![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
andrewduckerReading this article on advice to teachers in the UK about using AI, they suggest using it for things like "marking quizzes" and "generating routine letters".
And what really annoys me about this is that it's a perfect example of where simple automation could be used without the need for AI.
The precise example in the article is "Generate a letter to parents about a head lice outbreak." - which is a fairly common thing to happen in schools. So why on earth isn't there one standard letter per school, if not one standard letter for the whole country, that can be reused by absolutely everyone whenever this standard event happens? Why does this require AI to generate a new one every time, rather than just being a standard email that gets sent?
Same with marking quizzes. If children get multiple-choice quizzes regularly across all schools, and marking them uses precious teacher time, why is there not a standard piece of software, paid for once (or written once internally) which enables all children to do quizzes in a standard way, and get them marked automatically?
If we're investing a bunch of money into automating the various processes that teachers spend far too much time on, start with simple automation, which is cheap, easy, and reliable.
Also, wouldn't it be sensible to do some research into how accurately AI marks homework *before* you tell teachers to use it to do that? Here's some research from February which shows that its agreement with examiners was only 0.61 (where 1.00 would be perfect agreement). So I'm sceptical about the quality of the marking it's going to be doing...
And what really annoys me about this is that it's a perfect example of where simple automation could be used without the need for AI.
The precise example in the article is "Generate a letter to parents about a head lice outbreak." - which is a fairly common thing to happen in schools. So why on earth isn't there one standard letter per school, if not one standard letter for the whole country, that can be reused by absolutely everyone whenever this standard event happens? Why does this require AI to generate a new one every time, rather than just being a standard email that gets sent?
Same with marking quizzes. If children get multiple-choice quizzes regularly across all schools, and marking them uses precious teacher time, why is there not a standard piece of software, paid for once (or written once internally) which enables all children to do quizzes in a standard way, and get them marked automatically?
If we're investing a bunch of money into automating the various processes that teachers spend far too much time on, start with simple automation, which is cheap, easy, and reliable.
Also, wouldn't it be sensible to do some research into how accurately AI marks homework *before* you tell teachers to use it to do that? Here's some research from February which shows that its agreement with examiners was only 0.61 (where 1.00 would be perfect agreement). So I'm sceptical about the quality of the marking it's going to be doing...
no subject
Date: 2025-06-10 08:11 pm (UTC)no subject
Date: 2025-06-11 01:40 pm (UTC)(A hack to beat your description would be to mark every circle.)
no subject
Date: 2025-06-11 03:08 pm (UTC)no subject
Date: 2025-06-10 08:17 pm (UTC)Large language models are only accurate by accident. They're like the old game show "family feud," which gave points for guessing the most popular answers to survey questions.
The difference is, the show wasn't asking factual questions like "what's the capital of Russia?" or "find the hypotenuse of this triangle," they were subjective or specific to the individual. Things like "where do you eat breakfast?" or "where did you spend your last vacation?" An LLM might do well in that game.
no subject
Date: 2025-06-10 08:43 pm (UTC)I don't know much about educational resources in primary schools, but in HE lecturers are likely to have access to various packages that can automatically mark simple quizzes.
At a guess, AI allows for easy, fast customisation of quizzes to suit the needs of particular classes -- or even of individuals within classes -- and can then mark the results.
Re: standardised quizzes. That does remind me of Japan, where all children have the same lessons; the curriculum is even more centralised and standardized than the English one. Not saying standard software wouldn't be useful or timesaving. But it also seems a step in the direction of making teachers more passive and offering them less initiative and encouraging dependence. A criticism which has also been levelled at AI.
no subject
Date: 2025-06-10 09:16 pm (UTC)Our government has a similar policy planned: invest in AI to help teachers mark/make quizzes and lessons, as well as train our students to use AI (in areas such as making prompts, how AIs work or their impact on data management and the environment). However, our students still don't have IT classes in their curriculum, we don't have any kind of standardized tests we could use for quick evaluation (that would be really useful for new students who don't speak French for example), there is no central database for resources shared by fellow teachers, we have two computers in the teacher's room that we have to share. There's a 20 million euros budget for this.
no subject
Date: 2025-06-11 01:14 am (UTC)The attempt to apply AI to problems that could be automated more accurately with a more simple method has been really annoying. I know everyone's bosses told them to go use AI, but they should have added "but not for stupid things."
no subject
Date: 2025-06-11 02:42 am (UTC)"AI" (the "A" is correct but I dispute the "I", hence quotes) is very much the wrong tool for those jobs. It's one thing to say that when the only tool you have is a hammer everything looks like a nail, but in the case of this craze they are throwing away the real tools and then using, I dunno, maybe a kitchen spatula to pound nails. It's appalling.
no subject
Date: 2025-06-11 03:21 am (UTC)There is, it's called a Scantron.
