OLDaily

How Do Neural Networks Grow Smarter?
Robin Hiesinger, The Royal Institution, YouTube, 2021/07/07

Take the hour you'll need to watch or listen to this lecture (perhaps over lunch, as I did) by neurobiologist Robin Hiesinger. The video asks the basic question: how do you grow intelligence? Hiesinger gives us the story of the subject that winds through the history of artificial intelligence, discoveries about how bees, worms and butterflies learn, and how algorithmic information theory describes how the genome grows into a new neural network, and how this process, through evolution, produces smarter and more specialized neural networks. This talk touches on many topics familiar to readers of OLDaily over the last 20 years, but there is a whole second half that could (and should) be discussed, about how actual mechanisms of experience, training and development result in education and learning. Still, this is great stuff, and will certainly make you ponder the deep questions.

Web: [Direct Link] [This Post]

We tested AI interview tools. This is what we found.
Sheridan Wall, Hilke Schellmann, 2021/07/07

This article reports on the results of tests of two products specializing in AI job interviews, MyInterview and Curious Thing. They are (predictably) unreliable. For example, when an intrerviewee responded by reading the Wikipedia entry for psychometrics in German, Curious Thing awarded her a 6 out of 9 for English competency. Similarly, MyInterview "pulled personality traits from her voice," despite concerns that "intonation isn’t a reliable indicator of personality." What we should take away from this article, I think, is, first, the fact that companies are using AI to assess interviews, and second, that we cannot (yet) depend on the algorithms - though, it's when they get good that the real etrhical issues will arise. Note that this article is one of '2 free stories remaining' for me; I recommend using Firefox with UBlock Origin to avoid paying for the article, but if you can't, you can see the first half of the article here or here. Meanwhile, perhaps someone can explain to me why an educational institution with an $18.4 billion endowment needs to put a paywall on scientific articles meant to inform the general public.

Web: [Direct Link] [This Post]

Government needs to stop unethical practices in children’s education data gathering, report states
Dorothy Lepkowska, Educate, 2021/07/07

This is a good summary of a report (57 page PDF) looking at data gathering practices that emerged during the rise of online learning due to the pandemic. The report authors write "The governance of data for children’s learning in the UK state schools reveals significant regulatory and implementation gaps in data processing in education contexts." The report makes ten recommendations and these are summarized in the blog post. These include recommendations that government "develop an independent oversight mechanism" and to "set limitations on the legal terms EdTech companies use to contract with children in schools."

Web: [Direct Link] [This Post]

Play-Based Learning Isn’t Free Play and May Be Connected to Online Learning
Nancy Bailey, Nancy Bailey's Education Website, 2021/07/07

I don't want to get into the politics of this (which are probably deep and murky and unsavory) but Nancy Bailey does highlight a concern. "Articles repeatedly highlight play-based learning," she writes, but "it appears that the term has been hijacked." By 'play' we mean "a child or children being given the freedom to play," while 'play-based learning' has come to mean a case where "learning is adult-directed. The teacher organizes the play for a student, and it will likely be aligned to a skill or standard that will be tested." She also reports "screens may also be used, and this raises questions. Is play-based learning about transforming classrooms to online learning?" People should make sure what is being means by the term in their schools, she writes. "At times play-based learning is also helpful. But adults micromanaging and organizing all play is a concern." Reprinted in NEPC.

Web: [Direct Link] [This Post]

‘Social’ Mitochondria, Whispering Between Cells, Influence Health
Quanta Magazine, 2021/07/07

This article describes research into the ways mitochondria interact with each other and produce forms of self-organized behaviour, just as we've seen in other networks of connected entities. Mitochrondria are "are structures within cells that convert the energy from food into a form that cells can use." According to the article, "The repeatedly documented instances of the diversity, interactivity and apparent cooperativity of the organelles point to the existence of mitochondrial social networks in living things."

Web: [Direct Link] [This Post]

Madrid Micro-Credential Statement
Mark Brown, Deborah Arnold, Timothy Read, Don Olcott,Jr., Rory McGreal, Mpine Makoe, 2021/07/07

This (7 page PDF) is a statement, or perhaps more accurately, a series of questions with some answers, arising out of the EDEN 2021 conference on the topic of microcredentials. The questions include: what problems are microcredentials intended to solve? Who do they serve? How are they developed and implemented? What stands in the way of their general acceptance? Will they live up to the promise they offer? (I would add: should we spell 'microcredentials with or without a hyphen?) The authors write, "it appears that most HEIs are applying micro-credentials to short courses as low hanging fruit and easier entry point." Perhaps. But it's hard to see them progress much beyond accoutrements for out-of-program extension courses, because HEIs will cling to the existing model of certificates and degrees for as long as possible. Image: guroo.

Web: [Direct Link] [This Post]

Topic Transitions in MOOCs: An Analysis Study
Fareedah ALSaad, Thomas Reichel, Yuchen Zeng, Abdussalam Alawini, Educational Data Mining 2021, 2021/07/07

This is a really technical paper (11 page PDF) but one you're not going to want to miss. It introduces "the Topic Transition Map, a general structure that models the educational materials at the topic level. We model a course as a set of topics, and each topic is a set of concepts." The authors consider the question of how instructors order the topics in a course and then attempt to emulate that using four methods (read the first paragraph of the conclusion first to get this gist). This is a hard task because there are two major parts: first, creating a model of the different topics taught in the course, and then second, ordering them in the best sequence. See the diagrams (figures 5-8) to visualize this process. The four methods are evaluated against each other for factors such as clarity of the terms produced and the transitions between topics. (Here's the full list of papers from EDM2021).

Web: [Direct Link] [This Post]