OLWeekly

Gender, Race, and Intersectional Bias in Resume Screening via Language Model Retrieval
Kyra Wilson, Aylin Caliskan, Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, 2024/11/11

The topic of AI-based recruitment and hiring has been discussed here before and research continues apace. This item (13 page PDF), despite the characterization in GeekWire, is a fairly narrow study. It looks at three text-embedding models based on Mistral-7B-v0.1, and tests for gender and racial bias on applications containing name and position only, and name and position and some content (the paper discusses removing the name but does do it). The interesting bit is that intersectional bias (ie., combining gender and race) is not merely a combination of the separate biases; while separate biases exaggerated the discrimination, "intersectional results, on the other hand, do correspond more strongly to real-world discrimination in resume screening." Via Lisa Marie Blaschke, who in turn credits Audrey Watters.

Web: [Direct Link] [This Post][Share]

Feels Like Empathy: How “Emotional” AI Challenges Human Essence
Angelina Chen, Sarah Koegel, Oliver Hannon, Raffaele Ciriello, Australasian Conference on Information Systems, 2024/11/14

One of the major differences between AI and humans, we are told, is the capacity to empathize. But is empathy intrinsic to being human? We could certainly imagine an 'Artificial Emotional Intelligence (AEI)' that mirrors human empathy and is used as a much-needed salve for social ills such as loneliness. But offer a convincing enough emulation of empathy and we humans are inclined to conclude that it is empathy. And the authors of this paper (14 page PDF) consider the possibility of an unintended outcome of AEI: that we humans would change what we mean by empathy. "The assimilation of AEI into society gives rise to emergent empathy. This concept reframes empathy, not as an intrinsic feature but as an outcome of the dynamic interaction between humans, technology, tasks, and structures."

Web: [Direct Link] [This Post][Share]

Trashing universities is now a populist position
David Kernohan, WonkHe, 2024/11/11

I suppose I would agree with a lot of what David Kernohan offers in this long and winding discussion of the populist critique of the university, but I'm also finding myself disagreeing with a lot of it. With Kernohan, I reject the core of the critique, which is to assert that today's universities focus on "indoctrination" or "the priming of the younger generation around identity politics" and the consequent "rise of the bureaucratic class." It's a red herring. What the populist critique does tap into (and we see it even in Kernohan's retelling of the Kemi Badenoch origin story) is the idea that university is funded by an elite, for an elite, and otherwise blocks opportunities at the door for the rest of us.

When universities decided that raising tuition fees was a viable strategy, they laid the foundation for populist attacks against them. It's not so much the idea that "I could have gone to Oxford" as it is the idea of "a growing consensus that we can't just keep expanding university provision." If the doors to opportunity are being locked, and it's a different type of people being granted admission, why not tear down the institution? After all (goes the argument) it's not like they "disseminate truth and knowledge" or "train the next generation of young minds to actually anticipate the world as it is (and) to apply first principles to critical and difficult problems." Or, bluntly, if universities aren't empowering us (the people), what's the point of them?

Web: [Direct Link] [This Post][Share]

From Artificial Intelligence to Artificial Consciousness
Isaac Mao, Isaac+Mao, 2024/11/14

What's the difference between artificial intelligence and artificial consciousness? The difference is small enough for Daniel Dennett to warn against the deployment of "counterfeit people". Where the difference lies, at least as I read this article from Isaac Mao, is somewhere on the inside, echoing Thomas Nagel's idea that consciousness means that it is something that it 'feels like' to be something. A bat, say. This, presumably, would result in capacities an artificial intelligence doesn't have; Mao suggests that an AI can't be creative the way a human can. But my views align with Hume's, that consciousness is nothing more than experience, and that there's nothing in principle that would prevent an AI from having experiences.

Web: [Direct Link] [This Post][Share]

Magentic-One: A Generalist Multi-Agent System for Solving Complex Tasks
Adam Fourney, et al., Microsoft Research, 2024/11/13

Microsoft proposes a "generalist agentic system designed to solve such tasks. Magentic-One employs a multi-agent architecture where a lead agent, the Orchestrator, directs four other agents to solve.... complete complex, multi-step tasks across a wide range of scenarios people encounter in their daily lives." Examples of such tasks: find and edit missing citations in a paper; oprder a shawarma sandwich; describe trends in the S&P 500; and count the number of members of MSR-HAX. What's interesting is the use of WebArena to test the tool; this is a service that emulates interactive websites like GitHub, Reddit, and others. And here you see the current real implementations of such systems: automating reviews and comparing prices on vendor websites and autoposting to Reddit forums. Or, you know, maximizing prices.

