TikTok in effort, A-stage and Scottish Highers algo fiascos – Computer Weekly Downtime Upload podcast
Hear to this podcast
On this week’s episode of the Computer Weekly Downtime Upload podcast, Caroline Donnelly, Clare McDonald and Brian McKenna discuss about the fortunes of TikTok, the A-stage outcomes, and the algorithm controversies all the plot through which they and Scottish Highers bear been mired
On this week’s episode of the Computer Weekly Downtime Upload podcast, Caroline Donnelly, Clare McDonald and Brian McKenna discuss about the fortunes of TikTok, the A-stage outcomes, and the algorithm controversies besetting those and Scottish Highers.
Caroline kicks off the podcast by recounting the recent fortunes of TikTok, the short-create video publishing and community platform that US president Donald Trump sees as an arm of the Chinese language Communist Occasion.
Caroline is herself an avid TikTok devotee and especially loves staring at “correct dog” videos. She furthermore objects to the app being described as for Period Z, and not millennials. And it’s been a lockdown haven from the doom-scrolling corona dystopia of Twitter, she says. And it’s been downloaded two billion cases.
However the assign is beneath threat in the US, where the authorities sees the service, which is owned by Chinese language company ByteDance, as a privateness-busting malicious program, controlled by the Chinese language Communist Occasion, potentially at the least enabling blackmail and company espionage.
As Ryan Priest info in CW’s Downtime column, Scottish historian Niall Ferguson furthermore takes a dour peep of the service, describing it as “not true China’s revenge for the century of humiliation between the Opium Wars and Mao’s revolution. It is the opium – a digital fentanyl, to glean our children stoked for the arrival Chinese language imperium”. Heavy stuff.
TikTok’s salvation in the US may presumably well presumably are available in the create of acquisition by both Microsoft or Twitter, in any other case it can presumably well presumably spin dim on 20 September, its lip-syncing days over, and its users compelled to migrate to Instagram.
On this facet of the pond, issues seek less bleak for TikTok, that may presumably well presumably unbiased commence its first European datacentre in Ireland in 2022, because it appears to be like to be like to curry favour with European knowledge privateness watchdogs.
Whereas these are engaging and potentially dim cases for TikTok and its Period Z (and older) devotees, final week was once a with out a doubt mammoth week certainly for younger adults in England, Wales and Northern Ireland: the A-stage outcomes bear been announced on Thursday 13 August – outcomes with out tests, but with added algorithmic “standardisation”.
(The podcast was once recorded on Friday 14 August).
Clare has been closely monitoring the gender split for A-levels, especially with admire to computing. This yr there bear been 10,629 male candidates and 1,797 female candidates for computing, which is disappointing.
Nonetheless, in 2018 there bear been 1,475 female candidates, so there has been an create bigger, and, proportionately, 11% of the female candidates got Agrades, when in contrast with 7.5% of the males. Also, computing college students seem to bear been less adversely struggling from the Ofqual algorithm than bear A-stage college students extra on the entire.
The podcast team then spin on to insist relating to the controversy surrounding the statistical mannequin applied to trainer assessments (after which, later in the podcast, the mannequin applied by the Scottish Qualifications Authority to Highers, Developed Highers and Nationwide 5s).
For handy reference, this newsletter, from the Ofqual duration in-between portray on the full sweep of the England, Wales and Northern Ireland tests, describes the Ofqual mannequin:
“Our most in sort mannequin – is known as the Converse Centre Efficiency (DCP) mannequin – works by predicting the distribution of grades for every particular person college or college. That prediction is per the ancient performance of the college or college in that topic taking into narrative any changes in the prior attainment of candidates entering this yr when in contrast with outdated years. This was once unbiased appropriate-looking out-tuned to take narrative of known points equivalent to centres with tiny cohorts of college students, tiny-entry themes and tiered themes. Selections bear been furthermore made on the amount of years of ancient knowledge incorporated in the mannequin.
“The tiny print of these choices are assign out on this portray and are formalised in the regulations we set aside aside in assign for summer 2020. The assign schools and schools had a relatively tiny cohort for a topic – fewer than 15 college students when taking a peep in each place in the recent entry and the ancient knowledge – the standardisation mannequin set aside aside extra weight on the CAGs [Centre Assessment Grades]. Since tiny instructing groups are extra current for AS and A-stage than for GCSE, and on condition that the CAGs tended to be optimistic, it skill that the outcomes in some AS and A-stage themes are grand elevated this yr. Nonetheless, there may be no statistical mannequin that may presumably well reliably predict grades for in particular tiny groups of college students. We have got therefore ancient the most legitimate proof available, which is the CAGs.”
A Guardian editorial summed up a imaginable critique of this sort, which was once basically aimed, Brian comments, at suppressing grade inflation:
“By building in a criterion of previous college performance to this yr’s A-stage and GCSE outcomes, Ofqual has tied the fortunes of particular person college students to pre-present inequalities of consequence. To strive on this diagram to impose an fashioned spread of outcomes, which pupils are powerless to affect, is unjust.”
Clare comments in dialogue that an option of sitting an exam in the autumn would be nerve-racking certainly, on this yr’s circumstances.
Brian then takes the team back to early August, when the outcomes of the Scottish Highers, Developed Highers and Nationwide 5s (the identical of GCSEs) bear been announced. This was once the “first time as history” to the the rest of the UK’s “second time as farce”.
Any other time, the outcomes declared bear been the consequence of a combination of trainer assessments and the application of a statistical mannequin to outdated years’ college outcomes knowledge. The next text sums up the mannequin the SQA chose:
“In easy terms, mathematical optimisation (extra popularly known as ‘optimisation’) is a family of systems ancient to title the most easy imaginable resolution to fulfill a talked about object per one or extra defined constraints. Essentially, optimisation was once chosen as the most popular diagram for adjusting estimates, since it tests all imaginable solutions concurrently, in hiss to title the ‘most effective available’ value for an goal goal – given a assign of constraints, in a sturdy and efficient diagram.
“Furthermore, optimisation systems are tested and proven, both in trade and literature, and therefore present a credible plot for endeavor the changes required to make stronger this yr’s awarding. The optimisation plot applied was once per a blended integer linear program within a community framework to be definite the relativity of refined bands on a course as estimated by a centre, was once always maintained. The assign adjustment was once required, the predominant goal goal of the optimisation activity was once to minimise the amount of candidates moved between grades to fulfill the centre constraints for every grade and A-C rate.”
Right here’s from the SQA Technical File Nationwide Qualifications 2020 Awarding – Methodology File.
Brian talks the team during the legend of the Scottish case, where the Scottish authorities determined to revert to trainer assessments. First minister Nicola Sturgeon apologised and talked about the mistake had been to focus too grand on the overall system, and not enough on particular person pupils.
One in every of the most damning outcomes of the application of the statistical mannequin, says Brian, is that the most deprived one-fifth of the Scottish college students seen their cross rate spin down by 15%, while the wealthiest one-fifth seen theirs spin up by 7%.
Brian expresses his hold peep that both governments may presumably well unbiased quiet bear been less hung up on grade inflation this yr, given the stress all of society has been beneath due to the Covid-19 pandemic. Because it is, by imposing these statistical items – on the opposite hand correct they would presumably well or may presumably well presumably not be in technical terms – they bear merely created rods for their very hold backs.
Whatever the truth of that, this story will continue to trudge, especially given the drawing shut publication of the GCSE outcomes this coming week.
Postscript. In a dramatic U-flip, the authorities determined, gradual on the afternoon of Monday 17 August, that A-stage and GCSE college students in England will be given trainer-estimated grades.