Monday, October 21, 2013

Raising the Bar in Physics Graduate Education

By Meg Urry, Yale University (Department of Physics and Department of Astronomy) 

The following is adapted from a keynote address given at the APS Conference on Graduate Education in January, 2013.  

Reproduced from the June 2013 Issue of Status: A Report on Women in Astronomy



I am pleased to be addressing (and attending) this conference and I also know this audience is deeply committed to graduate education, so you probably don’t need to hear what I am going to say. Nonetheless, I thought a keynote address should be provocative, so I’ve done my best to push some buttons...

The invitation to speak tonight came shortly after the election last November. Front and center in the news was the Republican party’s concern about the shifting demographics in the United States: talking heads and columnists described the vanishing white male, the increasing diversity of the American population, and the sense that modern political parties have to adjust accordingly.

For example, here are some typical excerpts from an article by Michael D. Shear, NY Times, Nov 7, 2012:
“Before, we thought it was an important issue, improving demographically,” said Al Cardenas, the chairman of the American Conservative Union. “Now, we know it’s an essential issue. You have to ignore reality not to deal with this issue. 
The Republican Party “needs messages and policies that appeal to a broader audience,” said Mark McKinnon, a former strategist for George W. Bush. “This election proved that trying to expand a shrinking base ain’t going to cut it. It’s time to put some compassion back in conservatism. The party needs more tolerance, more diversity and a deeper appreciation for the concerns of the middle class. 
Tom Davis, who used to represent Dale City as a Republican member of Congress, said that the problem for his former colleagues goes beyond just Hispanic outreach. ... “It is time to sit down practically and say where are we going to add pieces to our coalition,” he said. “There just are not enough middle-aged white guys that we can scrape together to win. There’s just not enough of them.” [my emphasis]
The point, in cased you missed it, is that physicists can’t get away with educating only white men. This has been clear for quite a while. Since 2001 (despite homeland security issues) there have been more foreign citizens in our PhD programs than US citizens. We’ve admitted them to keep quality high, to keep our physics programs strong. Of course, these foreign students are often European or Asian men, so in some sense, our embrace of diversity has not changed the face of physics very much. For graduate physics education in the 21st century, just like the Republican party, we will have to expand our “big tent” to include more diverse participants if we hope to keep quality at the highest possible level.

The analogy to political strategizing is imperfect. Physicists are concerned with attracting the best and brightest to the profession not in order to get votes or stay on top in some power structure but to solve the difficult and complicated problems facing us today, and to ensure that the U.S. scientific enterprise remains healthy (and pre-eminent, or is it too late for that?).

As the population becomes more diverse, it becomes increasingly difficult to justify the selection effects that result in an overwhelmingly white, male student population in our graduate classrooms. Women remain below 10% of active physicists, and no more than 20% in the youngest, most diverse ranks. The latest AIP data show women receiving 18% of the physics PhDs granted in US institutions. People of color represent a much smaller fraction; for example, fewer than 3% of US citizens receiving PhDs are African-American and Hispanic. For comparison, together African-Americans (12.6%) and Hispanic Americans (16.4%) represent more than a quarter of the US population. By 2043, according to the US Census bureau,
our country will be majority minority.

I am suggesting that graduate education must diversify not because of fairness or equal opportunity, although that certainly ought to concern us, but because it’s vital
for physics.

Why Diversity is Vital for Physics


The first statement of the problem is simple: if we for any reason exclude from our laboratories and classrooms more than 60% of the population (roughly, half being women, a quarter being racial minorities), we are limiting the bright minds who could bring their talents to bear on some really tough problems. Absent compelling evidence that those excluded are less capable, this is not smart.

But there is an even better argument for increasing diversity and inclusion,
based on research on the roots of innovation: there is a competitive advantage in discovery fields to greater diversity among practitioners. As Sheila Tobias pointed out to me 20 years ago (when we wrote the Baltimore Charter after the first conference on Women in Astronomy in 1992), great civilizations have often arisen at the intersection of trade routes, where people of different societies encountered new ways of thinking. That is, the conflict of ideas stimulates new and better ideas.

