Link Dump 2015-03-31

2015-03-31

Standardized tests can be more useful if only they measured the right things. Well, yes. The problem with the suggestions here is that they are not scalable - we can't measure samples of every child's everyday speech and analyze it for grammatical complexity, or have conversations with every single K-12 student to establish their patterns of thought. What you could do is do this for 20 students in college classes. Which explains why the author is a psychology professor at Williams.
Intriguing point about math textbooks by Math with Bad Drawings: they only define a term once, extremely precisely, and often the implications of that definition are unexplained or left as exercise for the reader. I've never been able to work through math textbooks either, and I wonder if math textbooks can be written to more closely resemble normal English (in terms of repetition), even if the content doesn't need it.
People like tombstones to stand the test of time - but not so well that it remains sparkling white. Gravekeepers don't like them either - it ruins the overall aesthetic of the graveyard.
Where do we draw the line between being open to debate and not wanting to be assaulted by contrary opinions? Why is the answer different if the subject matter is conspiracy versus if the subject matter is rape? If college is not the place to discuss dissenting ideas, where then should we do it? I'm not opposed to personal or communal safe spaces, but I don't think safe spaces should be forced on others.
An online dating site for threesomes. I like the idea, although I'm curious about the limit to three people.
http://aeon.co/magazine/philosophy/intellectual-character-of-conspiracy-theorists/ FIXME something about not liking the reduction to personality - no long responsible for finding the truth?
I wasn't aware that there were "manly" languages. Although I guess I do stereotype programmers; for example, what do you mean you program in MATLAB? (I kid - I'm well aware MATLAB is perfect for a lot of engineering/number-heavy applications.)
Images that computer vision algorithms think are things... that are definitely not things (or even images of them). The art aspect of this is fascinating, since we're essentially using a learned classifier as a constraint (eg. how un-peacock-like can an image be (to a human) while still being an image of a peacock (to a computer)?). On the computer science side, I wonder if there are necessarily such images. These images are only interesting in so far as they highlight differences in how humans and computers view images - these are images where the computer thinks its one thing when people think it's not that thing. Of course, the opposite is not as interesting: images that we think is one thing when computers don't think it's that thing are simply called errors. Regardless, it seems to me that if we get computer vision systems to do exactly what humans do, that system could not have been used to produce these images.
Similar, urban camouflage for subverting facial recognition software.
If you've played Kerbal Space Program and have redirected an asteroid... well, NASA's not going to do that, but it's doing somethin'.
I've had a number of conversations about diversity in computer science in the last couple weeks, so this article is well-timed. Especially as I look ahead to helping Oxy establish a computer science department, I wonder what kind of ground rules and what kind of culture needs to be mixed into the foundations, and how I can do it. This may warrant a separate blog post eventually.
Even though I'm not teaching the same course next semester, I will probably have a hand in hiring the student instructors who will. This blogpost on advocating for minorities during hiring is something to keep in mind, and I'm sure I will have to do this again in a couple years at Oxy.