Stephen Downes

Really interesting paper (57 page PDF) that compares how well differently organized agents work together in a shared communication, culture, and world (which "acted as an external memory and transmission medium") or as unconnected individuals.  "The experiments support a bounded form of swarm advantage: interaction is most valuable when performance depends on building and maintaining a technological ecology, not when the objective is only to find one record-setting object." Also, "Emergence appeared in organization as well as outcome. No agent was assigned to be an explorer, builder, maintainer, or cultural broker," yet these specializations nonetheless appeared. "The mature society was neither a globally synchronized collective nor a collection of independent agents. It formed persistent, modular technological neighborhoods with a limited set of local hubs and cross-community connectors." Via Ross Dawson.

This feels like an introduction to the blog as its first and only post. It states its position clearly in its coverage of the new researchED and AEI partnership. It reveals how they "see the American education system: as a growth market for technocratic solutions that seek to reduce the cost and complexity of teaching by removing human beings, a decades-long political aim of AEI, now given the imprimatur of science. While American schools face decades of compounding crises and inequities, 'The Practice Project' offers privatized solutions to systemic problems that any school leader can 'add to cart' and solve with a subscription." This approach favours direct instruction and opposes hallmarks of modern pedagogy such as inquiry-based learning and transferable skills (such as critical thinking). Via Alfie Kohn.

The intent of this article is to introduce a 'bot check' method as a means of using AI to develop agency in learners. It takes a bit of reading to get to this point, as Christyl Lucille Murray first offers arguments against both AI abstinence and whole-scale adoption as strategies, offering the 'bot check' approach as a middle ground. The idea is to begin with human reason and collaboration, then to use the AI to check the thinking, befor implementing the result. The method, once described, is then assessed against the SHINE framework.

This is a report (40 page PDF) from an Ad Hoc Committee on AI use at MIT containing two major parts: a set of 'guiding principles' and a (much more comprehensive) set of recommendations. The recommendations have three sections: adapting education to an AI world, center the human community, and building review processes. These (in my cyclical view) address MIT's two major issues: how to make sure what it teaches remains relevant, and how to maintain MIT's exclusivity in a world where learning becomes that much more attainable. That's why (to my observation) the learning becomes more focused on personal experience (eg., out-of-class activities, more lab spaces) and the college community (value of residential education, strengthen social connections). These are the right recommendations, I think, but I don't think we need an MIT to do this. Via Fredrik Graver.

This article is one of a number (here's another one, from Doug Johnson) documenting the pushback to educational technology in schools, and especially measures limiting or blocking 'screen time' for younger children. "None of this comes as a surprise to me," writes Tim Stahmer. "Teachers and school administrators were starting to sour on edtech even before the pandemic. The panic-driven, poorly planned effort to take school online during that period, along with justifiable concerns over social media and AI, only accelerated the effort to remove technology from classrooms. Plus there's the glaring lack of evidence that the huge amounts of money spent on technology was actually improving student learning." 

I don't really deal in sales figures and pricing tables - that's the sort of grist for a corporate AI, not something based as I am in the humanities and (to a lesser degree) social sciences. So a lot of this article is orthogonal to what I would be looking for in a description of how to make my data AI-ready (assuming I have anything that would be considered 'data' in the first place). And yet, the model offered here by Pramod Sadalage and Prem Chandrasekaran is, with a little imagination, sufficiently generalizable to offer guidance to the rest of us. What we want, they argue, is a framework that supplies "the context, judgment, and skepticism to work around data that's incomplete or wrong." That seems like good advice even for a database filled with hot takes on learning technology. My readers are certainly grounded and sceptical, but I can't assume an AI will be.