The Fracturing of Knowledge
Classical Division of Labor
Division of Labor is understood as splitting up some major task into its component pieces, then assigning a single worker or small group of workers to solve each part of the process.
For example, say we are assembling an hypothetical car on the final steps of a manufacturing pipeline.
Imagine, for the sake of argument, that the chassis of a car goes through the factory floor in the following fashion: it starts with a team installs in the motor and transmission; then a team screws in the doors, hood and trunk; followed by another team which puts in the headlights, tail lights and electrical installations; then another team puts in the windshield and other window panes; and finally a team screws in the wheels and puts pressure on the tires.
Time ----> t1 t2 t3 t4 t5 t6 t7 Single team: Car1 Car1 Car1 Car1 Car1 Motor+Transmission: Car1 Car2 Car3 Car4 Car5 Doors+Hood+Trunk: Car1 Car2 Car3 Car4 Car5 Electrical: Car1 Car2 Car3 Car4 Car5 ... Windshields+Panes: Car1 Car2 Car3 Car4 ... Tires: Car1 Car2 Car3 ...
In the first model, say a single skilled team works in a single car for 5 units of time, from beginning to end.
In the second model -- the pipelined model -- five different teams work each in a different aspects of the car.
We can see that, at its peak, the pipeline may be working at five different cars at the same time! The time it takes to assemble a single car is still about the same, but the output is (ideally) multiplied five-fold. Meaning that, by the end of the run, we would have 5 cars instead of a single one.
A reason this pipelined model of manufacturing has “won” is because each team can specialize to perform their own work more efficiently. It's easier to find and train electricians in just the electrical innards of a car, than to hire and train an all-rounder team who can assemble the whole car.
Another reason is that each of these teams can do their work in parallel with regards to the other team by working on cars at different steps of the process, which increases the throughput of the pipeline a whole.
The Choreography of a Medical Clinic
The pipelined model has been insanely successful in manufacturing. One can attribute as the man reason we can have incredibly cheap manufactured goods -- it is often the case you can buy a T-shirt or electronic gadget for cheaper than a plate of food at a your local restaurant -- is precisely because they can be manufactured so cheaply and at volume, offsetting the costs by a significant margin.
We can see an analogous model happening at a medical clinic. What is interesting in this case is that it is not goods going through a pipeline, but instead services and, very critically, information.
Imagine this typical clinic visit: starting at the door, a receptionist may get your ID and reason of visiting; then a nurse will do a preliminary check-up and admit you in; you will get to see a physician after some time; a nurse will come back and administer the prescribed treatment; eventually, you will feel better, ask the nurse to see the physician again and finally be discharged.
Unlike as with the car scenario we've explored previously, on the scenario above there's plenty of points where things might end up differently and the people involved might have to make different -- and appropriate! -- decisions.
This is the challenge of knowledge workers: continuously making a judgment based on complementary training and experience; to know when they can deal with something as they do routinely, or instead bring in help or escalate when the unexpected happens.
Regardless, there is also division of labor: the bureaucratic aspects are handled at the door are handled by the receptionist; the initial care is administered by the nurse; finally the collected vital signals and information are handed to the decision-making physician who prescribes the treatment plan.
The physician is as unaware of how the patient has been ID'd as the receptionist is unaware of what was the prescribed treatment. Along the chain of service, each of the links trust that the previous one done their part of the job correctly, they are presented with the correct information in their own scope, and they produce an information output for the next link in the chain, as best they can.
In Information Technology
As computers and IT platforms grow in scale and sophistication, there is absolutely no single person which has complete understanding at every level of the chain.
Physicists and material scientists study semiconductors and pholitography to improve the silicon -- and manufacturing processes -- which underlie the processors and electronics at the physical level.
VLSI engineers design efficient arrangements of transistors -- assuming that those work and provide expected logical outputs -- into micro-architectures and designs for microprocessor manufacturing.
Other engineers design the operating system and drivers which runs on top of the microprocessor -- assuming the instructions do what's advertised on the manual -- allowing for...
Application developers to write end-user software which runs on top of the operating system -- assuming the provided interfaces will command the hardware as expected --, ultimately delivering what users see on their devices every single day.
If any lower link in this chain breaks, the whole chain ceases to function.
What we saw in the CrowdStrike incident last month was the result of the Operating System assuming the third-party driver would not be faulty.
Other vulnerabilities such as Spectre, uncovered in 2017, relied on flaws at the micro-architecture level, breaking the security assumptions taken for granted on higher levels.
Computers which are sent to space need to be carefully vetted, since cosmic radiation may flip bits and cause circuits to malfunction, breaking the lowest level assumption that the silicon will behave as commanded (on Earth, the atmosphere and magnetosphere provide protection).
Recognizing the beauty of this orchestra, in contradictory terms of integration and separation, in both its fragility when things go wrong, and sophistication when things go well every day, I think is what Engineering and the sciences are all about.