Thread by @fromarsetoelbow

A thread on the politics of that algorithm ...

There's a tendency to assume that the algorithm is complex & inscrutable, but it's actually very simple. There's no machine learning or AI here, just a crude adjustment that could have been done by hand if necessary.

I've seen some people poring over the equations used by the algorithm, but this is surely a case of obsessing over the mix of trees instead of the shape of the wood. What matters is not the algebra but the conditional logic that encodes the political spec.

That spec can be summarised as: 1. prevent grade inflation; & 2. be consistent with historic performance. Given that #1 is really only a concern re the state sector, & given that #2 will reflect the state/private disparity, then the spec is essentially: preserve privilege.

The algo logic (expressed in pseudo-code) then boils down to this essential assumption:

if $pupil_sector == "state"
then $final_pupil_grade = adjust_func($pupil-grade)
else $final_pupil_grade = $pupil_grade

One reason we can be confident this is the case is that there is no plan to dump the algo for BTECs. As public school pupils don't tend to be interested in vocational qualifications, the bias requirement isn't there. The algo may actually produce equitably adjusted grades.

Reverting to teacher assessments does not remove bias, as teachers will obviously reflect societal prejudices. Broadly, teachers at private schools will tend to be more optimistic about their pupils, teachers at state schools more pessimistic about theirs.

The way to overcome that systemic bias, and to simultaneously ensure that the number of pupils who can claim a place at uni is within the original planned capacity, is to divorce grading from placement.

What this means in practice is that the teacher grades should stand, but that their utility in respect of uni entrance should only be to provide a ranking of pupils within each school.

Uni places should then be allocated proportionately per schools in the first pass - i.e. a "bog standard" comp gets as many per capita as Eton. If a school has 15 places, then the top 15 ranked pupils get their 1st choice, assuming their teacher-assessed grades are sufficient.

There would still be clearing & some shuffling of quotas between schools based on variable takeup. Overall, this will obviously advantage state schools relative to private, but that should be considered a good thing both in terms of equity & (oh the irony) historical adjustment.

Of course, this will be resisted unto death because once you set the precedent, why wouldn't you do this every year? If you did, the premium of private schools (at least beyond preps) would disappear overnight.

We are now faced with a possible "crash" of the higher education sector because of capacity & funding constraints, exacerbating the underlying problems that have arisen from Covid-19 & the historic over-investment in foreign students.

The govt could bail out lots of struggling unis, but given its commitment to sector rationalisation, getting rid of "low value" degrees, & neutering the "lefty" culture of academia, it is more likely that it will let "market forces" rip, which will benefit the elite.

We are approaching a fork in the road. The current crisis will either see a return to something closer to the pre-1992 landscape or the end of the systemic bias in uni admission in favour of independent schools. The status quo seems less likely, even if exams return in 2021.

The media may be presenting this week as a challenge for Williamson & Johnson, but it may well turn out to be a bigger challenge for Starmer. With even arch-liberals like Sonia Sodha arguing for comprehensivisation, Labour must choose.
theguardian.com/commentisfree/…