1. Oxford University - extends exam times for women's benefit - https://www.telegraph.co.uk/education/2018/02/01/oxford-univ...
2. Prof Bumps Female Students’ STEM Grades (Because They’re Women) - https://reason.com/archives/2018/05/23/prof-bumps-female-stu...
#2 is a completely misleading headline. It ... never happened. An article in which the 2nd paragraph contradicts the headline isn't written in good faith.
[*] I would also argue that extending exams so that each student has a reasonable amount of time to finish is not a bad practice. A student that gives a correct answer in the extra 15 minutes is no less knowledgeable than a student that gives the correct answer in the original allocated time. We are selecting for breadth and depth of knowledge, not speed of recitation or ability to perform under pressure. In a CS exam, you either know the answer or you do not. Extra time is not going to allow you to falsify your level of knowledge. It will, however, give slower workers the ability to fully complete the exam.
Students that work quickly but are less knowledgeable than their peers are the only ones that would be penalized by this change. Those students have inflated scores relative to their knowledge, therefore this penalization should be encouraged. In an untimed test, the most knowledgeable student will always get the highest test score, therefore knowledgeable students should not be opposed to increasing test times, they should encourage them.
This is simply not true. Consider an algorithm proof, with enough time you might be able to derive a proof that you should have known cold.
The real world does have deadlines & performance matters, and the women who spent the time studying should get the better grade
My argument comes from the perspective of the real world. It is, in effect, the same type of argument that drives the "interview questions on a whiteboard" discussion -- which qualities are actually important in an employee? As someone involved with hiring for a company that consistently produces high quality, critical code used in important systems, my experience is that "working under pressure" is pretty far down the list of important qualities.
I'm not saying the person that came to the proof with 30 mins extra is not smart, but they were not able to meet the same expectation as the other students.
If everyone gets a 50, the grades will scale, and your final grade will depend on how you performed compared to the rest
I've brute forced a few proofs, and if i had done so in overtime, i would 100% stand by my viewpoint that i deserve less points than the student next to me who met the expectation.
In an interview it's different too, because you have not been preparing for a clearly defined expectation for 3 months.
Now on the other hand, if you were to argue for completely untimed exams, i can get behind that. I really enjoyed some of my take home CS finals, and it really let me perfect my solution to the best of my ability.
It really depends on what expectation is set (imo)
Secondly: any school that has GPA is essentially holding a contest. Many job postings consider GPA and may use it as a tiebreaker between 2 canidates from school X. It's not a perfect metric by any means, but it is relevant in the world today for new hires.
Work in the real world is largely the ability to deliver on expectations.
In college, I did well on math exams precisely because I had a strong understanding of what I was doing. I did not mechanically follow a recipe for a solution like some other students did.
The result was that I usually took the entire exam period (right up to the last minute) to finish, but I usually had a perfect score or close to it. I outperformed high scorers who had completed the exam 30 mins before me by almost 10 points.
The only exam that got the best of me was the first linear algebra exam, which was a lot of mechanical matrix multiplying. I only completed 3/4 questions before the time was up.
Maybe it's just me, but I would wager that a student that can derive a proof is more likely to understand that proof than one who writes it from rote.
Perhaps that person also makes a habit of not preparing & just winging it (Which is fine until it's not)
The student who was most prepared would be able to derive proofs (by the process that was taught, like knowing to apply a theorom, or being comforatable with induction, etc) in a timely manner.
The exam is there to see how well you grasp the material taught, not how good at winging it you are (which is a very useful skill of course, but not the one being tested)
But some questions in the exam required us to apply some algorithms by hand. I knew how they worked and I could also do them by hand. Did it on whiteboard with toy problems while studying for it. But what I didn't really expected was how fast the time run with manual execution on paper.. Even for modest data sizes.
I thought that the algos really could be improved to be much faster by hand, refactor to optimize for different operation costs when doing by hand given the available tools in the exam, reducing operation costs by speeding up hotspot manual steps by training right into muscle memory and better memory alignment on paper for faster data lookups in the first place.
I probably could have finished that exam in time, but with a baaad score. So I just quit and gave them a blank paper with my name to try again next time.
I knew what the real problem was: time, and I had a plan how to prepare for it. It worked =)
What's the alternative, be forced to drop out of university once you fail an exam?
We wouldn't have been forced to drop out (at least not due to that exam in isolation), but the 0 would have been considered as part of our final grade for the course.
