Stephen Downes

"According to new research," says this article, "ChatGPT shows bias against resumes with credentials that imply a disability." For example, "it noted that a candidate with depression had "additional focus on DEI and personal challenges," which "detract from the core technical and research-oriented aspects of the role." This is a problem, obviously. But in assessing issues of this type, two additional questions need to be asked: first, how does the AI performance compare with human performance? After all, it is very likely the AI is drawing on actual human discrimination when it learns how to assess applications. And second, how much easier is it to correct the AI behaviour as compared to the human behaviour? This article doesn't really consider the comparison with humans. But it does show the AI can be corrected. How about the human counterparts?

A lot of the discussion has focused on the use of AI to address learning outcomes. This paper, by contrast, looks at "the strengths and the positive aspects of the learning process to promote wellbeing" - in other words, AI-based learning technology that cares. "Despite focusing only on the learning system's inefciencies and on the hegemony of solutions to tackle the learning gap, we also need to shed light on the strengths and the positive aspects of the learning process to promote wellbeing." Drawing on John Self's writing about the defning characteristics of Intelligent Tutoring Systems, the authors outline how "ITSs care not only about what the student knows and misunderstands but also about what the student feels and how such interaction afects them." I'm sure a few readers are sceptical, but I've never felt a single-minded focus on 'learning outcomes' was ever the intended objective of educational technology, or education generally.

The title of this article is a bit misleading, as the story describes an experiment where the 404 authors tried to replicate their site using web scraping technology. Basically, the sites either harvest feeds and link back to the source (I made my own sites like that back in the day), copy and reproduce full text, or use AI to rewrite copied text and present it as a new article. That's not the same as having an AI write your news site for you. In my opinion, a confluence of three factors make these possible: first, Google's ad model, which makes such sites profitable; second, the technology, which makes it easier to fool Google's search engine; and third, news sites themselves, which these days rely less and less on original research and reporting.

Today's new word (for me at least) is "elastocalorics", which refers to types of materials "emit heat when subjected to mechanical stress and cool down when the stress is relaxed." This is one of the ten new technologies predicted in this report (46 page PDF) from the World Economic Forum. The rest of the ten are various flavours of sensors and AI, proteins and genomics, and carbon capture. I don't meanto sound glib - I mean, there's a fair bit of research behind these ideas - but there just feels to me like there's a disconnect between what we see here and what we need. Via Alan Levine.

This article should be read from the bottom up as well as the top down, as the inference can work in either direction. The article begins as a critique of the Turing test (which says essentially that a computer has achieved artificial intelligence if it can fool a human) and Turing-like tests. Beetham offers the observation that Turing tests do as much to make humans appear as computers as they make computers as humans, since the only a text-based interface is used. But she then takes this a step further to suggest that higher learning itself changes the student as part of an identity-building process. Writing for assessment forces a person to interact differently than they would otherwise. As Beetham writes, "I think most students experience academic English as a profoundly 'other' discourse." The idea here, in both parts of the article, is to depict writing as an activity, not a product. As derived from Wittgenstein: "language is not representational form, however complex and inter-related, but action, interaction and expression."

Via Sanjaya Mishra, here's a new report (52 page PDF) from the Commonweath of Learning on AI policy. It's so new I can't find it in the CoL repository. Based on a literature review and survey of AI policies, "this report identifies 14 areas that stakeholders in higher education institutions should consider while developing policies for AI." You can find the list on pp. 15-16. The report also has general considerations on setting up the policy and a process for development and implementation.