news (external)

The gist of this article is that the study of artificial neural networks - such as those used in large language models - can help in our understanding of the brain in a way that neuroscience itself has been unable to thus far. What makes it work is that, provided the networks have a sufficient base configuration, they can emulate many areas of human (and animal) cognition, including, for example, vision and image recognition. Good stuff here, and this article supports in a way the assumptions underlying the theory of connectivism George Siemens and I were on about some 15 years ago.

This document was discussed today here in Ireland and seems well worth passing along: "Volume one of the Reference Framework contains the model of competences for democratic culture... Volume two lists the descriptors of the competences for democratic culture... Volume three offers guidance on how the model of competences and the corresponding descriptors may be used in six education contexts."

These are lessons I thought I had learned alreayd, but you don't really internalize them until you live them, I think. Anyhow, based on my own experience, the observations offered in tbis article are on point: "A successful platform makes users highly autonomous; being intentional about what it offers, in order to bring velocity and efficiency to your users, making it easy to do what is right and hard what is wrong." Easy to say, hard to do.

Though this article is intended for online data journalists, it is a method and a compilation of resources that is quite useful to online instructors as well. The idea is that the various stages of data-based news article or investigation are presented with some background and resources, leading the reader from the development of an idea through compilation and clearning of the data to setting the context and telling the story, with numerous resources supporting each step.

"What is standing between people and their ability to do what they want online?" asks Jessamyn West. "The skills people were given—perhaps how to Google an error message or find an FAQ for an app—are increasingly getting manipulated not just by other people but by large language model-created spammy webpages that sound authoritative but are often nonsense. So, addressing this new digital divide is not just a digital literacy issue—although it is that too—but it also includes systems and network literacy."

I think the way to think about the AI here is as a conversational alternative to search fields or menu items. The AI doesn't actually contain the answers to questions, it refers you to the answers. "A startup called Decagon is helping Substack automate routine support functions... Substack provided Decagon with a library of relevant documents—for example, copies of all the blog posts Substack has written over the years about its technology."

Metaphor, similarity and conceptual blending are central concepts in human cognition, and not the niche topics they are sometimes represented to be. When we talk about processes of learning, inference and discovery, it is to these - and not abstrations and generalizations - to which we should turn. This post is also a good platform to plug Doug Belshaw's other blog on the topic.

Most of the research proposed here sounds pretty reasonable, though it might be worth surveying what other research is being done by other groups, since this article seems to suggest that little to nothing is happening. I would especially welcome research on access by income levels and by program. I'm less enthused about research into quality, not because I'm opposed to quality, but because the definition of quality is so vague that such research often verges into advocacy rather than measurement. My only other major thought on this piece is that it is very much devoted to traditional educational practices - typified by the comment that "the biggest area of student experience, though, is what happens in class." I don't think that was true even in my time and in this era of online learning I think it's even less true.

Alan Levine points to this document in one of his always useful OE Global Connect posts. "The publication addresses the issue of quality as applied to the performance of open and distance learning universities and intends to raise awareness about benchmarking methodology as a tool for quality assessment and improvement." 133 page PDF.

Miguel Guhlin shares this announcement on Perplexity Pages, touted as "a simple way to turn your research into visually appealing articles. With formatted images and sections, Pages lets you share in-depth knowledge on any topic...log into your Perplexity account, head to your Library to create a Page. Choose your topic, and let Pages guide you through the content creation process on desktop." Here's his can sample page.

Paul Wiltshire, we are told, "argues that we need a proper debate about student numbers." By this he means 'fewer'. His main argument is this: "it is obvious that the majority of the jobs that society needs performing (a high proportion of which are manual or administration based, low to medium skilled, or managerial – but not actually academic) don't actually need the incumbent to have spent their first three or four years of their adult life studying an academic course." He also argues against the idea that it is the academic education that explains graduates' higher pay: "the university cohort will be those who are not only on average more academic, but also more hard working and more conformist – so it is little wonder there is a career pay premium for this cohort." These are both stunningly bad arguments but not at all surprising in a post that argues against access to education.

I'll probably move on from this type of post in the future, but for now it remains relevant, especially for educators. Shannon Vallor writes, "the most ordinary human does vastly more than the most powerful AI system, which can only calculate optimally efficient paths through high-dimensional vector space and return the corresponding symbols, word tokens or pixels." Human intelligence, argues Vallor, is fundamentally different from AI. I think it's a bit of a straw man to point to the obviously limited capacities of today's AI. The argument is, "Why would a machine that works on silicon not be able to perform any of the computations that our brain does?" And what would the brain be doing other than computations?

