Cargo cult engineers: Will we ever realize that correlation is not causality?
By Dick Morley
Do you ever think that we are overdoing engineering? Are we playing in our elitist sandbox without thinking about the user? My porch in rural New Hampshire was built and designed by craftsmen and engineers. It had to keep the occupants warm and safe. Over the last 200 or so years, each owner has made some changes, maybe 10 in total over the years.
This is not so in software. If we used software over the same period, we would have thousands of changes. I use Apple computers and am up to OS 10.5.2. Do I need a hundred versions of the OS to type this column.
A few months ago in this column, I mentioned the scorpion and the fox fable. I think it bears repeating. A fox is on the bank of a deep stream and wants to get on the other side. As he gets ready to swim, the fox hears a voice ask, "Can I cross the stream on your back? I am a scorpion and cannot swim."
"But you will sting me and kill me if I carry you," the fox replies.
The scorpion responds that if he stings the fox, they both die. The fox decides that this logic is flawless and invites the scorpion onto his back. Off they go.
Halfway across the stream, the scorpion stings the fox. "Argh," cries the fox. "Why did you sting me in this deep water? We will both die."
The scorpion replies, "Because I am a scorpion." Are engineers the scorpion of the laptop?
Our human embedded code is sometimes our enemy. Where do our actions hide inside us? Actions are determined by correlation and emotion, not logic. If, for example, you are standing in line and you get $100 because you are the one-thousandth person to purchase a ticket, you feel good. However, if the person in front of you got 100 bills and you got 50, you feel anger. Why? You just got fifty bucks. What is the problem? Our innate position is that we strive for loss aversion and rank – we ain’t logical. Feelings and emotions dominate our behavior.
One of the most famous false correlations was the Cargo Cult. During WWII, we delivered supplies by plane to the Pacific Islands. Since no runways existed, the SeaBees bulldozed a runway and prepared flares for the planes. After the war ended, the locals cleaned the runways and lighted fires, hoping for appearing cargo – a misplaced correlation. One of my heroes, Richard Feynman, lamented the cargo cult scientists because they do "flawed research that fails to produce useful results." Maybe we should define cargo cult engineering as engineering that makes undesirable results.
The cargo cult also affects the medical profession. My bride of more than half a century had significant pain for half a year. We took her to specialists of all kinds. They all seemed to suggest that the problem could be solved with the tools of their particular specialty. We saw nerve, brain, bone and other specialists. Finally, our new family doctor said, "I don’t know." We then expanded the event horizon of tests and found the answer. Severe arthritis. The specialists are smart and concerned for Shirley’s health, but only had special hammers for the possible nail.
Many stories like these are oft told in the technical community. We are not condemning real mistakes, but are bringing up the flawed idea of correlation vs. causality. Sometimes bad stuff just happens. Let me list some cargo cult mistakes of the past:
• The Pinto market survey
• The Edsel market survey
• The PowerPoint "cause" of the shuttle explosion
• A 66-page manual for a flat screen TV
• The sinking of an oil well platform in Brazil
• The Big Dig vs. the Chunnel design
• The transistor watch
This last one is fascinating. When us techies discovered the integrated circuit and looked at our watch, we said, "Son of a gun. We could build a low cost accurate digital watch with this technology." Most of us can remember these watches. Today we are back with a broad spectrum of pricing and both analog and digital watches. The silicon guys don’t really design good watches. Why? Two important unseen factors. Watches are pieces of jewelry and geeks don’t do jewelry. And the other? Nobody wants to know what time it is, we want to know how much time we have. Digital watches only seem to work for the money people who deal in the digital world.
We all should reread The Mythical Man Month by Fred Brooks. He said that adding manpower to a late software project makes it later. Brooks said that his book is called The Bible of Software Engineering because "everybody reads it but nobody does anything about it!" Software projects take longer to do when more people are assigned to a project – a clear case of continuing to follow the old paths, even when data contradict the plan.
Don’t get me started on PowerPoint and flow charts. The links between objects or assets are exponentially related to the scalar growth of objects (people). It is very difficult to pen a collaborative novel. Most of us write a specific chapter and the other author(s) edit. A different approach to collaborative authorship is interesting. The first half of an adventure novel is written by David, who gets the hero in trouble, while Jason writes the second half and solves the hero’s dilemma. Neat. Gotta try that someday. PowerPoint gives bullet statements with little or no backup. It is a concept sell, not a design review. Working on detailed flow charts without understanding the user domain is folly.
We must be able to step back and use wisdom and analysis in a controlled way. Risk analysis, design reviews and, most of all, understand the basic physics and human interface. Are we playing in the sandbox? How about becoming professionals with balls? Do not let your tool shape the solution, otherwise, we will all drown together from the scorpion’s barb.
Dick Morley is the inventor of the PLC, an author, speaker, automation industry maverick and a self-proclaimed ubergeek. E-mail him at firstname.lastname@example.org.