More concretely, research shows that diverse groups are more creative and develop solutions to problems that are judged – by people unaware of the origin of the ideas – to be better. From the University of Wisconsin’s Women in Science and Engineering Leadership Institute’s booklet on Benefits and Challenges of Diversity in an Academic Setting, written by Eve Fine (historian of science and WISELI researcher on women and science) and Jo Handelsman (award-winning biologist then at the University of Wisconsin, now at Yale University):

A vast and growing body of research provides evidence that a diverse student body, faculty, and staff benefits our joint missions of teaching and research by increasing creativity, innovation, and problem-solving. Yet diversity of faculty, staff, and students also brings challenges. Increasing diversity can lead to less cohesiveness, less effective communication, increased anxiety, and greater discomfort for many members of a community.
Much of this research has been done in a business context rather than an academic or intellectual one, so it may be that the results do not apply in the physics world. However, (a) business organizations hire many physicists, and (b) business organizations are probably more aware than slowly changing academic physics departments of the influence of workplace culture on performance. So I believe this research is highly relevant to what we do.

A typical experiment is to create small groups that are, or are not, diverse in gender, race, class, or other variable(s). Each group works independently on a set problem. For example, in one management training class I took, we were told our airplane had crashed while off its intended route, probably hundreds of miles from the nearest city; that we had a limited list of supplies at hand (compass, bottle of vodka, salt pills, blanket, mirror, ...), of which we could choose only five; and that each group of 5-6 people should decide collectively what to do. Our group’s answer was probably pretty conventional, albeit almost totally wrong: we decided to take off in the best-guess direction (judged from the sun angle, some of us being astronomers), and we decided to consume the salt pills and carry the compass, bottle of vodka and mirror. I remember this quite vividly because I disagreed completely with the rest of the group. I thought we should stay put (someone would report us missing, they’d start a search, they’d get farther with search planes than we could walking) and that taking salt pills was a big mistake (dehydrating), but I was completely outvoted. (You can tell this still stings!) Despite the fact that I had good explanations, the rest of the group all agreed with one another. They had vaguely heard that salt pills were good for desert environments. I explained that this was correct if you took a salt pill and then drank a ton of water and then went to the desert. I also pointed out that it hastened death to drink saltwater if shipwrecked. But what mattered was that their similar opinions reinforced one another. They easily ignored the one outlier (me).

In their article on “Ethnic Diversity and Creativity in Small Groups,” McLeod, Lobel and Cox posed a simple problem related to tourism and asked experimental subjects to brainstorm answers. Experts from the travel industry then graded the responses, not knowing which groups produced each idea; they judged ideas from ethnically diverse groups to be “of higher quality – more effective and feasible – than the ideas produced by the homogeneous groups.”

Experimenters also report more strife in diverse groups. It’s much easier to talk to and work with someone who is just like you. But talking to yourself about a difficult problem doesn’t add as much value as talking to someone with a different perspective.

The value of diversity of thought has been demonstrated specifically for undergraduates. When confronted with a variety of different viewpoints, they are more likely to develop “cognitive complexity” (Antonio et al. 2004)

Many experiments, with different boundary conditions, collectively find that:

    1. Diverse groups experience more conflict.
    2. If diversity is welcomed (well managed), diverse ideas lead to better solutions (described in various papers at more innovative, more practical, more original and/or more creative; e.g., Hoffman & Maier 1961; Triandis, Hall & Ewan 1965).
    3. If diversity is unwelcome, diverse groups fail.

What is behind these results? As McLeod et al. (1996) explained, “Hofstede (1980) has shown that people of different ethnic backgrounds hold distinctly different ‘world views,’ as measured by the dimensions of individualism- collectivism, masculinity-femininity, power distance, and uncertainty-avoidance.” That is, heterogeneous groups hold a variety of perspectives. This means different ideas come into play, and perhaps the conflict between ideas challenges the group to improve its reasoning. It may also stimulate creativity (e.g., Nemeth 1992). So diverse backgrounds lead to different views and in the best case, to a beneficial refinement and resolution of those conflicting ideas.