I think there's a danger in speculating about biomechanical motives for actions. Such is the case here. Here, Mark William Johnson is responding to Michael Foot, who responds, "we are here to provide for all those who are weaker and hungrier, more battered and crippled than ourselves." Johnson, referencing epigenetics, suggests, "The uncomfortable fact may be that without the cruelty and selfishness, there can be no progress." That would be sad if true. But I am convinced that reality is rather more complex and subtle, at least to the extent that we can at least imagine doing away with cruelty, if not also selfishness.

This is from a few days ago but I want to make sure I don't forget it. Alan Levine, channeling some Mastodon discussion, points to a nice trick to get right of all the garbage in Google search engine results (by 'garbage' what I mean is all the stuff Google puts in the results page that isn't part of the search results). In particular, if you add '&udm=14' to the end of the URL on your results page, you can force it to display search results only. I do a similar thing with my personal Start Page, where I've created a small form to handle that for me. I also use '&tbs=qdr:y' to force it to return results only from the last year, so I'm not getting out-of-date results.

I've had discussions with people from time to time about the difference between networks and systems, and to me one of the main differentiators is this: a system has a purpose, and a network doesn't. Sometimes this difference can be difficult to parse, and this phrase, which originates with Stafford Beer, sits right along that dividing line. The idea that "The purpose of a system is what it does" (POSIWID) allows us to retroactively assign an intent to a system, even if it was not designed that way. A network, by contract, does not imbue cognitive properties; the structures and processes are purely physical, with no cognitive guidance whatsoever.

If you have beliefs about AI, one way or another, that you hold with religious fervour, then perhaps your faith is misplaced. AI is neither the stuff or prayers or nightmares; it occupies that mundane space in between. Moreover, as the Walrus puts it trenchantly, "The problems facing Canada or the world—not just climate change but the housing crisis, the toxic drug crisis, or growing anti-immigrant sentiment—aren't problems caused by a lack of intelligence or computing power... the fixes are difficult to implement because of social and political forces, not a lack of insight, thinking, or novelty."

There are two things we can draw from this short article. First, generative AI isn't to the point yet where it can reliably create things like employee handbooks (and therefore, other training materials). But second, the fact that companies are already using it for this makes it clear that when GAI is reliable, it will be widely used for this purpose. Via George Station.

The summary in Daily Nous says it all, I think: "You might not have heard the philosophy lectures of Michael Sugrue, who died recently, but hundreds of thousands of others have — 'The type of professor you'd ditch class to go and listen to,' says one YouTube commenter." Open Culture leads with, "If we ask which philosophy professor has made the greatest impact in this decade, there's a solid case to be made for the late Michael Sugrue." If impact is defined as reach, then maybe. Though it might be hard to surpass Peter Adamson's monumental History of Philosophy Without Any Gaps. But the main point here - which surely ought to dominate any discussion of online learning - is that the apparatus of colleges, courses, degrees and credentials is only a very small part of the picture, and that real contributions are being made outside the classroom walls, out in society, where such learning surely belongs.

This post comes from the larger publication, AI and Digital Inequalities (72 page PDF). Emma Ruttkamp-Bloem argues that such a policy should include "social values such as affirmation of the interconnectedness of all humans with each other, equity and human agency; human rights values such as privacy, transparency and accountability; and research values such as honesty and integrity." She also argues that "three of the biggest obstacles to attaining these goals include digital poverty concerns, the creation of monolithic societies and misinformation." This article is reflective of the publication as a whole (which is definitely worth a read): it is generally policy-based and founded in social justice themes. But the publication as a whole feels a bit lazy, in the sense that it essentially lists important global social justice issues and applies them to AI, with stipulations that AI should mitigate or in some way address these concerns.

This short article summarizes a longer paper in OTESSA on developing the Technology-Integrated Assessment Framework (19 page PDF) that "serves as a starting point to understand how to improve technology-integrated assessment practices in higher education in British Columbia and beyond." It consists "four components for instructors to consider when planning assessment." Each of these has three or four subcomponments and is based in previous literature on the subject - the purpose of assessment, based on the Bearman et al. model; the duty of care (in law, and also in the sense of communality); technology assessment (the UTAUT model); and assessment design (the five Rs framework). I would have looked for a more critical assessment of each of these (eg. why UTAUT and not UTAUT2?) but that may have taken more space than the journal allowed.