news (external)

"AI-generated avatars that look and sound like deceased relatives are increasingly popular to console those in mourning, or to hide the deaths of loved ones from the elderly and young children," Viola Zhou reports. What's interesting is the variety of reasons people want the avatars - not just for remembrance, but also to hide the death of relatives from young children. Also: "Lin hopes the bot will become his immortal doppelgänger, speaking on his behalf after his death. 'If my descendants ask 'What was Grandpa Lin Zhi like?' they could just talk to the AI version of myself to find out.'" It's not immortality, exactly; it's more like publishing an interactive autobiography. I need to get around to setting my own up.

This post references some good conference coverage by Phil Hill and considers the question of whether institutions are ready for the collapse of online program management (OPM) services. "Leaders in the sector, including 2U, Coursera and Keypath, never made a profit on the activity, and Pearson and Wiley sold off their OPM offshoots in recent months." The revenue-sharing model was never profitable, according to the articles. Moreover, as colleges acquired the relevant skills (especially during the pandemic) the need to outsource abated. The tech is complex enough that, if you're large enough, you want to do it in-house, if only to stay on top of features and costs.

According to this article, "Post News, a Twitter alternative run by the former Waze CEO Noam Bardin, is shutting down. Bardin says the platform 'is not growing fast enough.'" It was backed by venture capitalists Andreesen-Horowitz. It was "a social platform that also offers users ad-free access to paywalled content from publishers such as Fortune, Business Insider, Wired, The Boston Globe, and others." What they missed, I think, is that people not only want to read, they want to share, but they only place users could share was on Post News itself; otherwise, you'd just be sharing a paywall, and nobody wants that.

"The time has come to break free from the shackles of 'best practices' and embrace the power of effectiveness driven by the true experts in education—the schools and educators who implemented these strategies consistently and with a high degree of fidelity." Certainly there has been criticism of the idea of 'best practices' over the years for precisely this sort of reason. But are 'the schools and educators who implemented these strategies' really 'the true experts'? Certainly they would have valuable feedback, as would any practitioner. But they do not assess their practices scientifically, and they are unable to view outside their own context. The problem with 'best practices' isn't that we're talking to the wrong people - it's that we're asking the wrong question. Image: Perceptyx.

This is just a note to remind readers in Canada that graduate students - who teach a significant proportion of university classes - are expected to live on roughly $15-$25 thousand dollars a year. The explanation is that 'work' is 'capped' at 10 hours a week, so it's a good hourly rate, but nobody believes people teaching university classes are actually working only 10 hours a week. You can read about the strike at Western University here and here. As a graduate student association president some 25 years ago I worked actively on this very issue, and it's disappointing to see it persists to this day.

This post is directed toward ed tech companies and makes the point that "in a fragmented impact ecosystem, ed-tech needs collaboration to prioritize education over technology." In other words, "for a technology to count as educational, the market needs to be run as a partnership industry, where developers, educators, researchers, and students actively work together to develop, implement, and scale what works." The article makes four specific recommendations that seem reasonable to me (though I word them a bit differently): first, the company's self-interest needs to be subordinate to "the 5Es of impact, efficacy, effectiveness, ethics, equity, and environmental impact"; second, "focus more on the quality of evidence rather than solely on the type of evidence"; third, "contribute new ideas for impact measurement and understanding", and fourth, do more than just query practitioners for new ideas.

Meta (aka Facebook) has just released its new AI assistant, Meta AI, Built With Llama 3. Time Bray asks it a simple question, which it gets very wrong. "The problem isn't that these answers are really, really wrong (which they are). The problem is that they are terrifyingly plausible, and presented in a tone of serene confidence." I asked it a question about myself and got a short answer that wasn't wrong so much as very misrepresentative of my actual beliefs (see the image). Note that I didn't need to sign in to Facebook to use it (though I'm sure this will change).

