edtechdev

Public domain auto-updating knowledge base of recent open-access research on AI in educational contexts. Currently has ingested 950 research articles, synthesized into 180 concept pages. Static astro 7 site generated with Hermes and DeepSeek Flash-0731. Source and skill files: https://github.com/edtechdev/aied

Here's research showing the same exact effect 15 years ago, except it's about homework copying: https://news.mit.edu/2010/homework-copying-0318 Students who copied their homework more saw their grades go down more throughout the semester. A/B students became C/D students.

A name people use for this phenomenon now with AI is cognitive offloading.

I coincidentally took advantage of the cheap DeepSeek Flash pricing the past couple of weeks to compile & synthesize a knowledge base/wiki of hundreds of educational AI research studies like this one here: https://edtechdev.github.io/aied/ Here's the summary of this article: https://edtechdev.github.io/aied/articles/stromberg-generati...

Or get an AMD 395 laptop or mini PC for half the price of an equivalent mac device

which explains why this tool requires a NEAR AI account to use

I tried this out on huggingface, and it has the same issue as every other multimodal AI OCR option (including MinerU, olmOCR, Gemini, ChatGPT, ...). It ignores pictures, charts, and other visual elements in a document, even though the models are pretty good at describing images and charts by themselves. What this means is that you can't use these tools yet to create fully accessible alternatives to PDFs.

Cognitive load isn't a valid or useful concept: https://edtechdev.wordpress.com/2009/11/16/cognitive-load-th... https://www.tandfonline.com/doi/full/10.1080/00131857.2024.2...

There are separate contexts involved here: the coder, the compiler, the runtime, a person trying to understand the code (context of this article), etc. What's better for one context may not be better for another, and programming languages favor certain contexts over others.

In this case, since programming languages primarily favor making things easier for the compiler and have barely improved their design and usability in 50 years, both coders and readers should employ third party tools to assist them. AI can help the reader understand the code and the coder generate clearer documentation and labels, on top of using linters, test driven development, literate documentation practices, etc.

Some research:

Students learn and understand college math more when the classes are contextualized (usually engineering, biology, but you can also use everyday examples). See decades of research on situated learning and related approaches. https://careerladdersproject.org/docs/Contextual%20Approache...

Developmental courses can also be compressed https://blog.careertech.org/research-review-promising-practi...

Dual enrollment saves time and money and improves success. Let high school students take college math courses.

Corequisite remediation is the current best practice. Let students take regular (not remedial) math courses, and improve advising and support. https://ccrc.tc.columbia.edu/easyblog/future-of-corequisite-... https://strongstart.org/resource/corequisite-mathematics-too...

Good to see further development in this space. Would be interesting to see how it compares to Decap CMS https://decapcms.org/ and Static CMS https://www.staticcms.org/

Me personally I'd like to see something that supports easily creating and using different types of objects besides pages (such as: events, books, recipes, etc.), like content types and fields and views in wordpress or drupal, ideally aligned with schema.org like https://www.drupal.org/project/schemadotorg I think Hugo might support content types in YAML or something.

It struggles with basic math: https://www.wsj.com/tech/ai/ai-is-tutoring-students-but-stil...

There are some activitypub apps that support nomadic identity like HubZilla and Streams: https://codeberg.org/streams/streams

Another non-activitypub alternative is nostr, where your identity is a public/private key pair: https://github.com/aljazceru/awesome-nostr

MAA has a free, evidence-based instructional practices guide https://www.maa.org/programs-and-communities/curriculum%20re...

I'm not sure, but maybe GPT4All will eventually add support for it: https://gpt4all.io/

Educational researcher here. There's no such thing as a "science of reading." It's part of the highly politicized "reading wars" (see also the "math wars" which has been going on for decades). It's no coincidence that Republicans are pushing phonics as the end all be all solution to teaching reading, and you can cherry pick educational research studies that support or disconfirm various teaching strategies. Phonics has its place, contexts where it is appropriate and beneficial, but it is not the sole strategy that works or should be used in every context.

A recent meta-analysis https://www.researchgate.net/publication/338494581_Meta-Anal... and the What Works Clearinghouse have summaries of the evidence for different strategies for improving early reading skills: https://ies.ed.gov/ncee/wwc/practiceguide/14 Direct Instruction (also championed by one political side) is not an effective strategy: https://ies.ed.gov/ncee/wwc/EvidenceSnapshot/139

Here's just one post with a little more info on the political context: https://radicalscholarship.com/2023/01/14/does-the-science-o...

