Stephen Downes

I appreciate that Creative Commons is undertaking a good-faith effort to wrestle with some of the contradictions the emergence of AI has created around open licenses and open content. They conclude, "We believe in a thriving commons above all else. We believe that for a thriving commons to exist, reciprocity is required to sustain it. We believe that to encourage people to continue a full and sincere embrace of open sharing practices in this new world order, they have to have some agency. We continue to believe that copyright is not the hammer for every nail, and that new tools are needed. And we continue to believe that none of the experiments should come at the expense of public-interest uses, which must be strongly protected." I don't agree with the idea of reciprocity - as I've said before, openness and sharing are not transactions. And I don't want the model of a knowledge commons to be a transactional one.

According to the website, "We are piloting a knowledge-building library for AI agents in learning engineering. This hub curates evidence-based research, tools, datasets, and policy resources from trusted sources like What Works Clearinghouse, Evidence for ESSA, Mathematica, and the Learning Policy Institute - structured so AI agents can search, filter, and cite them directly." I can seethe usefulness of this, but once again a quick look at the literature raises more questions than answers. Why are there 169 frameworks, for example? This strikes me as a discipline less like engineering and more like creative arts. Here's the GitHub version.

This is an interview with Tom Sayer from Ello about their effort to build what he calls an AI teacher.The premise of the article is that since teaching requires mastering many little things, all of which go into the overall task of personalizing instruction, it follows that an AI teacher won't be the product of a frontier model, but rather an assemblage of more specialist models. I don't know about that, and in the end the difference probably doesn't matter anyway. I do think from reading this article that Ello is informed by very traditionalist pedagogy. For example: "At Ello, we focus more on core skills in reading, math, and language learning. We'd love to nurture creativity alongside these core skills, but right now, we just want to teach a kid to read. That involves drilling phonics, but it also means engaging a readers' comprehension too, and developing their vocabulary; all the parts of the reading rope."

This is a long article with the challenging idea that an 'agentic professor' is something that can and maybe should exist. What I like is that the authors offer a test of the concept in the middle of the article; you can copy their prompt into an AI of your choice and see what it does. "What distinguishes the Agentic Professor envisioned here from such present-day interactions is not any single function but the combination of capabilities sustained over time: agency, longitudinal memory, pedagogical plurality, cross-disciplinary synthesis, and continuous refinement. Its significance lies not simply in providing individualized support but in creating continuity across a student's educational experience, connecting ideas, skills, and feedback across courses, disciplines, and years of study." I don't think they're wrong in the sense that something like the agentic professor is probably achievable. But I wonder if an agentic professor is what learners and people generally are actually looking for.

According to this article, technologists need to study more of the liberal arts, while those in the liberal arts need to embrace technology. "By suggesting that we need to fuse technology and the liberal arts, we are explicitly affirming the need to balance topics like economic growth, political expediency, and human desires with the humanistic insights developed over the centuries." Otherwise, "technology's effects could well be harmful." Even assuming anyone has the time to study all that (and over my lifetime I have genuinely tried) it's not clear that any genuinely deep civilization-saving insight results. We already know that systems devoted to the single-minded pursuit of anything - whether paperclips or money - lead to undesirable consequences, but we build them anyway. But more, I just don't see this argument as particularly deep. Sure, we should pool out knowledge. But then what? Do we even know what we want and what we reasonably fear? I don't get a sense of that here. Via Mark Oehlert.

This article follows a well-established pattern for this sort of writing: a new capability is described, we get some examples of it at work, then we read some cautionary remarks about misusing the capability. In this case, the capability is very low cost automated assessment, courtesy System One + Jev. It uses a grading rubric and natural language student text as input, and saves money by avoiding costly next-token text generation in its output. The caution: "When you write to please a machine, you enter into a feedback loop that may not benefit your learning. While some writing tasks are straightforward and don't necessarily require much process work, the type of long-form and engaged writing many college courses call upon asks students to develop their responses via close reading of primary and secondary sources, multiple drafts, peer review, instructor feedback, and above all, time."