OLDaily

Big data, learning analytics and language learning / teaching
Philip J. Kerr, Adaptive Learning in ELT, 2019/03/15

This is a very real possibility in learning analytics: "But, for the data and analytics to be of any value in guiding language learning, it must lead to actionable insights. Unfortunately, as Jørno & Gynther (2018: 198) point out, there is very little clarity about what is meant by ‘actionable insights’. There is a danger that data and analytics ‘simply gravitates towards insights that confirm longstanding good practice and insights, such as 'students tend to ignore optional learning activities … [and] focus on activities that are assessed.'" Image: IMS Caliper Analytics.

Web: [Direct Link] [This Post]

Do Neural Networks Show Gestalt Phenomena? An Exploration of the Law of Closure
Been Kim, Emily Reif, Martin Wattenberg, Samy Bengio, arXiv, 2019/03/15

This paper posted to arXiv by some Google researchers shows that neural networks can be studied from the perspective of Gestalt psychology. In particular, the authors show that neural networks demonstrate the law of closure, that is, how "the human visual perception system has a tendency to 'close the gap' in order to perceive whole objects when only fragments are visible." It's based on the old connectionist principle of pattern recognition (see p. 48 of this excellent outline of connectionism). Anyhow, it seems to me that things like this show that as time goes by we are seeing a gradual convergence of psychology, artificial intelligence, and learning theory.

Web: [Direct Link] [This Post]

Facebook and Telegram Are Hoping to Succeed Where Bitcoin Failed
Nathaniel Popper, Mike Isaac, New York Times, 2019/03/15

This meshes with the idea that Facebook wants to follow the path set out by WeChat in China, which is widely used for financial transactions (people just pull out their phone and pay  I saw it a lot when I was in Beijing). Combining payments with a digital currency (Facebook acquired Chainspace in February) would address the pervasive problem (and cost) of currency conversion in the west. But it would have to be a 'proof of authority' currency because alternative models are too slow. Such an approach would also require what Facebook provides - a unique and knowable identity. We'll know we've passed the threshold from technology company to nation state when we are required to provide our Facebook ID for driver's licenses, credit card applications, passports, school records, etc. It's not that farfetched, and probably aligns with Facebook's end-game.

Web: [Direct Link] [This Post]

Arguing With AI
Lindsay McKenzie, Inside Higher Ed, 2019/03/15

The only link in this article is to the Harvard Debate Council Diversity Project, and not even to the IBM Project Debater web page that is the subject of the story, which in the light of recent admissions scandals strikes me as somewhat suspicious. There's a nod to "worries that access to the technology might only be available to students at the wealthiest institutions" in the very last paragraph, but otherwise the article is more concerned about whether the AI "really Understands" and how it's not able (yet?) to capture the nuance of an academic debate. More interesting is Speech by Crowd, an AI platform for crowdsourcing decision support inviting users to "share your arguments on debatable topics, and our AI constructs persuasive narratives both pro and con, giving you a fresh perspective."

Web: [Direct Link] [This Post]