Web: [Direct Link] [This Post][Share]

Human-AI collaboration: Designing artificial agents to facilitate socially shared regulation among learners
Justin Edwards, British Journal of Educational Technology, 2024/11/12

This article is dense but yields to a reasonable careful reading. In a nutshell: the authors consider what the use of AI could tell us about how to support "Socially shared regulation of learning (SSRL)", that is, "a complex process where learners collectively manage their learning activities, influencing each other's cognitive, motivational, and emotional states." They conceptialize an AI called a Metacognitive Artificial Intelligence (MAI) employing a design process called Echeloned Design Science Research (EDSR) (which readers will find is very similar to the Waterfall design methodology) and, lacking an actual functioning prototype, emulate its functionality using humans behind the scenes (aka the Wizard of Oz (WOz) approach), and apply it using a cohort of 52 pre-service teachers. It didn't work. "Although the theory-grounded design principles proposed in the domain appear somewhat effective, our results indicate that these are insufficient in real-world contexts, and insights from human-computer interaction are necessary for creating truly effective designs."

Web: [Direct Link] [This Post][Share]

The submissive university
Ingrid Robeyns, Crooked Timber, 2024/11/11

I have many thoughts about this article on education funding cuts in the Netherlands, but I want to focus on two items in this post. The first is the author's assertion that universities "are not supposed to serve any master other than truth, in all its dimensions." The second is that "politicians with authoritarian aspirations try to undermine public institutions because they are crucial for truth-telling." These is an element of truth to each of these assertions, but in a more important sense, they are not true: universities serve societies , and authoritarians would rather universities served the state. Now there is a lot of nuance here I'm eliding, but my takeaway is this: universities were strongest when their mission was based on providing knowledge and learning for all, and the further they move away from this ideal, the more reasonable the authoritarian case seems to be against them.

Web: [Direct Link] [This Post][Share]

The Different (and Modern) Ways To Toggle Content
Daniel Schwarz, CSS-Tricks, 2024/11/08

I spend a lot of time thinking about how to display and hide things on a web page. This article looks at some of the modern ways to do that, with an emphasis disclosures using "the Dialog API, the Popover API, and more." This is a bit of an insight into the whole world of html and javacript  APIs which I have been learning about recently.

Web: [Direct Link] [This Post][Share]

Stature and status: Height, ability, and labor market outcomes - PMC
Anne Case, Christina Paxson, PubMed Central, 2024/11/08

According to this article, "height is positively associated with cognitive ability, which is rewarded in the labor market." I would speculate that any such correlation would have to do with nutrition and health prebirth and in childhood. You want better learning outcomes? Ensure children are well-nourished. But there's a darker side to this argument, as in this report we read about height being used to screen applicants to higher education programs in Vietnam, where the restriction has generated outrage. "The school announced that female students must be at least 1.58 meters tall and male students at least 1.65 meters to be considered for admission this year." Both items via HESA.

Web: [Direct Link] [This Post][Share]

ChatGPT is transforming peer review — how can we use it responsibly?
James Zou, Nature, 2024/11/08

According to this article, "at major computer-science publication venues, up to 17% of the peer reviews are now written by artificial intelligence." Of course, the detection of AI writing isn't totally reliable, but as James Zou writes, "reviews penned by AI tools stand out because of their formal tone and verbosity — traits commonly associated with the writing style of large language models (LLMs)." The article has the usual caveats that AI should not replace human reviewers and that "guardrails" should be in place. But "the tidal wave of LLM use in academic writing and peer review cannot be stopped." Via Daily Nous.

Web: [Direct Link] [This Post][Share]

OpenAI defeats news outlets' copyright lawsuit over AI training, for now
Blake Brittain, Reuters, 2024/11/08

This will be all over the tech news today, but it's important to note that this is just the first step in what will certainly be a long legal process as the case works through appeals. But the finding is still significant. As Donald Clark summarizes, "Generative AI 'synthesises', it does not copy. This is central. It's a bit like our brains, we see, hear and read stuff but memory isn't copying, it's a process of synthesis and recall is reconstructive." More: VentureBeat, Engadget. Here is the full opinion.

