Quizzes & Puzzles1 min ago
Ofqual Chief Does The Decent Thing And Resigns. Meanwhile The Main Culprit, Gavin Williamson Clings On
https:/
There is a pattern emerging.
Cummings, Jenrick and Williamson all *** up massively, but brazen it out as though they are angels.
Bunch of arrogant twits ?
There is a pattern emerging.
Cummings, Jenrick and Williamson all *** up massively, but brazen it out as though they are angels.
Bunch of arrogant twits ?
Answers
Best Answer
No best answer has yet been selected by Gromit. Once a best answer has been selected, it will be shown here.
For more on marking an answer as the "Best Answer", please visit our FAQ.
I am also concerned that my granddaughter's results should not be looked at askance because of the year she sat them. She is still only 14 (was moved up a year a couple of years ago) and sat 2 G.C.S.E.s (History and French) early in order to leave more space next year for the rest. She sat last year's papers in Feb. as mock exams and was graded 7 in History and 8 in French (all using the proper grade criteria).
She is a worker and has worked hard all during lockdown - chatted to us about the Cold War and in French sometimes on the phone. The grades she has ended up with are: History 8 and a wonderful 9 in French. I think her mock results (plus all her school reports) suggest that she is in the right grading area. Very proud of her, so please don't assume that everyone's grades have been inflated -- a lot of pupils earned their good grades.
She is a worker and has worked hard all during lockdown - chatted to us about the Cold War and in French sometimes on the phone. The grades she has ended up with are: History 8 and a wonderful 9 in French. I think her mock results (plus all her school reports) suggest that she is in the right grading area. Very proud of her, so please don't assume that everyone's grades have been inflated -- a lot of pupils earned their good grades.
"A decent algorithm that actually did take notice of predictions would assume all predicted test scores were out by the same percentage and would lower all across the board until the overall grading was sorted."
What percentage, and how do you determine that? I assume the answer would be a comparison to historical results at the school, but that's not an easy comparison when you consider that the format of the exams has changed over the years (content, number of exams, exam board at the school, etc). So presumably you'd have to find a clever way of using the history to guide the results without merely saying "***, this year's class can have last year's grades"...
The problem is that, again, I am more or less describing exactly what Ofqual did. I'd encourage you, OG, to look at Ofqual's technical report on its algorithm -- because, as far as I can tell, you're basically trying to write the exact same thing. All the elements are there: trying to sort the overall grading (which, fun fact, is clearly horribly unfair to the individual); taking the teacher's "guess" as a weighted input; ensuring differentials remained (achieved by requesting that teachers rank their students); matching to previous years, etc. It's all there. All of it. At some point, your continuing to trot out all of the features of the Ofqual procedure while apparently thinking that they were absent from the design suggests that you haven't really looked into what they did, which is surely a major flaw if you are trying to criticise it.
https:/
To be fair, at 319 pages, it's a hefty read, but most of the key info that shows how close their ideas are to yours is on the first few pages.
What percentage, and how do you determine that? I assume the answer would be a comparison to historical results at the school, but that's not an easy comparison when you consider that the format of the exams has changed over the years (content, number of exams, exam board at the school, etc). So presumably you'd have to find a clever way of using the history to guide the results without merely saying "***, this year's class can have last year's grades"...
The problem is that, again, I am more or less describing exactly what Ofqual did. I'd encourage you, OG, to look at Ofqual's technical report on its algorithm -- because, as far as I can tell, you're basically trying to write the exact same thing. All the elements are there: trying to sort the overall grading (which, fun fact, is clearly horribly unfair to the individual); taking the teacher's "guess" as a weighted input; ensuring differentials remained (achieved by requesting that teachers rank their students); matching to previous years, etc. It's all there. All of it. At some point, your continuing to trot out all of the features of the Ofqual procedure while apparently thinking that they were absent from the design suggests that you haven't really looked into what they did, which is surely a major flaw if you are trying to criticise it.
https:/
To be fair, at 319 pages, it's a hefty read, but most of the key info that shows how close their ideas are to yours is on the first few pages.
I suppose I'd also ask a few key questions about what you might have done differently:
1. Would you have allowed the procedure to downgrade students by more than one classification?
2. Would you have ensured that, if there was even the small chance that a student at a given centre would get a U grade, one student *must* be awarded that grade, regardless of their teachers' assessment?
3. Would you have decided not to apply the algorithm to small classes?
4. What is the minimum threshold for predictive accuracy, by attempting to match to the 2019 data, that you would have regarded as a successful test? Ofqual's data suggests, for example, that it was able to "predict" accurately only 26.7% of all grades awarded in Italian A Level (the worst performance for their model), and had an absolute upper limit of 67.9%, for History. In raw terms, this means that the model was shown to get 15,000 grades out of 47,000 wrong. Assuming that this was the best possible result -- exam results are also, after all, driven by random chance as well as a student's skill -- then is this an acceptable amount of misgrading? If not, how would you propose to fix this?
5. Is it reasonable to adjust the teacher's predicted grades, but to treat the way they ranked students as "perfect"? As far as I can tell, in the Ofqual approach, if the teacher's ranking put one student ahead of another, that was preserved in the final prediction, with no attempt to account for the possibility that the "better" student may have had an off day. Would you have tried to correct for this, and if so, how?