Though I gotta say, I'm not a fan of multiple choice tests. What I'm a fan of is smaller class sizes and higher teacher pay.
no subject
Date: 2025-06-11 01:54 pm (UTC)Here in the UK you can buy your child smaller classes, but that tends to reduce the teacher's pay !
no subject
Date: 2025-06-11 04:16 am (UTC)I think that's fantasy though, because if you can't manage to work with a bunch of humans to explain/discover what your actual requirements are, you will have even less luck in getting what you want from any "AI" model. At a certain point of complexity, I would say the difference between "prompt engineering" and "coding" essentially vanishes.
Software development (and operational care) is a misunderstood and probably underestimated craft by most non professionals, and the talent, patience, mindset, training, learning etc. necessary to understand it is not quick or easy to acquire. Like so many many jobs/professions/crafts. The spirit of the age has a powerful minority that don't want to rely on "experts", and a larger population that want to believe they can understand all the jobs in the world around them, that 'anyone can do it" - but that's fantasy too.
As is proved by these nonsense suggestions for teachers. The people suggesting this don't understand LLMs, software or teaching. The beauty of a human society SHOULD BE THAT they don't need to - IF they work WITH the specialists and experts who DO understand.
no subject
Date: 2025-06-11 04:31 am (UTC)Which is done also for most online IT exams. Because people sharing answers to any set list of questions is, and always has been, a problem.
It took them 6 weeks of analysis to create their requirements documents for us to work with, 3 or 4 big meetings to discuss, took us (me) 10 days to create requirements (also talking to them during it), and I can't recall the development time but I think it was 1-3 of us (mostly me) for at least 2 months plus consulting the CAA as required, some weekly time from a project manager. A month of acceptance testing would have been standard but that's elapsed, not continuous. They supported it with in house staff after delivery.
I can't recall (and may never have known) the exact cost, this was in 1997/98. We were consultants. My time was charged out for about £500/day. The requirements were a solid dedicated block of time, I do know that so they cost 50k in themselves exclusive of the initial meetings.
no subject
Date: 2025-06-11 01:44 pm (UTC)Is that a human or an AI joke ?
no subject
Date: 2025-06-11 02:10 pm (UTC)no subject
Date: 2025-06-11 10:48 pm (UTC)Generating quizzes - which would still need to be reviewed for topic and accuracy, but if the AI will generate 30 useful questions out of 40 and the teacher only has to generate the last 10, that's a time-saver. And that's a case where the Autocorrect Generator is fine: For a lot of quizzes, you want generic, predictable (to someone who knows the subject) content.
Grading quizzes... I'm less sure of, but I can see the point of "we don't have an exact standardized method for asking questions/getting answers." If they're not using scannable answer sheets... sure they could point AI at them. Maybe. I am not sure if AI's accuracy problems are less time-consuming (i.e. less expensive) to fix than creating custom software, but it's possible it works better. However, it's not going to be "point the AI at the problem and it's done!" It'd be closer to, "point the AI at the problem and it has saved the teacher 30-60% of the time they used to spend on this task." Valuable, but not "we can cut teacher salaries by 20% because they're working less!" It's not saving teacher standard time; it's reducing the amount of unpaid overtime they were doing.
Generating routine letters - schools should indeed have a boilerplate letter for "lice outbreak"; that's common enough that every school knows it may have to deal with it at some point. They may not have a boilerplate letter for pinkeye, or, sigh, measles. If the AI can generate one - again, and then get reviewed by a human - that may save time.
Does it take long to generate a generic "we've had an outbreak of X; please check your child" etc letter? Not really. But it takes mental clarity to do it well, without revealing any kid's specific details, and to explain the health risks without coming across as condescending or accusatory. The AI has not been dealing with a wave of sick kids all day and some frantic calls to public health services, and can write the base letter without getting emotional.
And then a human can look it over, remove any bizarre hallucinations (hey bot, I said pinkeye/conjunctivitis, not covid), and polish the remaining language. Use of the AI may have saved anywhere from 10 minutes to an hour.
...It has not removed the need for teachers, secretaries, office assistants, or any other human staff member.
Formulaic generic content is what the ai bots are good at. Schools involve plenty of formulaic generic content.
...All of that has to be managed by actual people with brains and skills, and AI is not remotely going to remove the amount of effort involved with that. If it's done well (...it will not be done well), it could allow teachers to spend less time on boilerplate activities and more on the individualized support that students need.
no subject
Date: 2025-06-16 12:44 pm (UTC)[fill in blank] health concerns,
warnings to not have standing water lest mosquitoes spawn,
alert about holidays or school events -
and the blank was mimeo'd - or when copiers were available - were copied.
Tests were given a quick visual scan for completion - and this was even WITH classes over-stuffed with 30 or 40+ students - then usually graded in the teacher's office hour.
There were a LOT less tests, back in the 1960's & 1970's, it's true, but it was still possible to learn. ^_^