At my university the year was divided into three trimesters with exam periods at the end, and a fourth exam period at the end of summer holiday. Each exam period would have the exams of all the courses of the previous trimester and also exams of courses of the trimester before for people who still needed to pass that course. If you fail the exam twice, you have to try again when the exam is given again next year (or try to argue with the prof to get special arrangements, say if it's the only one you still need for your degree).
You might think A isn't as qualified as B. But it is known that inherent bias against A is very likely. So you might decide, even though I believe B to be more qualified, I will go with A, knowing that my judgement might be wrong, because of the likelihood of inherent bias.
In a way, yes, the known biases, can create a new bias. I bias against my known biases. This does mean that at an individual level, I favored someone else, when I initially wouldn't have. But does it result in better outcomes? I think this is where it matters. Let me explain.
If one is chosen and given opportunity over another. That choice is always going to be judgemental and opinionated. You're trying to predict who will make better use of the opportunity and contribute back the most for being given it.
But predictions, like in machine learning, show that a biased data set often perform poorly. So I think part of positive discrimination is to say, diversity could actually yield better results. Yet, because we don't have diverse data, we don't know for sure. But its still okay for the person who chooses to be of the opinion that it would. I find that no different then saying, well I felt they were more knowledgeable. All of it are intractable metrics in an attempt to predict outcome.
Edit: Some more thoughts related to fairness. Is it fair to give opportunity only to those who would maximise it? What if it was true, that there was a genetic quality, sex, or race, or a more specific one, like height, IQ, EQ, etc., which was the key variable to maximize the opportunity? Is it fair to say only they shall be given such opportunity, always? Or should we pick at random? Would that be more fair? Are there other scheme? What does it even mean to be fair? Your genetics aren't fair, should society be fairer then them? Clearly in the modern western world, we seem to define fairness as, whoever can maximize the opportunity the most, shall be given it. Aka, meritocracy. Is this best for our society though? If so, are we doing a good job at it? What if we try to do that, but instead have been giving opportunity to those who most resemble us? Not to those who can most maximise it? What then? Is that unfair?
Positive discrimination is like bombing a village known to harbor terrorists. Sure, you may kill some terrorists and drive the rest underground. You'll also radicalize a bunch of villagers. Suddenly ideas that used to be relegated to the lunatic fringe are going mainstream.
Sound anything like the US political situation in 2018?
Evidence and facts dont support the retaliatory bias position - retaliatory bias is currently too insignificant a force in comparison to previous systemic or societal biases.
Maybe if we lived in a highly equitable society in the first place, the position would make sense - the forces would be more relatively matched.
Currently, all things being equal - the systemic biases are far more significant and harmful to society as a whole.
Judging fairness by maximizing opportunity sounds a lot like the hypothetical counter-utilitarianism argument about harvesting YOU for organs thus saving more lives. Maximizing fairness to society instead of the individual is the basis of ideologies like communism and fascism.
There are numerous examples of how written exams are often biased against minorities due to either using cultural references/vocabulary or to frame problems using metaphors that more apply to an average white person than an average non-white person, so there’s that.
For example we had here in Sweden for a short time a program which gave bonus points for university student applications if the student would end being a minority in their class based on gender. The problem was that rather then seeing more women entering male dominated fields, it was more men who applied for seats in female dominated fields and thus ended up gaining the benefit of this program. It was then challenged by feminist groups as being discriminative which they won which then resulted in the program being scraped. A major issue is that on average the higher education here in Sweden in female dominated so any positive discrimination that is intended to countering inherent bias would primarily benefit men, unless those actions are applied explicitly to the minority of programs which is male dominated such as abstract math, IT, engineering and a few more. As such modern affirmative actions programs are much more selective in where it get applied.
Same issue applies to using affirmative action to counter gender segregation in the work sector. Both men and women have the same employment rate here in Sweden, and about 90% of men and women work in a profession which is gender segregated. Neither gender is more likely or less likely to be a minority or majority, which means that affirmative action in the hiring process would apply equally to both men and women. The only way to break the symmetry is to only apply affirmative action in places which the intended group is the minority and thus get the benefit, such as two current political initiative to get more women into sport management and management boards. By limit to only gender segregation in those areas they can be sure that only women will benefit from the initiative and political this seems to be a major success.
