Jump to content

Research talk:Account creation UX/CAPTCHA

Add topic
From Meta, a Wikimedia project coordination wiki

Notes

[edit]

I would change "prevent the creation of accounts" to "prevent or slow the creation of accounts". It's always about cost-benefit analysis, for both us and spammers. We know CAPTCHAs aren't a absolute firewall. The goal is to slow bots down, ideally by requiring human intervention or some sort of cost.

What is the source for "While the purpose of CAPTCHAs is to prevent spam, they currently deter only the most casual of malefactors. Dedicated spammers easily use a variety of methods to index our CAPTCHA images, either automatically or with human assistance, rending them a weak first step in our spam defense ecosystem." If we're explicitly saying it's weak, what is the research that backs that up? Also, weak compared to other anti-spam mechanisms, or compared to nothing? The same goes for "largely fail to prevent spam".

In the research questions, I would clarify that the first question is specifically interested in apparently good-faith human account creations (or add another question for this) in an equivalent time period (presumably, rates vary by time of day). Then, I would say that experiment one is designed to provide information on that. As I've said, I don't think it can answer the spam/illegitimate bot question in just two hours. Superm401 | Talk 19:57, 9 January 2013 (UTC)Reply

Changed prevent to hamper. You can just go ahead and make a little edit like that in the future, especially when it's a first draft. :) The source of the fact that they fail to prevent dedicated spammers is that's what we hear from Stewards, admins, and our comrades in Engineering (Chris just confirmed it to us yesterday, and Leslie Carr was just talking about how easy it is to break CAPTCHAs on Wikimedia-l). There is also some easily findable research literature about it. Steven Walling (WMF) • talk 20:11, 9 January 2013 (UTC)Reply

Logging actions which trigger CAPTCHA

[edit]

I think bugzilla:41522 and gerrit:40553 might be of some interest, as it would allow getting more data about the impact of the CAPTCHA on WMF projects. Helder 17:32, 14 January 2013 (UTC)

Obrigado, Helder. :) Steven Walling (WMF) • talk 02:04, 15 January 2013 (UTC)Reply