Web: [Direct Link] [This Post][Share]

Open Web UI Installer for Reclaim Cloud
Meredith Hoffman, Taylor Jadin, Reclaim the Blog, 2024/11/07

I had a ton of fun this afternoon following along with the top video and setting up a Web UI website in Reclaim Cloud. The point of this video was to highlight the new installer, which makes it pretty easy to set up the interface. What the interface does is to give you a window where you can access an AI interface (such as ChatGPT) or run an AI model locally as well as additional tools, functions, prompts and models. You can even build your own knowledge base. This isn't strictly a how-to video; it's more the two participants exploring the installer and the functions, and there are tangents and side-quests. At the same time, there are some really useful bits, like knowing you have to set the image dimensions properly (eg. 1024x1024) for the image generation to work.

Web: [Direct Link] [This Post][Share]

The duty to explain: a positive defence of academic freedom in the social sciences
Bruce Macfarlane, Xiao Han, Teaching in Higher Education, 2024/11/07

This paper (13 page PDF) argues that it is "unhelpful" to think of "academic freedom as a stark division between those who believe in 'balance' or 'neutrality', and others who insist on the right of the university teacher to express their personal beliefs without regard to their impact on the independent intellectual development of others." Why should professors get special rights to express beliefs that others do not have? Instead, it argues that the basis for academic freedom lies in the same principle as 'informed consent' found in other professions; providing the client with all the information they need to make an informed decision constitutes a part of the 'duty of care' that forms an ethical basis for these professions

Web: [Direct Link] [This Post][Share]

Universities need to leave X
Martin Weller, The Ed Techie, 2024/11/07

I have long ago left Twitter for reasons articulated elsewhere, so I understand the basis for Martin Weller's argument here. "It is explicitly an ideological platform now," he writes. "And unless your ideology aligns with that, then maintaining an account there is actively supporting it." Ah - but here's the rub: "I would hope that most university's ideology does not align with that of Musk." Why would we believe that? I won't offer the analysis here, but it's not hard for me to describe the educational system as one designed to advance the interests of the rich and powerful while maintaining class-based society and resisting the trappings of democratic control. People who are - or who want to be - the beneficiaries of such a system have no real issue with content and media that supports it.

Web: [Direct Link] [This Post][Share]

Human Made's S3 Uploads Plugin
Jim Groom, bavatuesday, 2024/11/07

One of the ongoing maintenance tasks I have been working on in the background is the gradual transfer of my media content to cloud hosting instead of my (more expensive) web storage. This is why you're not seeing some of my presentation content when you navigate my presentation pages. Working with cloud storage is not straightforward, and least, not if you're running a website that needs to access it. In this post Jim Groom discusses the use of a WordPress plugin to manage uploading and retrieval. I've been figuring out how code this this by hand for my own non-WordPress software. It's fun work, but definitely not for everybody.

Web: [Direct Link] [This Post][Share]

The uniqueness of learning: Rethinking the meaning of student-centred education
Adan Chew, Magdalena Cerbin-Koczorowska, Jess Gurney, Teaching Matters, University of Edinburgh, 2024/11/07

It's not clear to me that educators and educational institutions have embraced the findings discussed in this article. The authors assert, "an exploration of the predominant learning theories demonstrates that not only should each learner be treated as an individual with their unique learning readiness and abilities, but the dynamic and multifactorial nature of the educational environment cannot be ignored." They cite with approval Bates et al. ​(2019)​, "the inevitable diversity of contexts for learning and practice renders any absolute standardisation of programs, experiences, or outcomes an impossibility." But if it's true that "there is simply no 'one size fits all' approach to student-centred education" then how is any approach based on instructors, classes, and programmes to be successful?

Web: [Direct Link] [This Post][Share]

Canadian Government to Ban TikTok (the Company not the App)
Michael Geist, 2024/11/08

I'm on the verge of closing my account and removing TikTok from my phone. This partially because of the new Canadian action, but mostly because of its content moderation stance that silence counts as copyright content. It's two sides of the same coin, I think: on the one hand, exercising arbitrary control over the content, and on the other hand "unacceptably invasive-by-design, treat(ing) users as raw material for personal data surveillance, and fall(ing) short on transparency about their data sharing practices." And thus TikTok falls into the junk pile where Facebook and Twitter already reside.

