Monday, October 8, 2018

Good intentions and bad outcomes

Ed Jong is a pretty good science writer for the The Atlantic magazine. Somewhat a victim of Atlantic's editorial bias towards "clickable" articles, but at core, a good journalist, writing well about interesting topics.

But like most in his field of journalism, there is a blindness to their own ideological assumptions and its implications. The field is steeped in social justice postmodernism, intersectionality, etc.

I didn't see it at the time but I am now reading I Spent Two Years Trying to Fix the Gender Imbalance in My Stories by Ed Yong from this past February.

I empathize with Yong's objectives - he wants to make sure he doesn't have any possible unconscious misogynistic prejudices that blinds him from quoting women scientists as often as he quotes male scientists. He goes back and examines his past articles. He uses an excel spreadsheet to track his on-gong performance. Good, rigorous approach to tackling a plausible problem. Yong is asking himself whether he has a hidden bias against women and finding ways to document whether that might be the case.

But is it a problem? Does a differential pattern in sourcing between male and female scientists reflect a bias on the part of Yong that he needs to correct or does it reflect a reality? We know that differential patterns of behavior are manifest everywhere. There are virtually no sub-groups which display perfect proportionality of the super group.

I refer to this as the social justice fractal assumption of sociology. Fractals demonstrate similar/identical patterns at any scale of analysis. Correspondingly, social justice jacobins assume that, for a given super population, if a group X has a representation of 10%, then for any sub-group of the super-group, then the proportion of group X also should be 10%. If Asian-Americans are 5% of the national population, then Asian-Americans should be 5% of all journalists (or firemen or farmers or Rotarians or elected officials or whatever other sub-group is of interest).

This social justice fractal assumption of sociology is paired with an assumption that any deviance from the fractal pattern of correspondence, any disparity, must be, and can only be, the result of conscious or unconscious malevolent prejudice on the part of an individual or an institution. Rounding out the trifecta of evil assumptions, social justice jacobins assume that any such observable disparity, being the result of bigoted prejudice, must be addressed by the state in an educational or coercive fashion, not infrequently via quotas.

If people's personal choices do not yield the fractal sociology which social justice jacobins want, then the jacobins want to force people to change their personal choices. That is naively rational if individuals are making evil decisions but completely destructive if free individuals are making good decisions which optimize their well-being across multiple dimensions of what is good.

Which is kind of a problem because in no country, at any time, ever, is there a perfect, or even close, fractal sociology. Virtually no sub-group is a fractal mirror of the supergroup on any of the possible variables which could be chosen.

Social justice jacobins are enamored of racial discrimination, gender discrimination, orientation discrimination, ethnic discrimination, national-origin discrimination. They throw in fat-shaming and slut-shaming on occasion just to mix it up. But why are these particular identities preferred? Or privileged, as one might claim? This is one of the things to which SJJs are blind. They fail to recognize that their hierarchy of privileges is in itself a prejudice, and a malevolent one at that.

What are other identities that are pertinent, often more pertinent in terms of how people choose to identify themselves? Marital status, age, familial structure, handedness, height, morbidity, religion, territorial locality, social status, educational status, ableness, mobility, affiliative ability, folkways, work habits, culture, time discounting preferences, self-control, etc.. The list is endless. All of these are explanatory variables for disparate outcomes.

If you look at the average household income for Hispanics and Japanese Americans and see that the former earn $40,000 while the latter earn $70,000 there is no logic that dictates that that differential is due to either negative or positive prejudice. Factors which materially influence household income include age (middle aged people earn more than younger people), education attainment (more years of education and degrees attained is correlated with higher income), morbidity (healthier people earn more than sicker people), urbanity (city people earn more than rural people), etc.

Virtually all the household income differentials between Japanese and Hispanic Americans are explained by these factors. Japanese households are, on average, older, more educated, more concentrated in cities, more concentrated in professions, have better health, etc. Bigotry has nothing to do with it.

But that is reality and does not impinge on the SJJ article of faith that all disparities between groups can only be explained by bigotry.

Another stumbling block for the SJJ trifecta of evil assumptions is that those countries with the greatest degrees of cultural, social and legal egalitarianism (Anglo-phone, Scandinavia, northwest Europe) between genders also demonstrate the highest levels of gender differentials in career choices and income outcomes.

SJJs are unconsciously imposing an explanation for disparities in a bigoted fashion and against the evidence.

Back to Yong. He musters a lot of indicators which support the idea that women in sciences are discriminated against. He fails to present the other side of the argument, the other studies which demonstrate that there is no discrimination, that women are preferentially selected in both the academy and in jobs, etc. Both sides of the argument are supported by a lot of research but taken together the research is broadly inconclusive. It is only conclusive if you select only those studies which support one side of the argument or the other.

