Stephen Downes

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.

This is probably the best actual argument against learning styles I've seen, though the author does not address this issue in any way. Instead, Hollis Robbins is focused on the 'Big Five' personality traits: openness, conscientiousness, extraversion, agreeableness, and neuroticism. As Robbins explains, the taxonomy has its origin in Gordon Allport and Henry Odbert's 1936 Trait-names: A psycho-lexical study (176 page PDF) (interesting to me already because of my familiarity with Allport's The Individual and His Religion, which I studied in the 1980s). Through various processes of refinement these were reduced to the five, known as Norman's five. Long story short, Robbins considers various critiques, runs the Allport-Odbert list through Claude Fable, which concludes that "the compression didn't lose the relational, historical, addressed, and valuing vocabulary by accident at step forty; it discarded it at step one, as category error, and kept the words that predicate a lone individual in a vacuum." In other words, "character is built via a sequence of encounters in language, not an isolated trait item... calling correlations of self-descriptions a model of personality has always been a mistake."

As Mike Taylor writes, "Every year, Jane Hart asks people around the world to share the tools they use to learn. I always enjoy seeing the results." This is my list. I'm not sure how to communicate them to Jane Hart (her site is offline for a few days) but I'm always very much an outlier, so it probably doesn't matter. This will be her last year doing this, apparently. See also: Clark Quinn.