Stephen Downes

"AI technology tools hold remarkable promise for providing more accessible, equitable, and inclusive learning experiences for students with disabilities," writes the author, and this article does the service of listing some of them: automated image descriptions, audio description generation, support for cognitive and physical disabilities, inclusive design support, translations, captions, lip reading, and speech recognition. Each of these is described in some detail with examples.

Alex Usher offers a brief summary of the OECD's 'Education at a Glance' report. It's probably this report, though he does not bother to offer a link. There's not much that's groundbreaking in the report. I read far too much of it trying to figure out the 'public institution' data, which Usher (nor without justification) reports as "nonsense" - it reports 100% of Canada's and 0% of the U.K.'s institutions as being 'public institutions'. Now of course the OECD does not actually believe this - there are other tables in the report (eg. B.3.4) that show different percentages. I assume Usher is referencing this table (though he doesn't bother to reference his reconstruction, so I don't know for sure (just sloppy work here, he can do better)). Now I can't explain this table either (it actually shows Canada, Greece, Luxembourg and Denmark all at 100%, and doesn't include the U.K. at all).

This article has the magic word in it that always gets the attention of the media - "Harvard". The gist has been reported in other studies as well over the last few months: "Students were randomly assigned to learn a topic as usual in class, or stay 'home' in their dorm and learn it through an AI tutor powered by ChatGPT... Each student learned both ways, and for both lessons – one on surface tension and one on fluid flow –  the AI-tutored students learned a lot more." The study? "Kestin provocatively titled his paper about the experiment, AI Tutoring Outperforms Active Learning, but in an interview he told me that he doesn't mean to suggest that AI should replace professors or traditional in-person classes."

Dan Gillmor writes about this item: "The push for an 'age verification' requirement on the Internet is 1% aimed at protecting children and 99% aimed at controlling speech. It effectively bans anonymity, for starters. " The website itself expresses "serious concerns about the suitability, proportionality, and negative impact on fundamental rights of current age verification proposals." My feeling is that there ought to be a way to create a zero-knowledge proof of age. Though the failure to implement such a system seems to vindicate Gillmor's argument.

This is an excellent article that resonates with me in so many ways (not the least of which is that it is a long read from the Washington Post, where articles are normally locked behind paywalls - here's the same text in the Mastodon thread where I found it). It's the story of Chris Mark, the man who analyzed the data to solve the problem of coal mine roof collapse, and of the regulations needed to make the mines safe. But that's just the story - the value here is the meaning behind the story. "The father's work had received a lot of public attention and the son's had not. But that was just an accident of what people cared about. A lot of people cared about Gothic cathedrals; fewer were concerned with whatever was happening to workers deep underground.

This is why you should always read the papers you cite in your article. Well, I mean, you should read them anyway, but this is one more reason you should read them. Or, I suppose, more accurately, this is why you should read all the citations the AI puts in the paper it's authoring for you. Because some AIs make stuff up (which, frankly, is what they were designed to do).