Yong is testing himself to make sure he is not unconsciously excluding women scientists from his articles. Great goal.
Shortly after Adrienne published her analysis, I looked back at the pieces that I had published in 2016 thus far. Across all 23 of them, 24 percent of the quoted sources were women. And of those stories, 35 percent featured no female voices at all. That surprised me. I knew it wasn’t going to be 50 percent, but I didn’t think it would be that low, either. I knew that I care about equality, so I deluded myself into thinking that I wasn’t part of the problem. I assumed that my passive concern would be enough. Passive concern never is.

I’ve since been trying to actively redress the balance, by spending more time searching for women to interview. For any given story, I almost always try to contact several sources. If, for example, I’m writing about a new scientific paper, I will interview the scientists behind the work, but also pass the paper around to get comments from independent researchers. To find the right people, I’ll look at related work that’s cited by the paper in question. I’ll google for people who do similar research. I’ll check Twitter. I’ll look at past news stories. To find more female sources, I just spend a little more time on all of the above—ending the search only when I have a list that includes several women.

Crucially, I tracked how I was doing in a simple spreadsheet. I can’t overstate the importance of that: It is a vaccine against self-delusion. It prevents me from wrongly believing that all is well. I’ve been doing this for two years now. Four months after I started, the proportion of women who have a voice in my stories hit 50 percent, and has stayed roughly there ever since, varying between 42 and 61 percent from month to month. And of the 312 stories I’ve written in that two-year window, only 7 percent feature no female voices. (This figure excludes the small number of stories that feature no voices of any gender.)
What Yong doesn't address, indeed avoids, is what percentage is the right target? He says it won't be 50% but he thinks it should be more than 24% (his actual performance). What should it be? 25%, 30%, 35%, 40%? He doesn't say explicitly and that's kind of a crucial point. He seems happy that it is now 50% and that seems to be the acceptable level to which he has been striving.

If you are insisting on fractal sociology, you have to know the pattern. And obviously it is going to differ by field. If you have an article about forestry work and you are interviewing lumberjacks, it is no surprise if 99% of your quotes from lumberjacks come from male lumberjacks because it is a field that is virtually without women lumberjacks. Trying to ensure that 25% of your interviews come from female lumberjacks, or 50%, is obviously not correct, indeed misleading and possibly skewing.

The Department of Labor is a good place to start in terms of estimating the labor force pattern to guide what a representative base of interviews might look like. Here is one such study focusing on fields in which women are underrepresented.

But Yong is, presumably, not looking at the averages for the field in general. As a journalist, he is wanting to talk to the leading lights. There is patchy information on this, depending on the field. My research over the years covering medical, legal, accounting, government, authors and writing awards, movie directors, research scientists, academics, board memberships, etc. indicates that the broad parameters are that women constitute between 15-30% of the leadership or award winners in any given field.

Why not 50%? The elephant in the room is biology. Women take time off for pregnancies, many take longer breaks, some decide on long duration career interruptions, some never return to the workforce, some return part-time, some take the opportunity to change fields. They make those decisions, presumably in conjunction with their partner. They are influenced by their personal ambitions, the financial reward structure of their fields, the relative prospects of their field, their own personal values, social expectations, etc. A highly variable set of factors with highly variable weights. It is impossible to draw aggregate conclusions from highly heterogenous individual decisions.

Cutting edge performance, top positions, top awards, are all usually associated with distinctive accomplishments which are the result of long duration effort (many years), excessive effort (long hours), persistent effort (no breaks or interruptions).

The 15-30% makes a lot of sense if you are willing to acknowledge the heterogeneity of people's values and objectives in combination with the requirements needed to be among the top performers. Women can obviously be top performers in virtually every field. It is worth noting that gender disparities in patents, science papers, business founders, executives, journalists, author awards, etc. are the greatest in Scandinavia where the culture and laws are most empirically egalitarian.

In terms of proportionality, the US beats Sweden in virtually every field of endeavor in terms of having the highest number of top performers being women in a far wider range of fields.

Yong is trying to increase the number of women he quotes from? Why? 24% is solidly in the 15-30% range. What number should it be? If he is writing about theoretical physics, it would be close to 10% (not many top theoretical physicists are female). If it is biology or sociology, it is likely to be more like 40 or 45%. I cannot tell whether his 24% achieved number indicates that he is over or under representing women because it depends on the mix of fields about which he has been writing.

Given that he is a science writer (a set of fields with, in general, below average female participation) and making the assumption that he is writing about the leading performers in those fields (where female representation is de facto lower), we would expect that women would be at the lower end of the 15-30% range. I.e. Pretty much where he was at the beginning before he undertook this social justice experiment.

If indeed, women are, say 25% of the top performers in the fields about which he is writing, then what are we to conclude given that he now has 50% of his sources being women?

I think it can only be one of two things.
1) He is discriminating against men among the top performers. If 35% of the top performers in a field are women and he is quoting 50%, then he is actively discriminating against men with equal performance based solely on their gender. The Social Justice Jacobins are, potentially, behaving in exactly the bigoted fashion as the phantom strawman against whom they think they are fighting.

