maegul (he/they)

Ha yes … on the other hand, it was easy to forget how good damn expansive non-internet information is: the whole world ran on that shit for millennia.

Oh I hear you (and appreciate the response).

For me, I can’t help but think of another alternative, which I’m surprised I haven’t heard of yet …

stripping down one’s personal technological cognitive load to a stack of systems that can fit into one’s brain (like the Python mantra), focusing on learning that stack well building sustainable and stable systems, and then just detoxing from the increasingly polluted digital information stream (protected commons, traditional formats such as books and in person engagement … dunno).

Depends on what the end goal is, but AIs seem to be about using tech more or just opting out of sovereignty. Something like the above seems to me to be about using tech less (in the end) and pushing toward being a secondary tool rather than an end of its own.

Probably a shallow response …

But I always figured AI/LLMs are basically apocalyptic for all sorts of individualistic values in computing (including privacy but also independence and diversity).

Whether they’re good or useful etc, I just struggle to see how they will ever be justifiable against these sorts of values.

Sure, local models and our hardware will get better … but better than the state of the art from the big labs and providers? Given that data and training are the big bottlenecks on quality … I struggle to see how AI isn’t a complete feudal capturing of information computing and processing. Not to mention what happens to the pipeline that produces information content if everyone is only consuming it through the models that train on it.

So for me the big question is, what’s our call on a possible (likely even?) future where we are forever stuck using cloud provided AI along with all of its negatives, in the same way that basically all of us has been and still is stuck using MS windows, Google and the big-social-media hellscape?

For me, I baulk at this.

Thank you and interesting!

I’ve only skimmed the paper, but this line in the conclusion captures my impression so far:

the results should not lead to premature decisions in school practice or completely replace other existing teaching methods. In fact, digital tools show the largest positive effects on student learning outcomes if they are used in addition to non-digital material. Despite the potential of using digital tools in mathematics and science classes, teachers should always assess additional benefits in regard to the context they want to use it in, and learning content should still take center stage

…

there was only one variable that significantly influenced the overall effect due to differences between content-related categories, which was teacher training.

Additionally, their analysis highlights that the following are more impactful: simulation/ smart-tutorial tools, getting students working together (rather than using tools solo), and shorter durations of tool use (or, studies that ran for shorter times had better results).

All of which indicates to me that this study may well support the notion that digital tool use in schooling can be overdone and that a correction could very well be reasonable, especially if prior policies have focused on student-laptop provisioning all encompassing digital platforms.

Beyond all of that (and general scepticism I’d naturally have with any reaearch) … I’d still wonder whether it’s reasonable to combat the negative effects of saturated computer usage by leaning into non-digital education approaches, however “worse” the educational outcomes may be. Especially if digital education could be optimised with specialised and intermittent exercises and tools.

Interestingly, I don’t share your presumption. Personally, I’d start from the opposite end and ask where is the evidence that using computers is good for education. Anything that’s computer specific can be taught with computers as necessary. But generally everyone has computers now and so the basics likely don’t need to be covered by forcing complete use across the board.