Web: [Direct Link] [This Post][Share]

Falsehoods Programmers believe about DOIs
Joe Wass, Pardalotus, 2024/11/07

This is an informative post about digital object identifiers (DOI) and, indirectly, the Handle system. It's set up as a series of myths about DOI, which it then addresses. DOI was originally intended as a sort of permanent URL system (permaurl) for academic publications, so that even if the source URL changed, the URLs defined for the paper as a DOI would still allow users to read the paper. This guide is intended to be helpful (and it is!) but I also read it as a documentation of the degradation of DOI over time, as it depends on the goodwill and compliance of publishers, which has always been in short supply. I believe that at some point in the future DOI will be replaced with content addressing (CA), which will allow the retrieval of a document from any place it is stored, and not just the publisher's vault. Via Paul Walk.

Web: [Direct Link] [This Post][Share]

Oldest depictions of fishing discovered in Ice Age art: Camp site reveals 15,800-year-old engravings of fish trapping
IDW, 2024/11/07

I'm going to designate this as one of the earliest examples of an open educational resource (OER). It's an engraving of a fish and fish netting found in an ice-age camp site located at Gönnersdorf on the banks of the Rhine. The importance of fishing in an ice-age culture underlines the need to pass on the knowledge from one generation to the next. Here's the original publication.

Web: [Direct Link] [This Post][Share]

Researchers rejoice: the Canadian Common CV's days are numbered
Brian Owens, University Affairs, 2024/11/06

This is interesting. "Researchers applying for grants from Canada's three main federal funding agencies will soon have cause for celebration: the unpopular Canadian Common CV is being phased out in favour of a new narrative CV.... called the Tri-Agency CV, is a maximum of five pages (six in French) divided into six sections." The new format is focused much more on narrative than data, with a focus on things like your 'personal statement' and your 'most significant contributions'. I'm still not a fan of requiring people to fill out a form on a centralized website, but at least it's a bit of movement in the right direction.

Web: [Direct Link] [This Post][Share]

The Blight on the Ivy
Alex Usher, HESA, 2024/11/06

In case you thought nothing has changed over the last 60 years Alex Usher provides us with a treat of a review of A Blight on the Ivy by Robert and Katherine Gordon, published in 1962. "Before I go any further, let me stress that this is not a good book," writes Alex Usher. Beyond it's flaws, though, it's worth noting that "it's a book about the dissolution of a certain college ideal—one that involves residential living and enormous dollops of pastoral care from academics—under conditions of massification." It's hard to imagine ever returning to pre-1960s higher education anywhere in the world. And this is a good thing.

Web: [Direct Link] [This Post][Share]

What audiences really want: For journalists to connect with them as people
Mark Coddington, Seth Lewis, Nieman Lab, 2024/11/06

As always, my take here is that what we say about journalists can also be said about educators. And this article is a great case in point. The point of departure: "the core professional values that define good journalism... factuality, impartiality, public service, autonomy, and ethics." In this study, though, researchers find that the public wants "approachability, empathy, and skills in communicating clearly and in ways that emotionally resonate." I could recast this by saying people want to see themselves reflected in journalism  - and therefore, education - at least to the extent that what is important to them is reflected in what is considered important by the institution.

Web: [Direct Link] [This Post][Share]

The Ethics of Harmonizing with the Dao
Eric Schwitzgebel, The Splintered Mind, 2024/11/05

I'm not exactly going to endorse the view of ethics ascribed to Daoist thought outlined here, but it's worth noting that my own thinking has been influenced by such considerations. I was asked once what my core philosophy looks like and I described a scene where I was walking down a street in Edmonton and felt, for a moment, in perfect balance - harmony - with everything around me. I have described my philosophy of state similarly: in diversity, harmony. And in discussions online I've noodled around the concept of a 'pedagogy of harmony'. It has a basis in caring and community and culture, but it's centered also in a sense of self and identity and connecting. All as a form of harmony, not conflict, not competition (except in fun).

Web: [Direct Link] [This Post][Share]

On a Path to Open
Rajiv Jhangiani, Oya Pakkal, Catherine Lachaine, Robert Luke, eCampusOntario, 2024/11/05

This report (48 page PDF) briefly outlines open educational resources (OER) and open educational practices (OEP), outlining their benefits, and then details the responses to a survey of Ontario institutions' awareness, support and infrastructure supporting OER and OEP.

