Students at schools that previously performed badly will have their grades downweighted?
The article does not make it easy to understand how the actual model works. According to the IBO [1]:
We will use a calculation that is based on the relationship between coursework marks, predicted grades and subject grades to estimate the subject grades candidates would have received if the exams had gone ahead. If the relationship between these elements shows that in previous sessions candidates globally tended to achieve higher outcomes on their exams than their coursework, the calculation used this session will reflect that.
You can call this "reverse affirmative action", but really it's anti-grade-inflation. If you attend a school where students historically perform much better on coursework than on the exam, your predicted score on the exam will be lower. On the other hand, if your school tends to perform better on the exam than in coursework, then your predicted score on the exam will be higher.
So why would one school underperform on coursework and another underperform on exams? Grade inflation. Lots of schools make their coursework easier to try and boost the chances their students get into university. The whole point of standardized testing is to counteract that.
So in this case, it may be very unfair to some students who might have been outliers on the exam, but there's no way of knowing that without actually running the exam.
> If you attend a school where students historically perform much better on coursework than on the exam, your predicted score on the exam will be lower. On the other hand, if your school tends to perform better on the exam than in coursework, then your predicted score on the exam will be higher.
This is terrifying. It works for the average student, but distorts the performance for any student who isn't typical or better than their school.
It seems that it would be better to just ignore the testing component of the IB score this year when judging students. Instead they modeled what the students would have gotten on the test, and then judged them based on their modeled performance.
distorts the performance for any student who isn't typical or better than their school.
It distorts the performance of a student who underperforms in coursework relative to their peers but overperforms on exams, at a school where students overperform on coursework and underperform on exams. In other words, it punishes the student who does unusually poor on coursework that is already much easier, on average, than typical coursework from other schools.
How likely are such students to succeed at a competitive university? Not very likely at all, I would think.
I don’t agree. It also punishes students with good coursework and who would have performed well on the test, but happen to be in schools which historically perform poorly on IB tests.
yes, they would be higher than their peers at their school, but then penalized because the cohort before them performed poorly on the IB test.
For the students who would have gone on to test well on the IB test, this is clearly a disadvantage.
E.g. :
1) You are in the 10th percentile of your class.
2) Students from your school in the 10th percentile scored in the 20th percentile on the IB test last year
3) You are assigned a 20th percentile on the IB test you can not take.
This sucks if you would have done better than the 20th percentile on the IB test.
Why model an IB test score at all? just apply a weighing to the coursework grades if you want to standardize outcomes between schools. it would be more straight-forwarded.
They modeled the scores because they couldn’t hold the exams this year due to the virus. As for your description, that’s a gross oversimplification. What they’ve actually done is predicted people’s test scores based on their performance in coursework relative to their peers.
A person in the 10th percentile in coursework is already worse than 90% of their peers and is unlikely to go to university no matter what school they attend. On the other hand, a student in the 90th percentile of a low performing school is going to be fine. Their predicted score is going to be well above their peers.
> They modeled the scores because they couldn’t hold the exams this year due to the virus.
I said why make up a test score instead of just using weighted coursework.
> As for your description, that’s a gross oversimplification. What they’ve actually done is predicted people’s test scores based on their performance in coursework relative to their peers.
This is exactly what I described. taking someone's performance percentile in their school and predicting what they would score on the IB test. This appears to be exactly what they are doing
>A person in the 10th percentile in coursework is already worse than 90% of their peers and is unlikely to go to university no matter what school they attend
Sorry I was not more clear, I meant top 10th percentile.
>On the other hand, a student in the 90th percentile of a low performing school is going to be fine. Their predicted score is going to be well above their peers.
No, a student in the 90th (top 10th) percentile of coursework at a low performing school is the ones who are harmed. They will be assigned a higher IB test score than most of their school, but lower than the 90th (top 10th) percentile student in a school with a better track record on the IB test.
I said why make up a test score instead of just using weighted coursework.
That's what they're doing. They're weighting someone's coursework by an estimate of how they'll do on the test, based on how well students at their school have done on the test in previous years relative to coursework in those years.
No, a student in the 90th (top 10th) percentile of coursework at a low performing school is the ones who are harmed
Of course, they are harmed by the fact that their school makes coursework too easy. The alternative, of course, is using unweighted coursework grades which harms the students at schools that make the coursework extra hard.
The problem is that the IBO doesn't actually know what a student's abilities are until they write the exam. They also don't know how hard the coursework is at their school, in an absolute sense. They only know how hard the coursework is relative to the exam, based on a comparison of previous years' course grades with exam grades. Schools that consistently make coursework too easy will inflate the grades of their students. If those students underperform on the exam, the IBO will know that those schools are practicing grade inflation and adjust accordingly. That's all there is to it.