The claim that differences among people cause them to think differently is quite controversial – for example, there are reams of articles debating whether women inherently think differently than men. Without entering that debate, I think it is clear that the experiences of men and women in physics are different, as are the experiences of ethnic minorities and majorities. That is, how we approach problems, how we think about solving them, how we engage and mentor students, how we work with colleagues – in short, how we do our jobs as physicists – is informed by our individual histories. These tend to have been different for men and women, for different economic classes, for racial groups, and so on. So we have a lot to teach one another.

Perhaps we do best when we work with people who annoy us! I try to remember this when someone is really irritating me. “Hmmm,” I think (I hope), “I could probably learn a lot from this person.”

Not all conflict or diversity is beneficial. If minorities are seen as outsiders, their voices are not heard and their ideas do not hold (no one believed me about those salt pills!). This is worse than had the group been homogeneous because there is the burden of conflict without the attendant benefit.

In a nutshell: more conflict plus more ideas leads to chaos (if conflict rules) or superior performance (if conflict is managed). Which situation do we want for physics?

Let me return to the specific issue of graduate education in physics. Our goal is to train bright, young people to be outstanding physicists. It is always easiest to mentor someone who is exactly like you because if you can get inside their head, you know what advice they need to hear. For example, is it good to encourage students (positive reinforcement) or to challenge them (criticism)? My father was a professor of chemistry with a reputation for toughness. His three daughters are scientists yet none of the four women graduate students who worked with him finished a PhD (although at least one went on to a very successful career at Bell Labs, with strong backing from my dad).

Why Diversity Improves a Graduate Student’s Experience and Performance

Think of this issue as “impedance matching” with your students. My father treated his students as he wanted to be treated. That means that when they were going through a tough or indecisive patch, he pressed harder. This had worked well for him. At some low point in graduate school, when his PhD advisor implied that perhaps my dad should give up on his thesis, he came back with a resounding, “No, I can do it!” What he didn’t realize is how others – like me or my sisters – might have reacted in a similar situation. I remember vividly the Thanksgiving dinner years ago, when my older sister (now a biology professor and textbook author) and I told him that kind of approach would have meant the end of our graduate careers. We both knew we would have quit if challenged that way by our advisors. “No, no,” he insisted, “you are both too good to quit.” But we pushed back, and I like to think he learned something that day, unfortunately too late for his women students.

So, my dad (who was a great guy with a big heart) was a wonderful advisor for students with his confidence, his sense of belonging, his style of learning. He would have been a disaster as my advisor. What would happen if all professors in a department were like my dad? Or all of them were like me? It’s not good for the students.

Another way in which diversity is a benefit stems from the increasing role of teams in modern science. Today it is rare for a “rugged individual” with sharp elbows to make the big contributions. The Large Hadron Collider collaboration has upward of 5000 members, and the Hubble Space Telescope is used by thousands of astronomers world-wide, as will be its successor, the James Webb Space Telescope. Even smaller-scale “desktop” physics is typically done by collaborative groups.

Business, too, depends heavily on groups working smoothly together. Yet, much like physicists, they still hold tight to the idea of the top performer, the miracle man, the great leader – even when research shows women are better team players and leaders than men.

Many physicists still cling to the image of Einstein, toiling alone in the customs office, having brilliant insights all by himself. But that is not how science happens today – and it wasn’t really ever that way, even in Einstein’s time. Working together well is critical.

Many institutions have made a lot of progress in diversifying. When I started in physics it was very rare to see women faculty; more than half of physics departments had no women. Now, it is rare to find a department of any size without at least one woman faculty member, and many departments have done much better. There are also prominent physics leaders of color, though
far fewer. (The numerical reality is that women are within a factor of a few of parity, but under-represented minorities are low by an order of magnitude.) But these success stories – women or other minorities in physics – must be seen as exceptions rather than the rule. Typically they faced higher obstacles and thus probably had to perform at a higher level to succeed.

