To think that this is discrimination

[unnervingrabble]

The BBC have reported that 'The Scottish Qualifications Authority lowered grades using an algorithm - with pass rates for pupils in deprived areas downgraded further than those in more affluent parts.'

Surely this this discrimination against poorer Students who have been doing Highers? They have had their grades reduced by 15% in the most deprived areas but 7% in the most affluent areas.

It is discrimination and I've seen a lot of angry backlash against it by voters on social media. I dont know how they thought it was ok, or wouldnt be noticed.

It is discrimination.

I hope voters remember that when it comes to elections in Scotland next year.

Its disgusting. Those pupils don't have a chance to prove how well they could have done. If anything they should be given the benefit of the doubt, not downgraded. And certainly not downgraded based on where they bloody well live.

There will be a strong correlation between IDACI indicators and performance in school.

This will then feed into a statistical model, that gives the most likely outcome/expected results for a ‘normal’ pupil. There will naturally be exceptions but statistically, it’s proven to be an accurate coefficient.

It's absolutely ridiculous- I'm seeing news reports about students predicted 5 As from prelims and course work being given B and C grades now not getting places for Medicine and Vet science!

I thought so. I would love to see the SQA justification for this, I suspect that there isn't one.

Technically wealth isn’t a characteristic protected by law. Saw Sturgeon has ordered a review. I hope she gets on with it fast. Sounds like the algorithm approach is fatally flawed anyway. Why do a certain % have to fail?

@Pumpkinnose

Technically wealth isn’t a characteristic protected by law. Saw Sturgeon has ordered a review. I hope she gets on with it fast. Sounds like the algorithm approach is fatally flawed anyway. Why do a certain % have to fail?

"It is not enough to win. Others must lose"

Gore Vidal.

It'll be interesting to see if the same discrimination occurs with the A level results in England.

Is it discrimination though?
Did they know when they were sorting the grades what was the postcode/school? If so, would they necessarily have been aware of the meaning of that postcode? Some yes-inner city Glasgow I'd imagine is pretty poor, but other than that, would the average person really be able to look at it (assuming they were told the school, which I would tend to suspect they wouldn't be) and know?

It's possible that teachers in those areas were more inclined to overestimate. I don't know, but before saying how disgraceful it is, that is something that needs to be looked at.
The other thing to look at would be have those areas got lower grades than they have historically. If that is the case, then it does need to be looked at.

One local school here, which is a very middle class area had a similar thing happen over A/S levels a few years ago. Forms that had been predicted A/B levels got D/E. It was devastating for my friend's dc whose ds was expecting to be applying for university with a row of As and to be applying with CDE.

I know the school I went to in the 6th form would have rubbed their hands together, and said, firmly with their fingers crossed behind their back, that this particular consort was exceptional and was looking at As across the board. They'd have then made a huge scene on how mean and horrible the system was.

The Scottish appeals has also said that if an individual has got a lower grade and can provide proof that they were working at a higher level, then they will raise the grade. So if that's the case, I would expect to see a lot of shifting upwards.

But how likely is it that all the teachers in the less well off areas - every single one of them - overestimated the grades, and all the teachers in the better off areas didnt?

Of course the same thing will happen in England. This is the problem when you place too much reliance on machine learning and artificial intelligence. The biases from the developers are programmed into the system and it churns out more of those biases. If they had a diverse group of programmers from across social backgrounds, they could have mitigated against the poorer students being automatically downgraded.

This is one of the reasons I’m scared of Cummings obsession with big data and AI, if not done correctly it could increase inequality.

artificial intelligence

No such thing.

The system in England is that schools rank pupils in order and put the grade dividing lines in.
Then the exams boards / computer look across to see whether the expected proportions are getting each grade. If a school usually gets a certain profile and this year's cohort aren't significantly more able (as shown by SATs) then if the profile set by the school is too high (or low) then the grade dividing lines will be moved.

If they didn't do that then all schools could over grade. It's a standardisation process.

It might well be that poorer performing schools were overly optimistic to see what they could achieve for their pupils.
It doesn't mean it is wrong if they get downgraded more.

Did Scotland have a similar system?

There has to be something better - there really has to. None of this sounds good enough when we are supposed to be working to improve the chances for children coming from disadvantaged backgrounds.