A bigger scandal is how states like Florida game the system to make their reading score rankings look higher: https://www.tampabay.com/opinion/2023/01/05/floridas-educati...

The good news is there are a lot of strategies that help with reading in various contexts. There's even research on how reading to dogs (or even robots) helps students with reading :)

Wikipedia is unfortunately biased in many areas that have any kind of controversy/disagreement. Education is one other topic where Wikipedia has a lot of bias. Wikipedia's "neutral point of view" idea is not only false but unobtainable. Persistence wins over logic or evidence.

Would be interesting to see this connected to YouTube, to improve upon their auto generated transcripts. There is this command line version using YouTube-dl and OpenAI's API https://simonwillison.net/2022/Sep/30/action-transcription/

You could do a similar 4GB RAM server setup for around $7/month using the Yunohost server app manager (free) + a VPS server host like Hetzner (or see lowendbox for others).

See the instructions on the bottom of this page, for example: https://joinfediverse.wiki/How_to_host_your_own_Fediverse_in...

Here's a list of all the server apps Yunohost supports: https://yunohost.org/en/apps?q=%2Fapps

I was curious if there are plugins for [[wiki links]] and backlinks. Best I can tell, you'd need to use this wiki link plugin for the remark markdown processor: https://github.com/landakram/remark-wiki-link And then this github action to add backlinks to the markdown files: https://github.com/marketplace/actions/maintain-backlinks-in...

Chemicals from tires are killing off fish, too: https://www.acs.org/content/acs/en/pressroom/presspacs/2022/...

There have already been responses to this criticism before, such as: https://drbenvincent.medium.com/the-dunning-kruger-effect-pr...

including from David Dunning himself https://thepsychologist.bps.org.uk/volume-35/april-2022/dunn...

Logseq is free and open source https://logseq.com/ You have to sync yourself though (for now). You can also publish to github pages.

My only issues are: they are focusing more on the desktop and (now) mobile apps, instead of the web app. You can't quickly share items to the app or web version to quickly save links, etc. They did have github integration built into the web version but are abandoning that to work on their own paid sync solution that isn't out yet. But you can manually sync the files as I mentioned.

For now I'm mainly sticking with a wiki but keeping an eye on developments to logseq and similar open source apps like Athens Research, Bangle, etc. See this twitter list of different knowledge management tools: https://twitter.com/i/lists/1396562498002825240

This Falstad animated circuit simulation has been the best at conceptually understanding what's going on in circuits: https://falstad.com/circuit

Click on the 'Circuits' menu to see dozens of example circuits.

One issue with the hydraulic/fluid analogy is the "empty pipe" misconception - we forget or don't know that in electrical circuits, the circuit is a closed loop. An example of this misconception is that beginners sometimes think the current "wears out" as it goes along the wire. The Falstad simulation shows a line of moving dots that move faster or slower depending on the current - a little more like a train moving in a pipe - which helps counter this misconception, although it, too, isn't perfect. As a next level, I like showing animations/simulations that show the role of charge on the 'outside' of the wire in steering current flow, as well as magnetic fields surrounding the wire.

Cognitive load theory is unfalsifiable and wrong about many things https://edtechdev.wordpress.com/2009/11/16/cognitive-load-th...

Of course we have limits to how much we can do at once, but learning isn't like filling a bucket. Learning takes active work, effort. An analogy would be like saying 'weight lifting is easier with tiny weights' or 'don't lift too much weight.' Yes, but that's obvious, useless, and forgets the presumed goal of weight lifting, which requires putting in some effort. Similarly, you wouldn't force kids to listen to lectures about baseball or other sports and memorizing all the rules before letting them play.

Some of the other things mentioned on the poster article are false or only apply to rote, trivial learning. Remember most of this type of traditional psych research “includes participants who have no specific interest in learning the domain involved and who are also given a very short study time.”

Worked examples, like other passive learning situations, can cause an 'illusion of knowledge' - feeling like you know and understand, but not really.

Look at pre-worked answers, build rote knowledge, etc, before trying to solve problems? We learn more doing the exact opposite - see research on 'productive failure'. We learn best in context, when we have a need to learn.

Our intuitions about learning, teaching, etc, are often the exact opposite of reality. Here are just a couple of examples:

Khan Academy and the effectiveness of science videos https://youtu.be/eVtCO84MDj8

Measuring actual learning versus feeling of learning in response to being actively engaged in the classroom https://www.pnas.org/content/116/39/19251?cid=nwsltrtn&sourc...

TerminusDB is a graph database that supports versioning https://terminusdb.com/