Web: [Direct Link] [This Post][Share]

AI school opens - learners are not good or bad but fast and slow
Donald Clark, Donald Clark Plan B, 2024/11/05

"For the first time in the history of our species we have technology that performs some of the tasks of teaching," writes Donald Clark. "We have reached a pivot point where this can be tried and tested." Yet "you'd think it was Armageddon." I too am puzzled about the rage against AI. It seems to me to be out of proportion to its impact - critics exaggerate environmental impact, exaggerate social harm, exaggerate corporate ogopoly. But why?

Web: [Direct Link] [This Post][Share]

Dig a little bit
Andrew Jacobs, Lost and Desperate, 2024/11/04

Andrew Jacobs looks into the research behind an article claiming "most people want Instructor Led Training (ILT) at work" and finds shoddy and ill-considered research. In one case, "the source was easy enough to find but it was a single source which looked at the preferences of 336 Greek university students during the COVID-19 pandemic." In another case, a stat quoted in a Forbes article came from "a professional body made up of people with instructor led training (ILT) as part of their core offer." As he says, "we must be better in L&D." It's super-easy to find small scale studies or thinly veiled opinion pieces that support a certain point of view. But it's not exactly ethical.

Web: [Direct Link] [This Post][Share]

Five Guiding Principles of Successful Immersive Learning
Hannah McNaughton-Hussain, Discovery Education Blog, 2024/11/04

A lot of this may be old news for long-time readers, but it's useful to revisit occasionally the basic elements of immersive learning. This article identifies five dimensions: sensory range, compelling gameplay, narrative, producton quality, and agency. Why do these factors matter? Well, it varies, depending on what you're doing. But in general, "when students feel immersed in an experience, they are more receptive to the content, which enhances retention, fosters deeper understanding, and creates a lasting connection with the subject."

Web: [Direct Link] [This Post][Share]

When AI Meets Education: The Power of Diffusion Over Replication
Culture of Yes, 2024/11/04

"A decade ago," writes Chris Kennedy, "I wrote about how meaningful change in education spreads through diffusion rather than replication." He credits David Albury for the idea. This post adapts the idea to the spread of AI. "Just as we learned with previous innovations, the most impactful changes come when we allow ideas to diffuse naturally, adapting to each teacher's unique talents and each school's and district's unique context." This makes sense, since the odds of finding the killer AI application everyone in education should use are nearly zero. Not that there won't be a thousand PowerPoints suggesting otherwise.

Web: [Direct Link] [This Post][Share]

HTML link, or button, that is the question
Marijke Luttekes, 2024/11/04

I'm not sure there are many people creating learning technology applications by hand any more, but for those who are, this discussion of links versus buttons is useful (as an aside, I'm not sure any learning technology is being developed these days, beyond wrappers for AI interfaces). Anyhow, to make a long story short: use a link to open a web page or resource, and use a button to perform an action. Doing it the other way around will create usability problems. Image: Eric Eggert.

Web: [Direct Link] [This Post][Share]

mCaptcha: Replacing Captchas with Rate Limiters to Improve Security and Accessibility
Aravinth Manivannan, Sibi Chakkaravarthy Sethuraman, Devi Priya Vimala Sudhakaran, 2024/11/04

If you're like me, computers have long since become better at solving captchas than you are. I mean, how much of a motorcycle counts as being 'a motorcycle in the square'? I sure don't know. No, the battle to test whether a web user is a human or a computer is over, and computers have won. So now what? Enter mCaptcha. "Our broader goal," write the authors, "is to stop attempting to distinguish between humans and robots and return to captcha's original intent: providing denial of service (DoS) protection." They do this by requiring a 'proof of work' from the web browser. It's a single-use computation that can be performed pretty quickly by your web browser, but would bog down an AI or web scraper trying to make as many requests as possible. If it were me, I would want to see this 'proof of work' be some actually useful computation (which could then be thought of as the 'price' to access a website). But this is a good place to start.

Web: [Direct Link] [This Post][Share]

TikTok Muting
Stephen Downes, Mastodon, 2024/11/04

I'm not usually the source for a weird copyright story, but this time I am: The sound on my TikTok videos is being muted because it contains too long a selection from a copyright work - John Cage's 4'33". Here's the link to my TikTok page where you can watch (but not hear) my travel videos from Croatia and Slovenia. And here's a link where you can listen to (but not hear) John Cage's 4'33".

Web: [Direct Link] [This Post][Share]