@Pumpkinnose

Technically wealth isn’t a characteristic protected by law. Saw Sturgeon has ordered a review. I hope she gets on with it fast. Sounds like the algorithm approach is fatally flawed anyway. Why do a certain % have to fail?

But wealth is strongly associated with social minority group characteristics that are protected by law.

So the more pertinent point is why people under-achieve in those areas in the first place (hence programming the algorithm to account for it), and what needs doing about that.

Presumably they tabulated all the "estimated grades" provided by teachers and it's turned out far too high compared with previous years.

So then they've drilled down to find out which schools/areas have inflated grade estimates.

Perhaps teachers should have been more honest/realistic in the first place?

40% is a huge amount to be wrong by.

Perhaps, but it's not unlawful.

All this will do is increase the clamour for ‘outstanding’ schools. Nobody whose able to choose will send their kids to a school below that standard because as this situation shows, you could be marked down based on the schools past performance.

Also what happens to schools which have underperformed historically but have recently made some changes and are on the up? Do they have local knowledge based on current information on schools? I went to a school which was shit but then got a new head, made significant changes and is now one of the best in the borough. It wouldn’t be fair on the new intake to be judged on the old intake in that situation.

The algorithm will have years worth of data behind it. Unpalatable it may be but there are some very consistent patterns in the predictive grades given out by schools, I'm trying to remember the stats but I thought poor schools under predict, not over predict!!!

Off to search for Scotland specific data

Cards on the table - my DS at a nice "middle class", high performing school has benefitted from this.

On a nationwide level it makes sense - if a certain school always presents 100 candidates, 50 get an A, 25 a B and 25 a C, you'd expect the same this year. More or less.

But that only works for big schools, who present hundreds of candidates a year. For Highers especially the choices are more specialised - out of 210 in my DS's year, he had about 12 in his computing class. Last year there were 4 of them. So the numbers aren't really big enough. And of course you have the anomaly of a super-high performing child in a school which usually gets bad results being denied the grades they need for Uni.

But on the other hand, if they'd gone purely with what teachers were predicting, there would have been a 25% uplift on passes from last year and we'd all be screaming that the exams were devalued and the grades were meaningless.

It's a total fuck-up - last in a long line of Swinney fuck-ups. The whole SQA is appalling - we have personal experience of papers "going missing" last May, took until February to sort out. The chief exec of the SQA went to school with Nicola Sturgeon. The guy in charge of English exams at the SQA is Swinney's brother. Jobs for the boys/girls, irrespective of actual ability.

I had heard (but am willing to be corrected) that despite these reductions via algorithms that the overall results were actually increased from previous years by 1 and 2 % for both levels?

This was from ds college sendco when I said I was concerned about results not reflecting ability (they did mocks so it's not like they didn't have any data).

That's right. After moderation the pass rates were up a bit.

But the problem is that young people who are from less aflluent areas do worse. Always have done worse, for a whole host of reasons.

And for once, since education is wholly devolved, Nicola can't blame Westminster for this one.

@SockYarn

That's right. After moderation the pass rates were up a bit.

But the problem is that young people who are from less aflluent areas do worse. Always have done worse, for a whole host of reasons.

And for once, since education is wholly devolved, Nicola can't blame Westminster for this one.

The divide is education is definitely an issue that needs addressing.

Previous attempts over the years across the whole UK haven't been the most effective.

Seems a shame the divide is being used as a political tool now though if students really have just received what they likely would receive.

The scandal is why the difference exists. Not that an algorithm worked on the previous divide.

@CuriousaboutSamphire

The algorithm will have years worth of data behind it. Unpalatable it may be but there are some very consistent patterns in the predictive grades given out by schools, I'm trying to remember the stats but I thought poor schools under predict, not over predict!!!

Off to search for Scotland specific data

I know this is about Scotland but if there is any similarity with English schools then I would truly question this line...

The algorithm will have years worth of data behind it

as exactly where the problem lies. It is not in the statistical process applied to datasets but the methodology employed in building them that is cause for concern.

It is not a fair or accurate assessment of someone's ability and never has been - to put the focus on complex statistical methods and algorithms is wrong, imv - they are only as good as the person who utilises them to crunch the numbers.