In the future I just have to do really well in kindergarten and all the rest of my school and career will be predicted by a ML model with no correction whatsoever.
I’m an IB grad and 5/6 exams could have been done electronically. Should at least give students the option to take exams.
The article doesn’t include any commentary related to students that were mis-predicted in the other direction, i.e higher-than-expected scores. If the results stand, they’ll be an interesting case study in the predictive value of IB scores on college performance.
My 13 year old son just started doing classical mechanics on ocw.mit.edu
No one can keep students out of open courseware.
signaling is a zero sum game, but there is always a shortage of students who have the interest and aptitude to learn.
15 comments
[ 4.8 ms ] story [ 50.3 ms ] threadWait, this is like reverse affirmative action, no? Students at schools that previously performed badly will have their grades downweighted?
The article does not make it easy to understand how the actual model works. According to the IBO [1]:
We will use a calculation that is based on the relationship between coursework marks, predicted grades and subject grades to estimate the subject grades candidates would have received if the exams had gone ahead. If the relationship between these elements shows that in previous sessions candidates globally tended to achieve higher outcomes on their exams than their coursework, the calculation used this session will reflect that.
You can call this "reverse affirmative action", but really it's anti-grade-inflation. If you attend a school where students historically perform much better on coursework than on the exam, your predicted score on the exam will be lower. On the other hand, if your school tends to perform better on the exam than in coursework, then your predicted score on the exam will be higher.
So why would one school underperform on coursework and another underperform on exams? Grade inflation. Lots of schools make their coursework easier to try and boost the chances their students get into university. The whole point of standardized testing is to counteract that.
So in this case, it may be very unfair to some students who might have been outliers on the exam, but there's no way of knowing that without actually running the exam.
[1] https://ibo.org/news/news-about-the-ib/covid-19-coronavirus-...
This is terrifying. It works for the average student, but distorts the performance for any student who isn't typical or better than their school.
It seems that it would be better to just ignore the testing component of the IB score this year when judging students. Instead they modeled what the students would have gotten on the test, and then judged them based on their modeled performance.
It distorts the performance of a student who underperforms in coursework relative to their peers but overperforms on exams, at a school where students overperform on coursework and underperform on exams. In other words, it punishes the student who does unusually poor on coursework that is already much easier, on average, than typical coursework from other schools.
How likely are such students to succeed at a competitive university? Not very likely at all, I would think.
For the students who would have gone on to test well on the IB test, this is clearly a disadvantage.
E.g. : 1) You are in the 10th percentile of your class. 2) Students from your school in the 10th percentile scored in the 20th percentile on the IB test last year 3) You are assigned a 20th percentile on the IB test you can not take.
This sucks if you would have done better than the 20th percentile on the IB test.
Why model an IB test score at all? just apply a weighing to the coursework grades if you want to standardize outcomes between schools. it would be more straight-forwarded.
A person in the 10th percentile in coursework is already worse than 90% of their peers and is unlikely to go to university no matter what school they attend. On the other hand, a student in the 90th percentile of a low performing school is going to be fine. Their predicted score is going to be well above their peers.
I said why make up a test score instead of just using weighted coursework.
> As for your description, that’s a gross oversimplification. What they’ve actually done is predicted people’s test scores based on their performance in coursework relative to their peers.
This is exactly what I described. taking someone's performance percentile in their school and predicting what they would score on the IB test. This appears to be exactly what they are doing
>A person in the 10th percentile in coursework is already worse than 90% of their peers and is unlikely to go to university no matter what school they attend
Sorry I was not more clear, I meant top 10th percentile.
>On the other hand, a student in the 90th percentile of a low performing school is going to be fine. Their predicted score is going to be well above their peers.
No, a student in the 90th (top 10th) percentile of coursework at a low performing school is the ones who are harmed. They will be assigned a higher IB test score than most of their school, but lower than the 90th (top 10th) percentile student in a school with a better track record on the IB test.
That's what they're doing. They're weighting someone's coursework by an estimate of how they'll do on the test, based on how well students at their school have done on the test in previous years relative to coursework in those years.
No, a student in the 90th (top 10th) percentile of coursework at a low performing school is the ones who are harmed
Of course, they are harmed by the fact that their school makes coursework too easy. The alternative, of course, is using unweighted coursework grades which harms the students at schools that make the coursework extra hard.
The problem is that the IBO doesn't actually know what a student's abilities are until they write the exam. They also don't know how hard the coursework is at their school, in an absolute sense. They only know how hard the coursework is relative to the exam, based on a comparison of previous years' course grades with exam grades. Schools that consistently make coursework too easy will inflate the grades of their students. If those students underperform on the exam, the IBO will know that those schools are practicing grade inflation and adjust accordingly. That's all there is to it.
I’m an IB grad and 5/6 exams could have been done electronically. Should at least give students the option to take exams.