I.) Against the Logic of the Machine
The headline asked "if airpods can translate instantly, why learn a language?"
For this, I have an answer, based on my current habit of reading elementary-school level comics in a second language. Weird things are weirder, funny things are funnier, and interesting things are more interesting when I process them in a non-native language. Hell, invert that: cliches no don't hurt to read in an L2, as they are idioms instead.
Spending time in an alternate system of signification is one of the great joys in my life, and I have no desire to give that up to the machine.
II.) Using the Logic of the Machine
I have a few days off for Fall Break. The first day, I was too exhausted to do much. But one problem I worked on, and had a breakthrough in, was how to batch convert questions in a certain format on google sheets into google forms. I think this will save me at least 50% of my time on the front end of making assignments, and 90% of my time on grading.
This is me using the logic of the tools, so that I can not waste motion and energy, and also live more presently in the moment. Society is organized as a panopticon now. I might as well set up my own boutique panopticon within the panopticon.
Other benefits include students not losing papers and the ability to access make-up work anywhere they get on the internet. For me, this means I will stop trying to grade during class itself, and instead I will be watching more, more adaptable, responsive, and of course able to build better dossiers on student misbehavior.
I should point out that I could save even more time by just outsourcing this to the built-in google Gemini, but if I did that I would lose quality control, or ability to easily merge in existing questions from other sources. The Gemini > Forms pipeline proved there is some value in making such self-grading assignments, but like all things from the Modern Silicon Valley, it is preferred the tools only be used the way they envision.
I don't really care to get too bogged down in the details, just that I am happy that my computer nerdery allowed me to find something that is going to work -- relatively speaking.
===
Appendix.
I have seen it argued that AI just moves bottlenecks around. I don't think it "just" moves them around -- I mean, that's not all it does -- but it does move the bottlenecks.
I found that after the wait of a week, it was a bit on the frustrating side trying to set back up my little google form factory. So I guess I will go into some details of what I am doing, after all. If nothing else, it will give me notes for setting the system up.
I have to find a spreadsheet called "Form maker." I now have a link to it saved so it doesn't get buried under the pile of forms all dumped into root by every collaborative action that involves me in the Google ecosystem (it's a lot in a school system). My link for this is on a file saved on Desktop because, again, the Desktop metaphor is the best one and control of where information is stored in a nested hierarchy is prerequisite for my sanity -- and the fact that others haven't risen up to protect this leaves me unsure of whether I should grieve lives of quiet desperation or just grieve that the gap between me and everyone else is so vast.
Anyway, ChatGPT kept fucking up how to format the questions, spitting them out in ways that I couldn't actually put them in a spreadsheet, even though it had assured me it would just remember if I called it "spreadsheet format", so I had to work for over a half hour to work on a prompt that would make it consistently do this operation correctly. Fingers crossed that it works in the future. This prompt is now saved on the same document on my desktop. ChatGPT operates under the same "shit is just going to get buried" principle that is now our dystopia's default, so I need to have this on hand to paste in any time I want to set up my form factory.
If I can get this stuff right, this is where the magic can happen, part 1. I can take previous quiz banks and convert them, or have ChatGPT make them... I'll do some analysis of how this isn't really where my time gets saved at the end.
We're not quite done yet with the steps I have to dredge from memory to make this kludged-together thing work. The script I had Datum write to convert the spreadsheet to a form won't work on my school computer, nor my school network -- which I will admit does make solid sense from a cybersecurity perspective -- so I have to get on my home computer to run the script. Next, I rename the Google Form and make my school account an editor on the form.
Still not done. If I don't want this form to be buried forever.Forever.forever under everything else, I have to make a copy of the form -- otherwise, I can't move it. I move this copy over to a folder for current work, as I found that putting it in the quarter folder make already making for too cluttered of an experience. As I get done with forms, they will be moved into that quarter folder (in this case 2nd quarter), so they are available next year to really save me some time.
Oh, and then I modify the settings so it is a quiz and set the answer key for the multiple choice bits. I could have Datum take another crack at writing a script so the answers could be set in automatic-magic-part-1 step, but I am seeing value in using this as an opportunity to learn what the right answers are so I can answer student questions if they come up.
The first three of these forms I made today took me two hours of real time, as I also use time-boxes so chores can get done and my body doesn't suffer from too prolonged of sitting. That was 40 minutes a form. And though that is kinda awful, and almost seems to prove the "just moving bottlenecks" argument, 1) the ergonomics of doing these forms was better than having to think though each question 2) I think I can get the time down with more practice, and especially if the prompting to get the right formatting works the first time.
Still, I'm not sure that I can get it down to less than 12 minutes per form.
That would be quicker than making the forms myself, yes, but it is not quicker on the creation time than having printed textbooks and telling kids to answer the questions at the back of the book. It is not quicker than copying another teacher's existing assignment.
So the time savings really aren't that much on the creation end. Instead, the real time savings come in on the grading end. With the multiple choice questions, we've only had the ability to grade that by machine my entire life, but I can also add short answers, and have an LLM run through them. Now, that quickly becomes AI grading the work of other AI, but that is another story...
And for now, the kids don't have the ability to easily answer the multiple choices with AI. Hopefully that gap isn't closed within the next 1.75 school years. But even if it is, well, countermoves are available. Another story for another time, if I want to tell it.
No comments:
Post a Comment