2) He is dipping deeper into the performance well in order to get his numbers. He samples the best performers and only gets 35% who are female. He then goes the next level down in terms of accomplishments and capabilities to get the extra 15% he is seeking. But now you are dealing with mixed populations, apples and oranges. Whatever research he is doing is compromised.

As a hypothetical: Assume you are doing research on the effectiveness of corporate training on managers. 35% of managers are women. You interview 100 manager participants. For your resulting story, you only have 35% women. So you go back and you interview some supervisors (next layer down) and maybe even some front line workers until you have your extra 15. But what you have now are, hypothetically 50 male managers, 35 female managers, 10 female supervisors, and 5 female front line workers.

If you find a difference in training effectiveness between men and women, is it because the training is poorly designed for women or is it because your sample for women is no longer a sample of women managers?
These are the type of issues that creep in when you deliberately oversample in order to hit a quota that is not representative of the population.

Yong acknowledges, though he downplays it, that he is investing extra time in order to hit his self-imposed 50% quota. So there is a productivity hit to him at a professional level.

And it is even worse than he highlights. He notes:
As the year went on, I found that I would need to contact around 1.3 men to get one male quote, and around 1.6 women to get one female one.
So not only are women materially less represented in his field than men, compared to the 50% quota, but women are ~20% less willing to be interviewed when he contacts them. So not only does he need to spend extra time finding names to make up his 50% quota but he also has to chase more of them in order to hit his mark.

It may make Yong feel good, and it may earn him kudos in a field dominated by social justice thinking, but as a journalist is he doing his readers any favors and is he even doing women scientists any favors?

Readers are left to conclude that either they are getting biased reporting (he kept the quality standard but discriminated against a class of people in favor of another class of people) OR that they are getting misrepresented and subpar reporting (he had to go to the second and third tier of performers to hit his quota.)

Women scientists are similarly negatively affected. If they are part of a quota then their expertise is discounted even though they might be the experts. The reader can't tell whether they are getting the equal quality expert or the filler. If the woman is a filler, then she is being held to a standard of competence which she might not yet have achieved, and that can cut two ways - it either opens opportunities or she can suffer a setback from misplaced expectations.

Internally, within their field of endeavor, women are being privileged by journalists, and their male peers may resent the privileging with possible, but uncertain consequences. The women themselves, may end up wasting time on journalists that would have been more beneficially expended on field activities.

The social justice idealist assumes they are doing right by the target of their sympathy, but more often than not they can create bad outcomes. (See Doing Bad by Doing Good for the dynamic in economic development).

I have seen this in my career in management consulting. In the mid-1980s there was a concerted push to make it easier for new mothers to stay in the field. Management consulting involves frequent travel, unpredictable travel, long hours and unpredictable hours. Career results are a cumulation of typically 10-15 years sustained performance. Not a set of factors obviously compatible with personal pregnancy and reasonable expectations of time invested in infants and young children.

We were losing too many young women to reasonable personal objectives of family and parenthood. Our early efforts were ham-fisted. OK, we'll reduce the travel burdens on pregnant and young mothers. Except the travel requirements did not go away. Male consultants and single female consultants had to take up the slack which set up all sorts of undesirable team dynamics.

We got more and more sophisticated over the next fifteen years, primarily through trial and error. We ended up striking a sort of median where we were providing career support, preferences and benefits to women which we did not provide to men but not so many that it became self-defeating. But one of the noticeable unintended and undesirable side-effects by the mid- or late nineties was the differential pattern of public speaking requests for our male and female partners.

Male partners tended to be invited to speak at industry or sector events - "Supply Chain trends in the Utility Industry", for example. These type of speaking events aren't especially great at driving new business but they do to a degree. As a profession, it was something you did between managing the business, managing people and serving clients.

On the other hand, our women partners were getting invitations to speak on things such as "Women Executives in the Tech Sector" or "Managing work life balance in consulting". Speaking gigs that valued them as women rather than as experts. These type of speaking events were far less likely to drive business. And driving new business is an important partner role which in turn drives partner compensation.

Women partners, if they wanted to do speeches, had the opportunity to do so, to a greater extent than their male counterparts. But if they did so, they reduced the amount of time on activities for which they were rewarded such as managing the business, managing people and serving clients. And on an activity that wouldn't lead to new clients.

It was, in the scheme of things, a very minor problem that they were more than capable of managing themselves, but it was a notable pattern of how good intentions can lead to potentially negative outcomes.

Consciously choosing to discriminate against people based on their race or gender or religion or any other non-pertinent factor is bad - whether done by an ignorant bigot, or by a well-intended social justice-pursuing journalist. But we now have a lot of advocates asserting that discrimination is just fine so long as the discrimination is done against the right targets.

Yong is a good writer. I hope he thinks this through and ditches the quotas.

No comments:

Post a Comment