Black Magic

Math is hard for most people. It's confusing and requires different thought patterns that have to be cultivated. Starting with addition and moving up the ladder of operations and subjects, most people start to get a brain cramp around algebra, glaze over at geometry, and stop paying attention at trigonometry. This isn't any different in the military. The percentage of people who are capable with math is higher than the normal population (among the officers that is, no data on the enlisted ranks), but it still is quickly confusing to most soldiers. Especially the kinds of funky things we end up doing. Most of these guys were taught to shoot guns, drive tanks, or blow things up, not differentiate or run analysis of variance. We often joke about our models being the "Black Box of Mystery". Repeatedly adding numbers in a spreadsheet is treated like black magic. I'm sure calculus would be like alien technology. It would be funny if it weren't so darn inconvenient.

Math addled brains cause two large headaches in my life.

First: because most leaders do not understand math at a deep level, they trust you to understand it, and convey it to them correctly. If you gain their trust by presenting them with solid analysis, they will be more inclined to trust you in the future. If you screw up, and lose their trust, it is very difficult to get back.

Unfortunately for us analysts, numbers can be made to say anything. Inserting bias in numbers is extremely easy if you know what message you want to convey. (For example: a few years ago a river boat gambling proposition came up on the ballot in Missouri. The advertising campaign for this measure claimed that "84% of residents in X area voted 'Yes' on this measure in June in order to get it to a state wide vote". In actuality, 84% of the voter population had turned out to vote, and the measure had passed. The message of the ad implied that 84% of people had voted yes, when in actuality, 84% of people had VOTED, and the overall answer was yes. The difference is subtle, but striking, and was pretty tough to catch in a 30 second TV spot.)

As a result, those whose reputations are established on the accuracy of our number crunching, such as analysts like myself, must be very careful to represent our analysis as accurately as is possible. We are painstaking in our work, often having multiple co-workers check over our calculations for mistakes. We vet the assumptions our work is based on through the subject matter experts, so that we can reasonably conclude that our work is based on solid foundations. We are purposefully spirited away in a dark corner of the command structure so that there is no possibility of a general officer barging in our door demanding that we modify a spreadsheet. We regularly accompany briefers during their meetings to make sure someone is present that can accurately explain the analysis, and characterize the results. And, after learning the hard way, we never, ever, trust our raw data with anyone outside of our office.

That hard way I speak of. Oh yes. There are those who use the numbers we produce for evil. Aye. Evil. There have been instances where organizations have taken our data out of context. *Gasp* Unpossible, you say? It gets worse! There have been instances where organizations have taken our data files, changed the data points, and ran to the boss with the new "analysis" that fit their agenda. Dastardly villains they are. We don't like them. Of course if any questions are imposed upon these people they simply throw up their hands and say "The ORSAs gave it to us!" And short of having a duel with 9 MM pistols in the general's office, we usually have to swallow our vomit and take the beating.

The alternate route of deception is more underhanded. Recently we performed analysis for a group that I would consider shaky. The numbers were fine, but there were bunches and bunches of "IFs" that had to be met to get to those numbers. "If" we do this and "if" we do that then the whole world just looks better and better. And imagine, no one was talking about the probability of meeting all those "ifs", even when probed. Repeatedly. But that was ok, as we had been invited to the meeting where our analysis was being briefed in support of the larger topic. Only, at the last moment, our meeting invite was rescinded. No seats left in the general's office. Sorry. No credit for your work, and no reality check on our crazy assumptions. But we'll make sure we throw you under the bus when this all crashes down on our heads! Ok now? Buh bye! /simmer

Of course there are consequences for evil doers such as these. These naughty agencies no longer have access to our data unless we are in the room while they present it, or better yet, while we present it for them. They also don't get the raw data charts that can be edited, but rather picture files of the charts that show exactly what we intend to portray. Where trust has been breached, it is not easily re-earned. And after being burned a time or two, having our name and reputation attached to shoddy and untruthful work, we've learned it's just easier to keep all the data under wraps and only pass out the final, vetted results. Even the potentially legitimate requests get turned down now, for the most part, simply because we don't want two different answers being computed off the same data source using different methods. Math is scary. How would you explain to a three star general that the same data generated different answers and still get him to trust you at the end of the brief? It's not easy, so we just don't do it. It is detrimental to information sharing, I agree, but an unfortunately necessary action to prevent outright fraud with our numbers.

Second: because most leaders do not understand math at a deep level, they ask for impossible analysis. Why are we still in Afghanistan after 9 years? I want the answer shown in one chart, on my desk in the morning. Thanks so much.

This is by far the more common of my two headaches. Cretins intending to misuse our work do so at their own peril, and they know it. Leaders asking for two weeks of analysis to be done in 5 hours just don't know any better. Just today I had someone ask why attrition changed from last month to this month. My e-mail reply: Why did it rain Sunday instead of Wednesday? It just did, man. I can't point to ONE reason, or even five. Numbers rarely have a big neon arrow pointing at the cause of its perturbation. Rarely is there only ONE cause! In the case I was asked about today, there were literally thousands of tiny causes, few of which are being tracked and none of which can be directly affected. Asking this question is akin to asking an artist to draw the beach down to the individual grains of sand: time consuming, and adds nothing to the overall product.

We try hard to educate people on this, we really do. But the military mentality of "complete the mission, damn the cost" gets in the way. A lot. In the face of mathematical ignorance, what are genuine reasons as to why the task cannot be completed sound like excuses for failure. As such, a lot of the time we suck it up and pull an all-nighter or four. We save our push back for the most impossiblist of impossible tasks, so that people know we mean business whenever we say no. If they could be done easily in time, and we're asked to do them in no time, we generally say yes. Crazy tasks we have undertaken that fit this description include major modifications of the software we use to project how many people will be working in the Army and Police on less than 24 hours notice, and building Power Point slides from scratch less than 10 minutes before they are to be briefed to a general officer. Tasks that were absolutely bonkers, like rebuild our main model in one day (would take AT LEAST a week) have been successfully rebuffed. Our sanity is better for it.

One thing you can say for certain about this place: you won't be bored.

*****

My next post will not be about work. I promise. In fact, I hope that my next three or four won't be about work. I'm tired of writing about the office and all that stuff, and I think I've covered it well enough for right now. I will move on to other topics. I don't know what yet, but I'll figure it out as I go. I always do.

All for this evening. Out here.