All of these are questions you should really have tried to ask already, and really boil down to a well-known problem called the ecological fallacy (or perhaps a fallacy of division): what we can say about the student population as a whole means more or less nothing for any individual. The class of 2020 may well have appreciated that, taken collectively, their grades needed to be smoothed out to be compatible with previous years, but individually this led to injustice after injustice. There is simply no resolution for this problem -- or, at least, not by any algorithmic process.
None of this is to dispute that, taking the centre-assessed grades literally, this is the highest-achieving year ever by a phenomenal margin, but it's a mistake to assume that teachers were being deliberately optimistic, or were basing their predictions on nothing; at the very least, it is difficult for anybody to judge *which* predictions were optimistic, and by how much, with any degree of certainty for the individual. Put another way, the fatal flaw of any algorithm, designed by Ofqual or anybody else, is that it is treating exam results as deterministic when they are manifestly not.
1. Would you have allowed the procedure to downgrade students by more than one classification?
2. Would you have ensured that, if there was even the small chance that a student at a given centre would get a U grade, one student *must* be awarded that grade, regardless of their teachers' assessment?
3. Would you have decided not to apply the algorithm to small classes?
4. What is the minimum threshold for predictive accuracy, by attempting to match to the 2019 data, that you would have regarded as a successful test? Ofqual's data suggests, for example, that it was able to "predict" accurately only 26.7% of all grades awarded in Italian A Level (the worst performance for their model), and had an absolute upper limit of 67.9%, for History. In raw terms, this means that the model was shown to get 15,000 grades out of 47,000 wrong. Assuming that this was the best possible result -- exam results are also, after all, driven by random chance as well as a student's skill -- then is this an acceptable amount of misgrading? If not, how would you propose to fix this?
5. Is it reasonable to adjust the teacher's predicted grades, but to treat the way they ranked students as "perfect"? As far as I can tell, in the Ofqual approach, if the teacher's ranking put one student ahead of another, that was preserved in the final prediction, with no attempt to account for the possibility that the "better" student may have had an off day. Would you have tried to correct for this, and if so, how?
All of these are questions you should really have tried to ask already, and really boil down to a well-known problem called the ecological fallacy (or perhaps a fallacy of division): what we can say about the student population as a whole means more or less nothing for any individual. The class of 2020 may well have appreciated that, taken collectively, their grades needed to be smoothed out to be compatible with previous years, but individually this led to injustice after injustice. There is simply no resolution for this problem -- or, at least, not by any algorithmic process.
None of this is to dispute that, taking the centre-assessed grades literally, this is the highest-achieving year ever by a phenomenal margin, but it's a mistake to assume that teachers were being deliberately optimistic, or were basing their predictions on nothing; at the very least, it is difficult for anybody to judge *which* predictions were optimistic, and by how much, with any degree of certainty for the individual. Put another way, the fatal flaw of any algorithm, designed by Ofqual or anybody else, is that it is treating exam results as deterministic when they are manifestly not.
One needs no clever way. The idea of exams is primarily to give an indication of the ability of the individual to understand and take in the subject matter. That doesn't rely on format of exam, or the syllabus. The same percentage will be in each grade each time. As for this being what was done, it seems not as otherwise the opposition to the results would have had no basis nor support; save from those determined to get a better grade than deserved. The whole complaint was that the teachers' predictions were ignored. Unless you're saying there really was no justification for the protests.
“ The whole complaint was that the teachers' predictions were ignored”
No it wasn’t. The point of the “algorithm” as requested by the government, was to adjust everything down, it it did it very badly.
And I still can’t get over the idea that a scenario where one set of grades were given out, and then a revised largely lower set, were going to cause anything but mayhem. Even if the thing had done its job better.
No it wasn’t. The point of the “algorithm” as requested by the government, was to adjust everything down, it it did it very badly.
And I still can’t get over the idea that a scenario where one set of grades were given out, and then a revised largely lower set, were going to cause anything but mayhem. Even if the thing had done its job better.
// The whole complaint was that the teachers' predictions were ignored. //
No it wasn't -- again, I'm left wondering how much you've really followed the story. The complaints were because the adjustments made to individual students, in many cases, seemed arbitrary and unjust. Teachers understood that their predictions were going to be adjusted, but were left stunned as to how, or which students were affected.
It shouldn't be a matter of controversy here that the details matter, but it seems that you haven't got around to checking them, because all your arguments are based on:
1. telling Ofqual to do what they already did;
2. Criticising students and teachers for complaining about something that didn't actually happen.
No it wasn't -- again, I'm left wondering how much you've really followed the story. The complaints were because the adjustments made to individual students, in many cases, seemed arbitrary and unjust. Teachers understood that their predictions were going to be adjusted, but were left stunned as to how, or which students were affected.
It shouldn't be a matter of controversy here that the details matter, but it seems that you haven't got around to checking them, because all your arguments are based on:
1. telling Ofqual to do what they already did;
2. Criticising students and teachers for complaining about something that didn't actually happen.
Related Questions
Sorry, we can't find any related questions. Try using the search bar at the top of the page to search for some keywords, or choose a topic and submit your own question.