A Taxonomy of Predictability

I did this as a post, but it was too short. Here's a simplified list of types and levels of predictability

Most of these are standard terms, used by folks in the appropriate industry.

Format

Name -- definition -- example

-----

Easy -- Problems that you almost always get right -- 2+2 => 4

Complicated -- Problems with many steps, where each step has to be right -- 67*23 => 60*20 + 7*20 + 60*3 + 7*3 => 1200 + 140 + 180 + 21 = 1521. 7 steps

Polynomial -- Clear number of steps to complete -- 67^23 => taking multiplication as 1 step, that's 1. 67^2, 2. 67^4 3.67^8 4. 67^16 5. 67^24 6. 67^23 => 5 multiplications and 1 division, with each of those multiplication steps being larger than the 7-step one above.

Chained -- Step 1 impacts step 2 -- Many calculus problems require several calculus steps, some trig substitution, some trigonometric identities, algebraic manipulations, and some reasonable amount of arithmetic

Magnifying -- Each error compounds -- computer floating point multiplication or division. There is rounding in every step, and so the distance between the actual answer and the precise answer grows with the number of steps completed.

----

Non-Polynomial (NP) -- One cannot find a closed-form solution. One can only find the right answer by trying a (large) number of choices -- The classic example is: Find the shortest path between 50 cities. You do that by trying (nearly) all the paths. Modern public key cryptography relies deeply on this difficulty in the factoring of large almost-prime numbers.

Hard -- Easy to make mistakes -- A perfect dart throw hits the area you want to hit. It's 93.25 inches from the toe of the thrower to the back of the dartboard, and the best target, triple-20, is about 0.4 square inches in size. The best darts players in the world miss triple-20 more often than they hit it. Top darts-players average near 90 for a Three Dart Average, as opposed to the ideal 167. Some problems are too hard for human skill to succeed reliably.

Cooperative -- More than 1 person -- Doing things solo is comparatively easy. Making it work together, where everyone has to do their parts in a way that fits together involves timing and some amount of communication. An Alley-oop, where one basketball player throws a ball to another who jumps, catches it, and slam-dunks the ball is an easy example of coordination, because there's only 2 people

Opposed -- You vs. them -- Many interesting activities are somewhat competitive. Success is not purely a result of your own efforts, but rather it's a comparison between your efforts and those of your competitor(s). Tennis is a simple version of this type of problem, and market-leadership is a harder one.

Changing -- The System is changing regardless what you do -- Despite the simplifying assumption used by everyone and their dog, systems are rarely static. Instead, they're changing, drifting, and adapting regardless of what anyone intends to do. And most of the time, the changes are hard to predict. Almost all systems are changing, most of the time.

Imperfect Information -- You don't know all of the relevant variables to make a good decision -- Poker is the classic example here. When your opponent raises in the small blind, and then reraises you and a straggler in limit holdem, and then raises again when a rainbow A93 flops, you know something about their hand, but not much. What you have is a range of probability, and that's all.

Probabilistic -- The results are random in a range -- Craps, Lotto, etc. -- At best you are betting on the probability of things happening in the future. There's a 1/6 chance that two dice thrown total seven. There's a 1/9 chance they total 9. You have to bet on probabilities, because we only have a range of low-probability outcomes.

Chaotic -- The formal name for the math of Chaos is Nonlinear Dynamic Systems. Systems where small mistakes don't stay small -- In weather systems, there is the famous adage that suggests that a butterfly's single wingbeat today in California can change the weather from sunny to a typhoon in China next week. Very tiny initial condition changes eventually make it so the system is thoroughly unpredictable. Many or most physical systems work like this.

Adaptive -- the discrete version of Chaotic -- The system is composed of many individual actors that each change based on local conditions. Conway's game of life is the most famous of these, but the behavior of a flock of birds or a school of fish matches tremendously well here. So does traffic flow between lanes.

Non-repeatable -- Your actions ripple through and change the system as well -- This means that you can't act, and then act again the same way to get the same results. In cooking, when a sauce is thickening insufficiently, then adding a little oil and flour may thicken the sauce to positive effect. But if the sauce thins again, and you add more, you may ruin the flavor as the flour/oil mix becomes too dominant in flavoring. This is a major concept in database theory.

Agentic -- Many actors, different goals -- There are different players in the system, and each of them has different goals. Adaptive, non-repeatable, and opposed. Harder as the number of agents increases. Getting a business to work well, when the shareholders, the executives, the management and the employees all have sufficiently different goals is hard.

Political -- Actors have goals they won't admit, the organization is too big to be person-to-person -- This is true in every group I'm familiar with over a dunbar size or so. 50-150 people maxes out the extent to which person to person interactions can dominate. Above that, political concerns begin to dominate outcome-centric concerns excepting in very rare leadership cases.

CAS -- Complex Adaptive Systems -- Many of the above pieces put together. Agentic, Changing, and Non-repeatable. A Free Market is the classic example of a CAS, where each actor changes the system when they add their knowledge, and then the system itself becomes (effectively) smarter than any of the individual participants, or even any small group. See Warren Buffett's 2007 bet against the hedge funds

VUCA -- Volatile, Uncertain, Complexity, Ambiguous -- The term the military uses for tricky situations. Roughly all military action combines (V) change, (U) imperfect information, (C) lots of moving parts, and (A) an inability to clearly understand the information one has.

Wicked -- The complexity capstone are the wicked problems. Problems where there isn't even agreement on what the problem is, and if there were, then the unpredictability and adaptive nature of the system makes any effort to change things once-off -- Peace in the middle east. Taxation. These are wicked problems.

----

Claim: Everything we do is infused with complexity/uncertainty at levels no one wants to admit.

The software business in particular, has no interesting problems in the 5 easy categories.

And that is why we do Agile.