If we use data sets and machine learning to counter bias then it would likely work but it would also benefit both men and women. Political this is not acceptable. People claim that positive discrimination is to be used to countering inherent bias but the way it is actually used is all about political power and who can benefit from it.
It's worse than that.
What doctor would you go to, a white man, a black man, or a woman, if you knew that there's affirmative action for blacks/women, and you wanted (and could afford) the best doctor?
Mind you, picking white man isn't even racist, it's rational in the Bayesian framework (the white/male had to have better exam scores to be admitted to the school than blacks/women).
Only if you assume no inherent counter-bias, which is of course ridiculous. Without affirmative action, blacks/women would have to have equal exam scores coupled with adverse conditions (statistically, on average) to the white/male candidate. By that logic, they would need to perform much better on aggregate.
Anyway, my original question was about fairness :).
I guess the hope is that increased participation will slowly reduce the severity of adversity, but whether that is effective I'm less sure of (I don't really believe it will ever disappear completely tbh).
As for it being more rational to choose the person who contends with adversity on the assumption that the adversity never disappears (and therefore continues to impair), this only works on the assumption that affirmative action is a complete, perfect remedy to adversity: otherwise those contending with adversity must still overachieve to a certain degree to get the same positions (and therefore represent a potentially better rational choice).
Of course I wouldn't really, I'm just using your logic.
Although technically that's a different effect. Affirmative action is, AFAIK, meant to counter negative bias before (i.e. positive discrimination at university entry is meant to counter worse education and poverty at a primary- and high-school level), whereas you're talking about overcoming biases after graduation ("old boys network" etc).
Just philosophizing.
However, to take the bait and make an attempt at analysing it:
- (race/gender-based) adversity in education usually has the effect of reducing your chances of succeeding in exams, admissions, etc.
- (race/gender-based) adversity in professional practice usually has the effect of reducing your chances of acquiring customers, partners, progression within an organisation.
So while she may still face after admission, that adversity should have a much more significant affect on her paycheck than on her medical performance.
How would that work? If we talked about oral exams or job interview then surely there can be a bias. But written exams? I can't imagine how it would be affected by your race.
What can be however affected is your access to quality education prior to the admission, which will affect your admission changes, which should not be an issue after the admission because now you are getting the same quality education as others.
While I can imagine that socio-economic issues (like being black=poor) may affect your access to quality education prior to the admission, I can not imagine how being a woman would affect your access to quality education. Maybe stereotypes in case of STEM?
What is it with this obsession that every requires the best doctor? I don't want the best doctor, I want a good enough doctor who is close to my house location or I will stick to my current doctor because I had good experience with them. Once I stumble upon problems with that doctor I might reevaluate. Sometimes good is good enough, you don't always need the best of the best. My dentist is a woman, and not white. She's great. My doctor is a white male, he is also great. Skin or gender mean nothing to me in this context.
Being only willing to settle for the absolute best doctor in the world is ridiculous, but it is perfectly rational to want to choose the best one out of a small group of candidates. I can't imagine approaching it any other way when my personal health is on the line.
[1] https://nl.wikipedia.org/wiki/Medisch_tuchtcollege
[2] https://en.wikipedia.org/wiki/Antoni_van_Leeuwenhoekziekenhu...
Feeling safe is related to rapport with the person which is again related to ethnicity, gender, age, behavior, etc.
Maybe you don't actually think that the issue in the Tokyo medical school was such a big deal after all?
1. Seems like a legitimate choice on what a test is designed to measure. Exam times are the same for all students.
2. "Fortunately, the plan appears to have been vetoed." - so it didn't actually come to pass.
In this case extending the time because "female candidates might be more likely to be adversely affected by time pressure" negates the advantage people that can work under pressure have.
I'm curious how they're going to do this in real life when (unsurprisingly) they will be expected to work under pressure and it will highlight a big mismatch between academic and professional performance levels.
The pressures STEM employees face at work are usually very different from exam pressure, so I would lean towards "very little" as the answer to that question.
Everything about academic studies is an arbitrary "neutral point". The quantity of information for one topic, the difficultly of the exam, the time allotted.
It's exactly like saying that you don't actually need to show all that knowledge during the exam since Wolfram Alpha can do that for you. And in real life deadlines are no longer like an arbitrary exam time or even a standardized interview test. They're usually as tight as the best people on your team can afford. That becomes the "neutral point". And that's when it will matter that two people that previously looked equal start performing differently.
