Cooking the Figures

This is a story that really won't be news to some, but to many it might be. It's the classic case of thinking that the end justifies the means.

He knows what he wants his results to be. And the original numbers aren't giving him that data. So the agency he works for won't be able to persuade people to fight the war he wants to fight.

Well, that's not acceptable.

He starts with his software. There are certain procedures that are normal and accepted in his line of work. But if he makes just one little mistake, his program does a weird little recursion and if there's any data at all that shows the pattern he wants it to show, it will be magnified 139 times, so it far overshadows all the other data.

He can run it on random numbers and it gives him the shape he wants. Unfortunately, the real-world numbers aren't random -- they have a very different shape. All the numbers. Even his jimmied program won't give the results he wants.

All he needs is any data shaped the right way. And so he looks a little farther, and ... here it is. It looks, on the surface, like all the other data that he's been working with. Other researchers working in his field, just glancing at it, will assume it is, too.

But it isn't. Because the source that gathered this batch of data had some other key information that takes it all away. The numbers don't mean what they normally mean. In fact, this number set is absolutely false.

If you use these numbers along with all the other data, however, the clever little program will pick them up, magnify them radically, and voilá! The final report shows exactly the shape he needs the numbers to have.

The trouble is, these numbers are supposed to be doublechecked. Anybody who looks closely at his numbers and at his program will see what he's done. It's not hard to find, if you have the original data sets and can examine the program. He will be exposed as a fraud. It will do his cause more harm than good, if it's made public.

But he's not afraid. He knows how this works.

He doesn't show the program or the lists of his data sources to anybody.

To those of you familiar with this story, you find some familiar names: Mann, Santer, McIntyre. Those of you who don't recognize these names, go and read All In A Good Cause by Orson Scott Card.

h/t: SDA