tea.mathoverflow.net - Discussion Feed (Reading test before posting an answer)

Harry Gindi comments on "Reading test before posting an answer" (5445)

2010-05-07T10:32:47-07:00

The recaptcha on MO is completely useless. Why would a computer take 20 minutes to automagically generate a post? Also, for people who have at least, say, 100 reputation, this simply shouldn't be an issue.

Harald Hanche-Olsen comments on "Reading test before posting an answer" (5437)

2010-05-07T07:40:04-07:00

The two words aren't supposed to be related. MO is using recaptcha. The clever idea behind it is that these are words from scanned books that the OCR software had trouble with, so they solicit the help of humans to figure out the correct meaning. One of the words has already been solved by another user, and is used for verification. The other is entered into the database to be used for future verifications.

Of course, in an ideal world, recaptcha should have data from several languages, so you can select a more familiar one. But this is not such a world.

Olivier comments on "Reading test before posting an answer" (5436)

2010-05-07T06:53:06-07:00

I see. Well, this means I can escape these tests: all I have to do is prepare my answer some place else and then paste it to MO. Thanks a lot.

Unknown G. comments on "Reading test before posting an answer" (5434)

2010-05-07T05:52:55-07:00

This seems to be related to a previous discussion: http://tea.mathoverflow.net/discussion/346/yes-i-am-a-human-being-at-least-for-mo-posting-purposes/#Item_0

If I understand correctly, the "reading test" appears only if it takes more than 20 (?) minutes to compose/edit a question.

Olivier comments on "Reading test before posting an answer" (5433)

2010-05-07T05:22:44-07:00

Before posting an answer (or for that matter a question) on MO, one has to succeed at a reading test safeguarding against automated spam. Now, I don't consider myself handicapped, but I happen to find these tests extremely difficult. For one thing, they usually involve uncommon word strings in English, but almost by definition, this implies that non-native speakers will have a hard time inferring the second word from the first (except if I missed something about this system). And though I usually don't have serious difficulties in reading hand-written note (when they have a meaning), I find myself hesitating in about one out three cases with these human verification tests. I have checked with many people that I am uncommonly incapacitated in that respect. So even though I fully understand the practical use of this system, I wonder if there are easy to implement ways to make those tests easier to complete (maybe there are various levels of blurring).

Before posting this, I tried to make a test to see the alternatives available by default, but sure enough, for some reasons I can't fathom, my dummy answer was not subjected to a test for the very first time.