I was reading the wiki article about the Tokyo Stock Exchange, and came across this paragraph:
During the initial public offering of J-Com on December 8, 2005, an employee at Mizuho Securities Co., Ltd. mistakenly typed an order to sell 610,000 shares at 1 yen, instead of an order to sell 1 share at 610,000 yen. Mizuho failed to catch the error; the Tokyo Stock Exchange initially blocked attempts to cancel the order, resulting in a net loss of 347 million US dollars to be shared between the exchange and Mizuho. Both companies are now trying to deal with their troubles: lack of error checking, lack of safeguards, lack of reliability, lack of transparency, lack of testing, loss of confidence, and loss of profits. On 11 December, the TSE acknowledged that its system was at fault in the Mizuho trade. On 21 December, Takuo Tsurushima, chief executive of the TSE, and two other senior executives resigned over the Mizuho affair.
What the! I can perfectly imagine the scene in front my eyes: there's this dialog box with two text fields, the employee enters the details, and clicks 'OK'. A confirmation box pops up, for probably the 18,000th time that day. No patience to even click, his right pinky hits the Enter key, and whoooooooooosh! 347 million dollars, and three top executives go right off the cliff. And if you really think about it, there is nothing anyone could have/can do about it. Selling 610,000 shares at 1 Yen is a very probable transaction, and so is 1 share at 610,000 yen, so no 'common sense' software check could have caught it. Confirmation dialogs are useless in such high volume work, we all know that. From the side of the employee, it simply is an error that can't be avoided!
Nai had some things to say about this during Shaastra Main Quiz, as these kinds of errors being related to Freudian Slips. I too have been searching high and low for research on 'casual' or 'silly' mistakes that are so common in numericals, but haven't been too successful. At their best, these casual mistakes affect performance assessment (in my experience) by about 5% , and at their worst, render the assessment meaningless (a mistake in CAT, for example).
Is these something 'behind' casual mistakes, something more that 'silly errors' or 'it happens'? I'm trying to design a survey form, and send it by mail to everyone in the insti, and try to get some statistics about it. If you know of any such survey being conducted before, or have some suggestions for the form, please do let me know!