Stephen Downes

I've seen student representation on boards, committees and agencies from both sides. There's truth to what's being argued here, but it's not the whole truth. Emily McIntosh argues here that student representatives feel like they do not matter - no matter what they say, the decisions have already been made without them. McIntosh argues for "Authentic Student Voice (that) is not simply being listened to. It combines influence, ownership, and shared responsibility." I don't disagree with any of that. But if we go back to the beginning of the column we read that the student "had been invited to a significant institutional meeting as the 'Student Voice'." They were 'selected' or 'chosen', presumably because they could be counted on to do the research and to be responsible. Something like that. But an authentic student voice is often much less compliant that an especially selected student voice. They won't take the same truths for granted, and won't share the same background assumptions about presentation and decorum (and not should they; for all the good intentions in promoting student agency, we all know that as soon as agency becomes pushback there is the risk that it will be shut down).

There's a lot of bait to rise to in this article (for example, the economics of having expensive professors teach a mere 25 people at a time) but Steve Krause's response to a paywalled NY Times article has some nice misdirection in it that merits our full attention. Krause is responding to Joe Cruz, who is arguing against the idea that most students learn to think only by writing. The clue to this misdirection is in the 'gotcha' at the end: "isn't it interesting that Cruz decided to write this down?" I'm not sure how else Cruz was going to get paid for a NY Times article. Similarly, Krause notes that Cruz favours "real-time interactions between people talking with each other." Which is also language. Which also involves the same skills of composing series of words, just in a different medium. Which means they are saying - and doing - the same thing, since they both teach using dialogue and composition. To really get at the heart of the idea that learning to think does not necessarily require text we need to get away from institutions (like a university or the NY Times) that are inextricably text-bound. 'Thinking' isn't the same as 'thinking in text'.

This article reminds me of my presentation and article on the role of the educator. Alexandra Mihai is, of course, defending the educator against encroachment by AI. "We are architects... we are hosts... we are role models... we are artists." Etc. My point back then was that each of these is a specialized skill, and it made more sense to think of them as individual roles, and not something one single person could, or ought to, provide. I mean - a classroom teacher has a lot of skills, but are frankly limited by imagination and resources when it comes to "design learning spaces that support learning experiences." Mihai adds three human skills to supplement these roles: attention, intuition and empathy. I think this overstates the unique capacities of humans - we already know AI pays attention (it's in the founding documents). Intuition and empathy (or equivalent artificial states) could easily be next.

Even though there's no formal requirement to register a work to establish its copyright status, "CommonsDB is an EU-funded pilot project that has established a functional public registry of rights information for public domain and openly licensed works." This way people can verify that the open work they are using is indeed openly licensed. As the CommonsDB website states, "The registry allows anyone to check the public domain status of media files found online." But a big centralized registry is an expensive single point of failure, so the plan is to decentralize it. As this strategy paper (31 page PDF) summarizes, "The feasibility study committed the project to federation as a core design principle, and the pilot has delivered on its architectural promise. The registry includes a distributed layer, implemented as an open-source service over the Iroh peer-to-peer protocol." As so many others have discovered, "Identity is the hard part. Registries require proper authentication of the declaring party... Qualified organisational credentials remain rare among the institutions that hold relevant collections." And you need to make it cheap or nearly free to register a work as open or public domain.

The highlight of this post has to be the 'Bitter Lesson' video crafted by AI to explain the concept Ethan Mollick has in mind: "the hard truth, learned over and over again, that things that we thought required elaborate human rules and thinking can be solved with the brute force of better machine learning systems and more AI." At least, I thought the video was pretty good. In the current case, the task is AI agent orchestration, which people thought would require a lot of human oversight. Not so much. Now we have Dot and Muse and Grok Bot, Instinct, and Gemini Spark all doing this work. Next up: swarms of AI agents. 

Here's the problem: "Agents are a very, very strange animal, and I always like to say it's like Drew Barrymore from 50 First Dates. Every time they open their eyes, they forget everything that happened before." Microsoft's response is called "Fabric IQ, a shared intelligence layer that connects enterprise data, business metrics, and operational context across the Microsoft ecosystem." There's a lot more tech stuff in the article, but what isn't clear to me is whether your AI is consuming tokens every time it reorients itself. It's an issue I've run into as my own environment gets more complex, leaving each agent with that much more to learn to catch up.