One of the first things any problem solver does on the way to a solution is to ensure a correct understanding of a problem. The next step is understanding the desired results. These two steps are remarkably consistent from math to hard physical sciences like chemistry and physics to any of the so-called soft sciences like sociology or psychology and even economics. Knowing your starting and ending points will greatly accelerate resolution.
Failure to nail the first two steps can result in much frustration and wasted time and resources. But sometimes we can’t even describe the problem accurately because our vision of a solution requires new or not yet invented technologies and history is full of examples. For instance, from very early times humans have wanted to fly like birds and unfortunately “like birds” was a major stumbling point because humans don’t have the anatomy for avian flight.
A human’s strongest muscles are in the legs whereas birds have skinny legs and big chest muscles to power their wings. Birds have also adapted to life in the skies by jettisoning unnecessary weight, they have small heads with beaks instead of teeth and legs and feet are just enough to do their jobs. Birds also have very rapid heartbeats to deliver a constant and rich supply of fuel and oxygen to their muscles. For instance, hummingbirds’ hearts can beat at 250 beats per minute–at rest–rising to 1200 beats during flight. A good resting human pulse is 60 bpm.
Yet human history is replete with stories about heroes fashioning wings to fly with, such as the story of Icarus who made wings of gossamer and wax then flew too close to the sun where the wings melted and a crash resulted. Leonardo drew plans for wings intended to be powered by human arms and there are even vintage early movies of people attempting to fly by flapping wings only to crash.
We all know that the Wright Brothers were the first to fly a heavier than air machine. They invented modern flight by disregarding all of the received wisdom about it and conducting their own inquiry that included multiple hypotheses and experiments on the windy shores of one of North Carolina’s barrier islands, Kitty Hawk.
The Wrights were really good at the first two steps. One of their first decisions was to give up on human powered flight. Next, they studied wings using a wind tunnel that they had built to discover the optimal shape for creating lift. Finally, they had to devise a steering system that would enable control in all three dimensions.
The Wrights didn’t get everything right by today’s standards. For instance, their control system required enough wing flexibility for twisting to enable control. Modern aircraft use flaps for the purpose. Although their control system was later superseded, it was at least consistent with the successful principles they developed. The Wrights succeeded where others failed because they were first to truly understand the problem. They were also lucky to be alive when the internal combustion engine had been perfected. The bicycle mechanics built their own, out of aluminum to save weight. In that they were thinking like birds.
So what does all this have to do with Facebook? Simply put, Facebook has a quality control problem that many parties have variously and mistakenly defined as
· A size problem, the company is too big and needs to be broken up (maybe, but other companies are bigger).
· A management challenge, Zuckerberg doesn’t have the management chops for this (quite possible though he’s been learning on the job for quite a while).
· A technology issue, we just need more tech like AI and machine learning to spot the bad guys. This is the fallacy of sunk cost, we’re in too deep to change course.
Time out. Let’s consider this from a quality control perspective.
Understanding the problem
According to multiple sources including a recent story in the New York Times, the company is working hard to come up with artificial intelligence and machine learning tools to catch abusers of the social network before they can do harm. They’ve effectively defined the problem as catching bad guys.
The effort is being led by Mike Schroepfer, Facebook’s chief technology officer but according to the Times,
“…Every time Mr. Schroepfer and his more than 150 engineering specialists create A.I. solutions that flag and squelch noxious material, new and dubious posts that the A.I. systems have never seen before pop up — and are thus not caught.
The trouble is that it’s an arms race and every time Facebook comes up with a solution to one issue, others inevitably arise. The company is playing Whack-a-Mole and by definition with this approach it will never catch up.
The situation is analogous to what the US manufacturing industry went through in the 1970’s and 1980’s. At the time US manufacturers, especially in the car industry, were having their clocks cleaned by the Japanese who were, ironically, effectively using American expertise first developed by W. Edwards Deming–statistical quality control.
US car makers initiated their own quality programs that attempted to add a quality control step at the end of the manufacturing process. It was essentially an inspection regime and cars that failed inspection would be diverted to rework so that defects could be repaired. This worked well enough for cosmetic issues, but it was costly and for things like engines and transmissions that were intended to run for 100,000 miles or more but broke well before that, no inspection process could suffice.
The Japanese in contrast used a system of statistical quality control. Statistical quality control is just what it sounds like. It’s an effort to capture deviations from specifications in all components as they are made. Parts that deviated less than tolerance limits could be passed on in the manufacturing process, those that deviated more would be rejected. In this way, Deming’s idea was to build quality into a product at every step in the manufacturing process.
Facebook’s lack of QC
Facebook isn’t manufacturing anything, so QC for them would be different. But at their core Facebook is focusing on policing everyone’s finished products and removing the bad ones rather than trying to build quality into their product.
Some people might take minor issue with using the term “product” because it’s well known that with the advertising model, the product is the Facebook user sliced and diced and delivered to the real customer, the people paying for information. But in fact, Facebook is delivering eyeballs of increasingly questionable value to its advertisers. The people paying for information increasingly don’t know if the demographics have been messed around with; they don’t know if there are bots in the data that they really want to be live people with at least the potential to make a purchase. Advertisers also don’t know if prospects are avoiding their ads just because.
No doubt about it, Facebook has a quality control problem.
Despite the huge numbers involved (i.e. billions of users), Facebook’s problem is amenable to a solution but not the solution that Facebook fervently hopes would be its salvation, which is to say, more technology. As noted in the article cited above, the company is currently investing huge sums on artificial intelligence and machine learning to identify bad actors but that’s like inspecting a car for the first time when it’s already rolling off the assembly line. What Facebook needs is a way to build quality into its use processes as its product is being used.
Technology might be Zuckerberg’s blind spot; it’s what he knows and it’s his hammer in a world ostensibly populated by problems that all look like nails. If only.
I’ve noted before that in our democratic approach to free markets, we let the marketplace more or less police itself. The IRS exists but its enforcement efforts all concentrate on catching violators after they’ve had multiple chances to do the right thing by hiring accountants and tax attorneys or simply hiring a tax preparation service. The IRS isn’t looking over your shoulder waiting for you to screw up which would be par for the course in a totalitarian regime.
The IRS and tax preparation are far from the only examples too. Virtually every profession is governed by laws that professionals adhere to because of training and certification. This system works well with barbers, plumbers, doctors and lawyers and many others. In every case a practitioner takes education and/or performs an internship to learn the profession and then proceeds to a professional certification. That’s how these professions largely govern themselves and keep the public safe but it’s not how social media is administered.
Anyone can just begin using social media whether it’s to connect with friends or to attempt to influence a million people. For the majority of users that just want to connect with friends that’s fine. But for influencing a million people there should be guard rails.
So the proposal is this: let ordinary users with 100 or 1,000 friends continue doing what they do. But for corporations, governments and bad actors who want a cheap and easy way to spray paint the world, time out. You need to prove your credentials first. You also have to identify yourself when using the social network so that problems can be traced back. Traceability will do much more and do it more simply than a massive investment in AI and ML.
Breaking up Facebook might be a good idea but that’s not the point. Frankly, anything that gets them to convert from a business model that sells customer information and into a model that provides a safe marketplace is an improvement. What’s needed right now is a way for the company to get back to delivering high-quality service without the distractions, exploitation, and outright sleaze inherent in the service today.
Taking a quality control approach instead of one that polices and punishes bad behavior will make the service more robust while greatly reducing the number of bad actors corrupting the service today. A quality control approach is also inherently less expensive so why not try it?