rank my reading list alongside its actual book reviewing publications, the odds of such a thing being equivalent to zero). (9)
And yet, something's there. People at organizations like The Stella Group or Vida are devoting their entire professional lives to the cause of championing women's work in literature. Their work identifies positive role models for girls and ensures the display cases of our stores and libraries reflect merit and accomplishment. Such mission based organizations don't just pop up because of one or two unfounded accusations about bias in the world of publishing. (10)
From my own limited experience, I think the bias we observe in hindsight results from processes that stick too long with what one knows, lack imagination in defining requirements, or utilize criteria that contain their own bias. The malicious intent of discrimination isn't there. But the comfort zone of routine makes it difficult to pinpoint how the process is influencing selections. (11)
I think this helps explain what's happened to my reading list. I, like many, rely on publications to review new books. I lookup award-winners to point me toward the forgotten classics. But given the skewed proportions highlighted by organizations like The Stella Group or my understanding of how gender bias in the past prevented many women from being published at the same rate as men, my quality-seeking approaches are sure to lead me somewhat astray in living up to some of my big-talk about using merit to choose my next read.
But in the grand scheme of things, 56% isn't too bad. It's not 50%, though, so I'll keep my eye on the metric in the coming years. If this proportion drifts upward, I'll consider some adjustments to my book selection process.
For now, a better use of my time will involve continuing to honestly assess my own process as frequently as needed. Otherwise, I'm sure to fall into biased behavior as the unintended result of seeking comfort, familiarity, or routine.
That said, it might not hurt to have a second look at those blogs I read.
Thanks for reading.
Tim
Footnotes / imagined complaints
1. Fine...
No need to suggest I am making a biased recommendation here.
2. Non-resolution is now a thing? Like humblebrag?
No hashtags, please.
My basic rule of thumb is that any statement without a specific plan of action is a non-resolution.
3. Math whiz gets four more authors than books!!!
I did need to make some design concessions here. For authors I read more than once, I counted them each time a book of theirs appeared on my list. For books co-authored, I counted each author once. I did not include authors of picture or children's books.
If I did this over, I would probably just count each author once regardless of total books read. Is this convenient preparation for next year's post when I count the twelve George Saunders books as one male author?You bet your ass it is!
4. AKA...
*'The Passing The Buck' Stage
*'Don't Blame Me, I'm a Product of My Time' Stage
*'Tim is Officially Out Of Ideas' Stage
5. But my library only stocks male authors! How can I be blamed for my 100% male reading list?
I think in the short-term, its OK to read the books available to you. But eventually, the answer shifts. Go find a better library, man! At some point, everyone goes from 'helpless victim' to 'needing their suffering'.
6. What about Porter Square Books? Or the Cambridge Library?
I considered checking other branches or bookstores but thought it might be overkill. I browse the shelves at places like Porter Square Books, the Harvard Book Store, or the Cambridge Main Library about once or twice a year. So, including these as 'recommendation sources' alongside those I use once or twice a month did not seem like a very robust method.
7. Talk about a method only a goldfish would love....
For books published long ago, I assumed that any change in gender bias since the publication date would reflect the larger societal movement toward equality.
In other words, if I found evidence of bias in today's landscape, I would assume that things were at least just as biased in the past.
8. This is kind of beside the point, but...
The survey also cited that male reviews tended to prefer reviewing male authors while female reviewers tended to review male and female authors with near-equal frequency.
Have a look at the full survey here.
9. Relatively speaking is all well and good...
Of course, when it comes to questions of bias, being 'relatively' less biased is like having a bone that is 'less broken' than your ER neighbor's fracture. I'm either biased or I'm not. So, though I might look good by comparison to my, er, 'colleagues' Down Under, it's a completely irrelevant comparison to make. What's important here is doing the best I can with what I have control over.
10. This is not actually related, though it seems like it is...
It's an interesting sign that my Google searches for 'bias against women' returns 2.3 million hits compared to the 1.1 million hits I get for 'bias against men'.
11. I spent an hour and half trying to decide if this was a footnote, its own blog, or junk...
I'm always intrigued by findings that note an example of how good intentions lead to unintended results. A common one involves the admissions process for highly selective schools. Most schools do not use the homogeneity in their student populations as a selling point to prospective students and yet somehow 'balanced' admission processes around the country are producing that exact outcome. Even if the only 'discrimination' in the selection process is against inferior applicants, the accepted method of filtering prospective students by standardized test scores, grouping them into intelligence ranges by using criteria like admissions essays, and breaking any ties by using something like a legacy preference might lead to outcomes consistent with what a 'biased' process would produce.
The underlying wealth-based factors that impact things like primary school performance or the correlation of education level to wealth makes me suspect that the college application process is just as likely to measure a prospective student's surroundings as much as their ability.
On the other hand, it's not hard to find other observations of the same phenomenon. Take this Guardian article, for example, which cites a study by the US-based Vida confirming a similar general pattern in the US and UK book markets.