My colleague Peter Parker – some of you know him (no, not Spiderman, but a long-time nuclear astrophysicist) – told me about arriving at Yale in the 1960s to find all-white-male faculty and AWM students – and, he said, “I didn’t even notice there was anything wrong!” He went on to say, “It’s much better now.” Indeed, our top students are frequently women (30-40% of our physics majors) and/or minorities.

My contention (no real data) is that top men and women succeed. The difference is one step down, where men can pass through the evaluation filters and women and minorities generally cannot. We will have achieved equity when women of slightly-less-than-world-changing ability succeed as easily as men of similar ability.

Graduate Admissions and the GRE

Let me touch on one more topic about graduate admissions: the physics Graduate Record Exam. Data show that women and minorities score lower on these tests than white men do (see presentations by Ted Hodapp and Casey Miller, this conference). I know that my own score was low, perhaps because I went through a less rigorous physics program than students at, say, Johns Hopkins (where I went to graduate school) or MIT (where I was a postdoc). For example, I discovered that the textbook we used for quantum mechanics my senior year in college was a sophomore-year textbook at MIT. As a first-year grad student at JHU, I took courses alongside their seniors and occasional juniors. On the other hand, low GRE score or not, I was one of only 3 of 12 first-year students to pass prelims that winter. So the physics GRE didn’t predict my performance level very well relative to my male colleagues. Perhaps, as is known to be the case for the SAT, a given score predicts higher performance for women than for men.

Other women physicists I know – including incredibly successful, outstanding scientists and leaders – also had low physics GRE scores. There are undoubtedly a variety of reasons for this. One interesting explanation is “stereotype threat,” the under-performance of a group because of a negative stereotype rather than lack of ability, a field developed by Claude Steele and colleagues. The classic experiment is as follows: a class is told they are to take a very difficult math test. The men score an average 25 of 100 points on the test; the women score 10. Some journalists would stop right there and write yet another article about how women are not as good at math as men. As Larry Summers famously put it, at the high end of the distribution of abilities, men vastly outnumber women. (He neglected to mention that Japanese women out-perform American men, or that the gender difference within the U.S. has changed markedly over a 30-year period, ruling out any kind of Y-chromosome-based explanation.)

What Steele and company did instead was to repeat the experiment with one small change: they typed, “This test was designed to be gender neutral” at the top of each test. Now the women score 20 of 100 points, double their previous score. Interestingly, the men’s average score is the same, 20/100, the idea of gender neutrality somehow having suppressed their performance. In any case, the gender gap is gone.

These experiments are repeatable, and subsequent testing turned up other interesting results. The gender gap is exacerbated when the stereotype is activated or emphasized. If you were to say to your students that women generally don’t perform very well in physics – well, you would probably cause the performance of women in your class to drop. A similar effect is seen for under-represented minorities in science and math. Note that these gaps occur regardless of the educational background of the students; even those who have performed well in rigorous programs can be induced to under-perform under the pressure of negative stereotypes. Now think of the atmosphere of a graduate physics cohort that includes very few women or minorities: the subliminal (and sometimes explicit) message is that those minority groups are not top performers in physics. Lo and behold, perhaps this leads to low GRE scores and issues with qualifying exams.

The dominant skill the physics GRE tests for is the speed at which the test taker does physics problems – which measures familiarity as much as understanding. Indeed, it favors rote learning over conceptual thinking. There are way too many questions for the time allotted. We use the GRE because it is quantitative and standardized. It provides a uniform way of measuring something about students, independent of the vagaries of their institution’s grading or letter-writing practices. But does it actually measure what we want to know?

There are many qualities needed by a successful physicist, which each of us has in different measure: intelligence, curiosity, creativity, determination, persistence, openness to new ideas, speaking ability, writing skill, ability to multitask, integrity, teaching ability, sense of the big picture, willingness to ask questions, ability to work with others, etc. Qualities that are sometimes used as proxies for excellence but in my opinion have little to do with the ultimate impact of one’s work: aggressiveness and assertiveness, as well as how little sleep a person can get away with.