An exam tests you for a lot of things and they just lessened one. Which isn't in itself a necessarily bad thing if it didn't give ammo to every person claiming that men are better at STEM. Now their bias is backed by universities so good luck fighting it.
*Just highlighting again that this is discussing the "increased exam time" topic, not the "let's change results to exclude women" topic.
> It's exactly like saying that you don't actually need to show all that knowledge during the exam since Wolfram Alpha can do that for you.
I don't get how this is at all equivalent. Could you please clarify?
And I'm not even touching the situations where it could be a matter of life and death and you have to fix it now.
> I don't get how this is at all equivalent. Could you please clarify?
I was referring to this:
> The pressures STEM employees face at work are usually very different from exam pressure, so I would lean towards "very little" as the answer to that question.
Exams rely on a lot of memorizing which can easily be replaced by Google and WA and is not a reflection of how you will do real work so why test that "knowledge"? The pressure to properly allot time to every question and finish in the exam in time after providing all or most of the solutions is an ability to work under pressure perfectly equivalent to the one you feel doing a task at work. You can't replace the working under pressure skill with almost anything else.
You have to know if those future engineers can do it fast and well even under pressure. Lessening one burden is no different from lessening the other. And when you do it for the benefit of a single group it can't not feel like it's strengthening arguments that some people need help to get through STEM. That's the last thing women in STEM need, having to fight yet another prejudice.
A being a better sprinter than B neither means B can't work under pressure nor that B is the worse mid-distance runner. The same is true for mental tasks.
As an employer, "working under pressure" is not very high on my list of considerations. I would much rather have a complete picture of the candidate's ability to learn and retain information. Well-designed systems take months and months to build, I would prefer information about the candidate's general aptitude than their ability to work in crunch time.
If I was looking for employees for a job in which "working under pressure" was a central component, I would rather test for that in the interview process and have their university scores be uncompromised by pressure.
It's because academic performance is by design an average of skills displayed over multiple years. You can't fulfill the expectations of all employers at once with a standardized test. The degree (and the marks attached to it) just show the average performance per topic over the years. How fast you can finish your assignment is an integral part of any academic program. That's why you get tight(ish) deadlines for any assignment, and exam times.
By extending that time you can no longer make the difference between 2 people with good knowledge but different working under pressure skills. That's the difference between good and best.
I'm not saying the exams become irrelevant, just that this kind of thing does more to discredit the image of women in STEM than a slightly lower exam mark. This just strengthens preconceptions that some groups need help to be equal. And it's backed up by universities and whatever science they used to reach this conclusion.
> By extending that time you can no longer make the difference between 2 people with good knowledge but different working under pressure skills. That's the difference between good and best.
This is true, but if two students have the same score from elongated tests and one does not test well under pressure, it's straightforward to understand that one of the students would have performed better on a time pressure exam. However, consider:
Student A achieves 92% on a 1.75 hour exam but would have achieved 90% on a 1.5 hour exam. Student B achieves 94% on a 1.75 hour exam but would have achieved 88% on a 1.5 hour exam.
In a 1.5 hour exam, the clear conclusion is that student A is the better student. But is student A truly the better student in this scenario? The exam length is a rough estimate made by a university, and it seems like an arbitrary line to draw.
Allowing every student to have enough time to show the breadth of their knowledge on the subject material is not "helping" women. The same principle is applied to both men and women. It is an attempt to provide a level playing field and work towards meritocracy.[1]
Consider an exam on a technical subject that contains questions that are difficult to parse in English. Non-native English speakers have trouble parsing the questions which results in lower test scores. If I suggest that we simplify the language in order to accurately gauge the students knowledge of the subject, you may suggest that this is discrimination against native English speakers. This is undoubtedly true in the context, but it is an attempt to move towards a level playing field which tests knowledge of the subject, which is the entire point of the exercise. You can test for English proficiency separately, that is not the point of the exam.
[1] Contrast this with link 2 in the GP which is straightforward discrimination against men and was roundly rejected for that reason.
If they decided to lengthen times to better fit real world conditions, it is fair. If they went searching for 'fair' changes that could be made that would statistically boost one group over another, it isn't a fair change.
You don't know how common the reader thinks this is occurring, but you're implying its common. Given your sources are both debunked by another reader I'd appreciate a plethora of other sources.