Can tags lie? Of course they can. But this is usually not a problem because
incorrect or misleading tagging typically causes trouble for the very same
people who are doing the tagging. This gives them an incentive to get the tags
right. And, if the tags aren't right, there is an incentive to fix them.
Consider an XML-based publishing application. I want to get the tags right so
that the presentation comes out right. Or, in a syndication application, I
want my tags to be semantically, not just syntactically correct, because I want
someone else to use and link back to my information. Even in an XML-based
commercial transaction, where there might in fact be more incentive for me to
have the tags tell lies -- increasing the quantity of goods shipped, for example
-- the external controls already built into the transaction (counting the
quantity of goods received, for example) create an incentive to ensure that the
tags tell the truth, reducing overall processing costs and ensuring repeat
business.
All of this changes when we use XBRL to communicate financial information to
analysts and investors. The incentives to misrepresent information or, in some
cases, to hide it altogether, are substantial. This makes XBRL different from
many other XML applications and requires a different approach to validation.
This is not just a detail. The shift from intrinsic incentives that help get the
tagging right to a need for external controls changes the way XBRL is used. It
also adds to the list of capabilities that must be in place to build an XBRL
market.