What qualities does the physics GRE score correlate with? As reported at this conference by Frances Hellman (UC Berkeley) and Casey Miller (U Central Florida), performance in graduate school is uncorrelated with physics GRE score, for scores above roughly the 30th percentile. Some students with low GRE scores turn out wonderfully well, while some with high GRE scores wash out. Nor is performance correlated with Verbal or Quantitative GRE score, or with GPA in college, or with any other easily reported variable. Perhaps this should not surprise us, given the range of qualities that affect a scientist’s work.

At the same time, the dependence of GRE scores on gender and/or ethnicity suggests that the physics GRE is not an unbiased indicator of performance. Using the same cutoffs or ranges as for white male students will unnecessarily exclude diverse students who will perform comparably.

Diversity After Graduate School

I’ve talked a lot about the diversity of students entering graduate study. Let me finish with just a few words about the diversity of outcome: not every student will become an academic.There are obvious reasons why the academic route is better for the advisor: her work will be multiplied, her papers will be referenced, her students will send her their protégées – in short, her career will be helped. Having worked in both a laboratory environment, at the Space Telescope Science Institute, with minimal access to students, and in a university bursting with excellent students, I can tell you it’s much easier to be productive with students.

But not all students are well suited for the academic track, which they will realize along the way. Encouraging their aspirations is our job. We should train them well to do whatever physics-y thing they decide is their direction. I almost said “their passion” but that goes back to the medieval priestly version of academia. Indeed, at the same 1992 Conference, Sheila Tobias described the origin of the modern academy in medieval monasteries, and the persistence of that culture into the culture of physics today:
    •  “Calling” – priests/physicists are born to the profession, or not; it’s not something you can learn to want to do, or learn to do better.
    •  “Dedication” – physics takes precedence over everything else; you must devote every waking hour to it.
    •  “Celibacy” – god forbid you should have a personal life, like a partner or a family.


Does this sound like the physics profession? I contend it’s not a good template (though it may explain why the template attracts white male dominant populations). The truth is: people can come late to physics and find amazing things, they can learn (as opposed to always knowing how to do everything), and many, many physicists have combined work with family life. So the monastery is the wrong model.

The analogy works in this way, though: if an older monk mentors an apprentice monk, he feels a failure if the apprentice leaves the monastery. The term itself – “non-academic” careers – clearly reveals our bias toward keeping our protégées in the priesthood. We will have other talks about this so let me make two quick points:

We should help our students succeed regardless of where they go. (Professors, colleges: your future endowments are more likely to come from outside academia.)

This means we need to train students broadly, beyond the standard Mechanics-E&M-StatMech-Quantum skill set. Actually, management skills will probably help students in standard academic positions as well.

Final Words

We are fortunate to work in an important, exciting field that we love and to be well paid (even graduate student salaries exceed the poverty level for a family of four). Let’s also remember that we are ideally suited to better the world. I’ll end with one final anecdote:

Two weeks into the fall term, shortly after I became department chair, one of the incoming graduate students asked to meet with me. She was thinking of leaving graduate school, she said, because she wanted to “help others.” She had spent the summer working for a non-governmental agency in South America, and felt that was much more valuable work for the rest of humanity than solving the fluid equation.

For a quick moment, I saw her point – saw how it looked to her – and I realized in that same instant that we are teaching students that what we do is an intellectual exercise, gratifying to ourselves and other weird creatures like us but not ultimately useful to others.

This is wrong. As the provost at a large Midwestern university once said to me (he was an economist and son of a physicist), “Physicists are a lot like economists. They think they are the smartest people on the planet; they think that if they have not addressed a problem, it has not been solved; and they think there is no problem they could not solve.” Sounds about right. ;) So let’s put our money where our mouths are: let’s teach our students that they can solve the problems of the world, that physics tools are essential. Look at climate change, global warming; biological systems; even finance: physicists are there in the thick of it (for better and worse). There is nothing we can’t do. Even if you won’t go quite that far, we can do as much or more than anyone to address the challenges facing this nation and the world. So let’s find students who reflect the constituency and interests and concerns of the world, and equip them to make the world a better place.