by Stephen Downes
Apr 09, 2015
Machine Learning Algorithm Mines 16 Billion E-Mails
MIT Technology review,
I'll leave aside the question of where they got 16 billion emails and pause for a moment to ponder the implications of this: "Human e-mailing behavior is so predictable that computer scientists have created an algorithm that can calculate when an e-mail thread is about to end." (I really thing 'created' is the wrong word here - I think the appropriate word is 'found'.) If we're that predictable, what does it say about us? It used to be that one of the major objections to causal theories of the world was the apparent phenomenon of free will. But suppose the data tells us it's just an illusion. If - if they can tell us when an email thread is about to end, is there any way to telling us whether anyone learned anything from it? See the full report here.
The 60,000 Times Faster Claim Gets Dialed Back to 1982
Have you heard this? "We can process visuals 60,000 times faster than text?" In my own mind I would question it right away because of its overt employment of a computer metaphor to talk about cognition, which to me is prima facie questionable. But Alan Levine actually looks for the source of the quote. He traces it back to an unreferenced claim made in a Business Week advertising section in 1982, attributing it to Philip Cooper, president of Computer Pictures Corporation. Philip Cooper is still around, but hasn't responded yet to Levine's enquiries. And me, what I wonder is, just how much mythology was created over the years by the business and advertising press? Possibly the bulk of what people today call 'common sense' was at one time or another someone's ad copy.
The Math Ceiling: Where’s your cognitive breaking point?
Math With Bad Drawings,
"A student who can answer questions without understanding them is a student with an expiration date." You'll understand. Good discussion of the question of whether there is some mathematical knowledge that is beyond the innate 'limit' of students, or whether every student is (theoretically) able to master every mathematical concept. The disclaimer in the blog ("I can't draw") also makes me wonder whether there's an art ceiling.
Los Angeles Police Department taught the Canadian way when it comes to using force
What's interesting is not that the Los Angeles police are now using training methods employed by the RCMP, but rather, the manner in which training is now being conducted: "When I went through the academy, everything was compartmentalized... Now, recruits are trained the same way they operate in the field: in teams. Training the whole person rather than training on a specific task, training so that the constitutional policing and use-of-force policy and all the things that are important to the public are contained in every exercise and everyone is graded accordingly." Cadets are run through a simulator which runs through a variety of simulators. The instructors watch the cadets' performance. ""I'm not going to tell you what to do... (if you) understand policy, they will be the one to make that decision when and if to use deadly force."
Digitally Connected: Global Perspectives on Youth and Digital Media
Sandra Cortesi, Urs Gasser,
Social Science Research Network,
This open (I think; it uses an SSRN redirection service) online book contains the proceedings of a conference funded by an array of charitable institutions and United Nations agencies that only an institution like Harvard can assemble. From the abstract: "With a particular focus on voices and issues from the Global South, the symposium addressed topics such as inequitable access, risks to safety and privacy, skills and digital literacy, and spaces for participation, and civic engagement and innovation. The event also marked the launch of Digitally Connected — an initiative that brings together academics, practitioners, young people, activists, philanthropists, government officials, and representatives of technology companies from around the world who, together, are addressing the challenges and opportunities children and youth encounter in the digital environment." 130 pages PDF.
I can't even remotely do justice to this large work (which I'll be nibbling at for days) in a single post, so let me confine myself to a quibble. Nishant Shah writes, "The edge, then, is not an outer limit, but a route that marks the transfer of data from one point to another. Moreover, the nodes are also not predefined permanent points but rather points in a network that gain intensity (and hence value and valence) because of the frequency with which data travels and intersects at that particular point." No node is more or less valuable. If the node gains sufficient intensity as a result of incoming signals, then (according to a probabilistic signalling function) it fires, and the intensity is set to zero (or some other function-defined number). A node that fires frequently is not more valuable than one that fires rarely; indeed, extremes at either end subject the node as a candidate for deletion. Do not, in a network, confuse frequency with value. Ever.
Mean What You Say: Defining and Integrating Personalized, Blended and Competency Education
Susan Patrick, Kathryn Kennedy, Allison Powell,
International Association for K-12 Online Learning (iNACOL),
The sponsored a post in EdSurge linking to this white paper, and the paper is heavy with self-referential linking (so take some of it with a grain of salt) but it is on the whole worth a read as an outline of the major elements (and supporting technologies) for personalized learning. The key point (and probably why it's the subject of a marketing campaign) is that personalized learning and competencies go hand-in-hand. Well, this is true, if the object of learning is some sort of certification or standardized outcome. That's a pretty big 'if' and publishing companies are pushing hard toward making it a fact, because they see a gold mine in standardized learning materials. Non-standard outcomes, though, notwithstanding questions about certification, are far more valuable toward individual growth and development. Interestingly, many of the same technologies described in the report promote both standardized and non-standardized outcomes. (Note that OLDaily does not accept sponsored posts and receives no remuneration for any content posted in this newsletter).
This newsletter is sent only at the request of subscribers. If you would like to unsubscribe,
Know a friend who might enjoy this newsletter? Feel free to forward OLDaily to your colleagues. If you received this issue from a friend and would like a free subscription of your own,
you can join our mailing list. Click here to subscribe.