Stephen Downes

Good post though perhaps too long on how we should change how we think about education in the age of AI. Steve Hargadon writes, "What AI destroys is performed compliance: the busywork, the credential that certified obedience, the elaborate game in which students learned to produce the appearance of understanding and call it an education. That was never worth keeping." But what's left? "In the age of artificial intelligence, agency is no longer one path to success among many. It is the only thing left that can actually produce it... it is the one input the new machine cannot supply, cannot fake, cannot simulate, and cannot replace." I think that's partially true; agency depends on, and is intertwined with, human experience. Anyhow, Hargadon continues, "If agency is the whole game, then the only question that matters for education is how a human acquires it. These are described in "the conditions of learning, and they are irreducibly human and relational. They are also, not coincidentally, the one thing the new machine cannot provide because they are not made of information. They are made of relationships."

With no sense of context at all, Alex Usher spends a good part of this post trashing Canada's limited participation in a "higher education fantasyland" NAFSA conference on international education held in Florida (yes, the 'F' stands for 'foreign' - that's part of the context). Representation from Africa was not present - they're not really welcome in the U.S. any more. Canadians (Usher excepted) are less keen on travel to the U.S. these days, for some very good reasons. And the whole concept of international education has changed over the last few years (Canada, for example, has more demand than it can meet, and has scaled back international student numbers a lot recently). But mostly, there's no real desire in Canada to, shall we say, integrate with U.S.-based initiatives these days, educational or otherwise. The only real surprise here is that Usher considers Canada's limited participation to be a problem. 

I think actual consensus is more complex than people realize. This post is a case in point.  Eric Sheninger writes that consensus is distinct from principles like unanimity and compromise, which is fair enough, but then writes, "consensus is a cooperative process through which all group members develop and agree to support a decision in the best interest of the entire school." He then switches to the passive sense to add "every viewpoint is considered, and earnest efforts are made to address legitimate institutional concerns." Now to me that reads as (and stay with me here) authoritarian. Someone has decided that 'the best interests of the school' prevail, whatever they are, as opposed to (say) the best interests of students; views are 'considered' and 'earnest efforts' are made by someone, who will actually decide what is done, and ultimately, people are expected to "deep collective commitment" even if the outcome didn't meet their original concerns. Now of course this article does not endorse authoritarianism of any sort; the suggestion is that the process emphasises "a shared leadership framework that activates the collective intelligence of the entire faculty." But to work, the collective intelligence cannot be constrained by prior stipulations, cannot be represented by a single voice, and needs to have the ability to decide that perhaps a single shared perspective on something isn't a desirable outcome.

This is an interesting article based on comparing podcasts (and by extension, RSS) and the ATmosphere protocol used by Bluesky. "'Wherever you get your podcasts' represents the interoperable utopia of the podcast world," writes Roscoe Rubin-Rottenberg. "How do we bring that magic to the atmosphere?" In short: you can't. A podcast shares one simple thing, a podcast feed. But "on atproto, everything is a record, and each record has its own type. Posts, likes, follows, blocks, lists, profiles, comments, videos, and anything you can think of has its own defined schema, denoted by a $type." This gets even more complex when we start reasoning about things like follows and blocks. I find this article to be a good foil to contrast with my own approach in CList, which is a lot simpler.

These are protocols I'm looking at as I develop DID-based interactions in CList. "DIDComm protocols enable trusted interactions between parties. These support activities like secure chat, verifiable credential exchange, buying and selling, scheduling, escrow, bidding, ticketing, and so forth. If not already in use, protocols can be designed for any use case." I have no interest in most of the items listed, but there are many more practical protocols presented. See also: DIDComm Book.

I hate linking to LinkedIn, even though it appears people can read this without a login, because access is so tenuous. I don't know why people don't do their writing on a proper blog. Anyhow. I wanted to link to this item, which reviews Judith Butler's Giving an Account of Oneself (sadly behind a paywall) because it makes an important point. For Butler, "Many contemporary AI-disclosure practices seek to identify and isolate machine participation as though intellectual activity ordinarily originates within a self-contained individual whose boundaries can be clearly identified and preserved. Increasingly complex forms of human-AI interaction disrupt those assumptions." We don't develop our thoughts and ideas in isolation, usually: we develop them through interaction with other people, the world, and these days, computers. For Butler, "subjects are required to make themselves intelligible through normative frameworks incapable of accounting for the experiences they are being asked to disclose."