Though this is mostly an exercise in taxonomy, and though it is also badly titled, this post on what Carlos Ortegon, Matthias Decuypere, and Ben Williamson call 'edtech brokers' is an interesting glimpse into an infrequently-discussed branch of the field. Edtech brokers position themselves between educational institutions and the (usually commercial) vendors and services that support them. The authors identify three types of edtech brokers: ambassadors, which act as representatives for specific brands; service engines, that function as search portals offering such things as 'what works' indices; and data brokers, that manage data flows between institutions and vendors. They mediate edtech in three ways: by supporting infrastructure building and standards development, by producing evidence of 'impact' and 'efficacy', and by 'professionally shaping' via development and training programs. I think both taxonomies could be extended with a little thought. See the full paper, Mediating educational technologies (17 page PDF).

According to the authors, "studies on teacher-student interaction (TSI) support tools often focus on teacher needs while neglecting student needs and autonomy." This raises the question of how to enable student needs to be expressed during a class session. They describe the development and testing of a tool called NaMemo2 (built on NaMemo, a tool for remembering student names) to address this need. In so doing they propose a TSI framework called STUDIER (i.e., Sparking, Targeting & Understanding, Designing & Implementing, Evaluating & Refining). NaMemo2 is based on an augmented reality (AR) tool "that allows students to convey their willingness to interact to the teacher as well as show their names to the teacher in physical classrooms." 36 page PDF.

Beatriz Toscano writes about the Kinderstadt or Ottopia project in Magdeburg, Germany. You can read more about it here (the article is in German but easily translated using your browser). "For two weeks, up to 450 children in around 40 trades were able to pursue professions, start-ups, learn at the Children's University – and rule the city." If you want to learn about democracy, the best way is to practice democracy. "Do systems built by the so-called innocent lead to more just and equal societies, or do they devolve into selfishness and corruption?" asks Toscano. "Ottopia offers a playground for these experiments, with children experiencing governance firsthand."

This article describes a project called Student Hubs, which "exists to empower university students to become active citizens, equipping them with the tools, behaviours, and skills they need to make a positive change." In addition to the sorts of skills and attitudes needed to become successful later in life, a project like this also builds connections between the students with each other and the community, which can be vital in creating opportunities for further employment and personal development. This is the sort of thing elite universities excel at (Student Hubs began at Oxford and was inspired by a similar project at Cambridge) and what we should be thinking about beyond simple grades and content knowledge when we talk about equity in education.

The scikit-learn module is a set of machine learning algorithms for Python, and scikit-lego is built on top of that. But that's not the important bit here; no, what we have is an engaging story of how the author got engaged with and learned about scikit-lego. "I often found/find myself looking up at the source code of the libraries I use, trying to understand how specific features work and how they are implemented," writes Francesco Bruzzesi. "This is certainly not necessary, but it works for me as a way to learn and understand better the tools I use when in doubt about something." It's an example of advice I gave on Reddit recently.

Alex Usher interviews his "favourite commentator on all issues related to educational technology and higher education institutions," Phil Hill. The best bit is this, where Usher comments: "25 percent of post-secondary enrollments in the U. S. are now fully online and another 25 percent or so are at least partially online. I had no idea the market was that big." But they are, and as Phil Hill makes clear, a big part of the reason was MOOCs. "MOOCs, along with 2U's initial business model, those were the two things that forced traditional higher education to get over themselves and say, we need to take online education seriously." The two of them are far more interested in markets and money than I am, but on these they know their stuff. Image: Phil Hill.

Your college degree, sponsored by Google? " Interim (University of Nebraska) President Chris Kabourek announced Tuesday that the university will soon offer Google Career Certificates in a variety of fields... Google experts teach the programs, which are vetted by leading employers. NU students, alumni and Nebraska residents can get a special first-year rate of $20 per enrollment." According to Kabourek, "the new partnership advances a 2022 legislative goal, which NU supported, to increase the percentage of Nebraskans with postsecondary credentials by 2030 to 70%." I'm not sure that's what the legislature meant, but these days, who can be sure? Via The 74.

