Fool Me Twice, I Might Believe YouFool Me Twice, I Might Believe You

A friend of mine seeking a cure for a bitter heartbreak was contemplating a Machiavellian remedy: If she could seed a rumor that would reach her ex's new flame through multiple, seemingly independent sources, the rumor would surely work to split them apart. The mechanism at work would be the Repetition-Validity Effect, a principle of human psychology that we tend to believe a "fact" the more often we're exposed to it. As insidious as the Repetition-Validity Effect is when consciously manipulated, as you can see abundantly demonstrated during an election season, it's even more damaging when it's unintentionally put into motion — when there's nobody to recant, no lie to identify. In an increasingly automated world, mistakes can be replicated at lightning speed, and what seems like corroborated intelligence can actually be the echoes of errors.

People complain about having meeting-room battles over multiple versions of the truth. But at least that situation invites scrutiny of the "facts." When everyone agrees on numbers that are bad but unsuspected, the result is an ironclad form of groupthink ... and we all know what sorts of disasters groupthink can lead to. Josh Greenbaum's Enterprise Applications column goes into more detail on this subject. It provides a necessary reality check even while we enthusiastically explore the positive possibilities of the future of BI and enterprise (including "extraprise") applications.

Greenbaum cautions that the increasing automation and interdependencies among companies creates a greater need for quality assurance. One way to improve our ability to discern truth from data is better control over metadata, such as reference data. Rajan Chandras discusses this problem and a limited solution in his RazzaDS product review.

In principle, intelligence can increase with the amount of data, the number of people able to access the data, and the number of sources of data. In "The Land Beyond Transactions", Stewart McKie describes, for example, how location analytics take advantage of new sources of data from RFID chips and GPS devices. But those and other sources of data can lie outside our ability to manage: "As companies make use of external Web services over which they have little or no control, the garbage-in/garbage-out problem becomes even scarier," Greenbaum writes. McKie believes, however, that XML standards such as XBRL will make data from a multitude of Web service-accessible sources suitable for something he calls metadata mining. In the case of financial data, he says, metadata mining "can drive new and improved financial analytic processes that will use Web services to find, download, compare, contrast, and consolidate financial data without the need for a business or financial analyst's intervention." Considering the penalties public companies can face for putting out inaccurate data on SEC filings, this source could be pretty safe — but bear in mind that no company is required to ensure the accuracy of the XML markup of its SEC documents.

As with everything, moderation seems to be the key. The more mission-critical or potentially damaging an instance of automation or the analytic conclusion is, the more important it is to control or verify the sources.