Again, I'm not making any firm conclusions with the data I've presented above. If at any point a blog I read, a website I browse, or a friend I trust has looked at a book and said 'you know what, I was just about to read this, but look at that first name, sounds girly...' I remain unaware of it (and ready to react with righteous shock were I ever informed of such an incident).
And though I'm open to examining my process and studying the flaws, it does seem like my 56% male author percentage from last year would rank pretty well, at least in Australia (assuming The Stella Group would rank my reading list alongside its actual book reviewing publications, the odds of such a thing being equivalent to zero). (9)
And yet, something's there. People at organizations like The Stella Group or Vida are devoting their entire professional lives to the cause of championing women's work in literature. Their work identifies positive role models for girls and ensures the display cases of our stores and libraries reflect merit and accomplishment. Such mission based organizations don't just pop up because of one or two unfounded accusations about bias in the world of publishing. (10)
From my own limited experience, I think the bias we observe in hindsight results from processes that stick too long with what one knows, lack imagination in defining requirements, or utilize criteria that contain their own bias. The malicious intent of discrimination isn't there. But the comfort zone of routine makes it difficult to pinpoint how the process is influencing selections. (11)
I think this helps explain what's happened to my reading list. I, like many, rely on publications to review new books. I lookup award-winners to point me toward the forgotten classics. But given the skewed proportions highlighted by organizations like The Stella Group or my understanding of how gender bias in the past prevented many women from being published at the same rate as men, my quality-seeking approaches are sure to lead me somewhat astray in living up to some of my big-talk about using merit to choose my next read.
But in the grand scheme of things, 56% isn't too bad. It's not 50%, though, so I'll keep my eye on the metric in the coming years. If this proportion drifts upward, I'll consider some adjustments to my book selection process.
For now, a better use of my time will involve continuing to honestly assess my own process as frequently as needed. Otherwise, I'm sure to fall into biased behavior as the unintended result of seeking comfort, familiarity, or routine.
That said, it might not hurt to have a second look at those blogs I read.
Thanks for reading.
Tim
Footnotes / imagined complaints
1. Fine...
No need to suggest I am making a biased recommendation here.
2. Non-resolution is now a thing? Like humblebrag?
No hashtags, please.
My basic rule of thumb is that any statement without a specific plan of action is a non-resolution.
3. Math whiz gets four more authors than books!!!
I did need to make some design concessions here. For authors I read more than once, I counted them each time a book of theirs appeared on my list. For books co-authored, I counted each author once. I did not include authors of picture or children's books.
If I did this over, I would probably just count each author once regardless of total books read. Is this convenient preparation for next year's post when I count the twelve George Saunders books as one male author?
4. AKA...
*'The Passing The Buck' Stage
*'Don't Blame Me, I'm a Product of My Time' Stage
*'Tim is Officially Out Of Ideas' Stage
5. But my library only stocks male authors! How can I be blamed for my 100% male reading list?
I think in the short-term, its OK to read the books available to you. But eventually, the answer shifts. Go find a better library, man! At some point, everyone goes from 'helpless victim' to 'needing their suffering'.
6. What about Porter Square Books? Or the Cambridge Library?
I considered checking other branches or bookstores but thought it might be overkill. I browse the shelves at places like Porter Square Books, the Harvard Book Store, or the Cambridge Main Library about once or twice a year. So, including these as 'recommendation sources' alongside those I use once or twice a month did not seem like a very robust method.
7. Talk about a method only a goldfish would love....
For books published long ago, I assumed that any change in gender bias since the publication date would reflect the larger societal movement toward equality.
In other words, if I found evidence of bias in today's landscape, I would assume that things were at least just as biased in the past.
8. This is kind of beside the point, but...
The survey also cited that male reviews tended to prefer reviewing male authors while female reviewers tended to review male and female authors with near-equal frequency.
Have a look at the full survey here.
9. Relatively speaking is all well and good...
Of course, when it comes to questions of bias, being 'relatively' less biased is like having a bone that is 'less broken' than your ER neighbor's fracture. I'm either biased or I'm not. So, though I might look good by comparison to my, er, 'colleagues' Down Under, it's a completely irrelevant comparison to make. What's important here is doing the best I can with what I have control over.
10. This is not actually related, though it seems like it is...
It's an interesting sign that my Google searches for 'bias against women' returns 2.3 million hits compared to the 1.1 million hits I get for 'bias against men'.
11. I spent an hour and half trying to decide if this was a footnote, its own blog, or junk...
I'm always intrigued by findings that note an example of how good intentions lead to unintended results. A common one involves the admissions process for highly selective schools. Most schools do not use the homogeneity in their student populations as a selling point to prospective students and yet somehow 'balanced' admission processes around the country are producing that exact outcome. Even if the only 'discrimination' in the selection process is against inferior applicants, the accepted method of filtering prospective students by standardized test scores, grouping them into intelligence ranges by using criteria like admissions essays, and breaking any ties by using something like a legacy preference might lead to outcomes consistent with what a 'biased' process would produce.
The underlying wealth-based factors that impact things like primary school performance or the correlation of education level to wealth makes me suspect that the college application process is just as likely to measure a prospective student's surroundings as much as their ability.