Offering a little light reading, this report (502 page PDF) "tracks, collates, distills, and visualizes data related to artificial intelligence (AI)." The authors ease you in; you can read the 'top 10 takeaways' (p. 5), chapter-by-chapter 'report highlights' (p. 14ff), or read the chapters themselves - each page is almost like an individual slide with some key points and a graph or table illustration, while some coloured 'highlight' pages (like 'Will Models Run Out of Data?', p. 52) are interspersed. It's worth the time to just flip through this document, where you'll find everything about AI considered, including, for example, "their capacity for moral reasoning, especially moral reasoning that aligns with human moral judgments, is less understood... Of all models surveyed, GPT-4 showed the greatest agreement with human moral sentiments," a statement that is remarkable on several dimensions (p. 122 ).

You may not think of children's vaccines as educational technology, but I think of it as the second educational technology, following only the first: proper nutrition. This tech stuff we do? That only helps once we've addressed the rally major issues created by child poverty and misinformation about health. I'm old enought that I got all the diseases when I was a kid - measles, German measles, chickenpox, mumps. Everybody got them, and a few of the kids died. What would have been better than running that gauntlet? Immunization. "We have forgotten how many children used to die before their fifth birthday," writes Kristen Panthagani. I haven't forgotten. That's why I get a flu shot every year and make sure I'm up on my Covid shots. I've remained flu-free for years and years, and entirely covid-free. I hope to stay that way. Via Robin DeRosa.

It's not that this article is wrong. But as Gerald Ardito comments, "there is more than a little irony in this post, given EdSurge's long time tech boosterism." Tanner Higgin writes, "Within this technocentrist frame, education is sick and edtech is like medicine." How many studies do we see like this, asking "what works?" in education, as though the same goal were shared by all. But as Seymour Papert writes, "The content for human development is always a culture, never an isolated technology." And in particular, "The potential of technology is significantly affected by the humans that use it and their context."

This short article references a 2022 paper that studies how the order of questions in a test impacts how well the test-takers do. "A perfectly logical test-taker should do equally well no matter the order of questions," writes the author, "but research shows that humans are influenced by the order." In particular, "students answered more... questions (correctly) when they began with easy questions compared to when they started with difficult ones." The relevant question to ask here is this: if the result depends so much on the order of the questions, what is it exactly that the test is measuring?

I often ask what it is we're doing when we're teaching and learning, that is, what counts as success? This paper offers some insight into the sort of question I have in mind. To 'learn' something is more than just to come to know that something is true. That's just memorization. No, as David Bourget would argue here, it's to grasp what is being taught. "It is one thing to believe something, and it is another to grasp it. For example, everyone knows that life is short, but most of us arguably do not fully grasp this fact." But what is it to 'grasp'? He argues, "we grasp to the extent that our thoughts are grounded in experience, whether occurrent or non-occurrent... , what we experience matters to how we reason because that is how we are wired: consciousness isn't a late addition to our minds; it is the
most central, causally potent form of mental activity." I think this argument works, overall. Image: Ding.

This paper (21 page PDF) is a review of papers proposing AI competence frameworks for teachers (CFT) and organizes them into five categories: integrating AI competencies in existing CFT; modelling new AI knowledge domains; process-driven; AI systems-driven; and competence level-driven. True to their field, the authors stress the need to theorize. "The empirical and design-based nature of AIED requires a solid theoretical foundation. The adoption of theoretical frameworks serves as a unifying force, fostering shared concepts and terminology among researchers and designers." But it's not clear there is (or is going to be) agreement on just what to theorize. This depends on what we want teachers to do, and as the authors note, "it is important to understand what kind of problems AI teachers' competencies will solve and what kind of solutions AI teachers' competencies convey."