When we talk about risk there are lots of confusing concepts floating around. But risk management, guided by paradigms found in industrial controls, flight systems, enterprise class projects, and "fail safe" systems are a good way to look at the term "Black Swan," in the domain of software development.
Project Managers.net has a post about Black Swans. But this article suffers from the same issues most articles about Black Swans do, it's not about Black Swans, it's about poor risk management. Like those mentioning Black Swans in books and articles based on Taleb's work, there is a serious disconnect between how risk if managed in places where people die - either by accident or on purpose - and the much softer side of risk where political or editorial adjustments are common.
Here's the list from the post and a question about how "black" these Black Swans actually are:
- Original Apple computer - TRS 80, original mail order INTEL 8080 kit predated the Apple machine. What was not understood was the marketing power of Steve Jobs, not the machine. Then came the IBM PC, and then came the Apple marketing machine in 1984. No to Black with the history of PC's in place
- Google - legal search engines pre-dated Google. What was cleaver is the architecture of having users define what is at the top of the list. Google is a popularity contest.
- Terrorist attacks on Sept 11, 2001 - full prediction. The terrorist were known by intelligence agencies, but those capable of stopping them didn't have that information. The attack on the US was forecast long before 9/11, but no one listened.
- Facebook - yea OK, maybe no one could have forecast the power of vanity, except maybe Tom Wolf. But the notion of a community was around in the time sharing days of Boeing Computer Services and the orginal ARPAnet I used iin the defense industry. We had "groups" of geeks working on specific software for the US Air Force, where we exchanged all kinds of information around technical and personal topics. All the pieces were there, what was needed was 10 Million users, not just our small community.
- You Tube - obvious once you have Google in place. Oracle long ago (maybe 18 years) had a project with AOL in Vienna VA to stream on demand video using Time Warner's emerging cable backbone. Didn't pan out for lots of reasons. Worked with two forward tinkers in the newspaper business who used the term "convergence," for what is now the merge of video, internet, voice, and streaming over broadband and wireless (4GLTE).
- Twitter - yea, maybe
- Global financial meltdown - No Black Swan here, fully predictable. Read The Big Short and All The Devils are here, and the University of Chicago economics professors.
- Earthquake and Tsunami in Japan - nope, fully predicted. They just didn't listen. And they did every other bone headed mistake of operating a nuke plant. Look at the nuke plant operations journals. The risk assessment submission was one (1), yep that's ONE page long, and was written 10 years ago and NEVER updated.
- Recent massive tornado damage in the southern American states - nope, grew up in the Texas Panhandle (Pampa), where tornadoes came through every year and tore the living hell out of North Texas and Western Oklahoma. Predicted every year. Some years worse than others. I wnat to know why they always hit trailer parks?
A Black Swan is NOT defined as something that changed the world. It is a ...
Low Probability High Impact Event based on a long tailed probability distribution.
Attempts to redefine Black Swans just confuses everyone. I know that is defined later in the post, but those things in the list are not Black Swans, they are bad forecasts of completely predictable events, expect maybe Face Book, Google, and Twitter.
As mentioned, being prepared is important, but in actual high risk, high impact domains (where I work) like Nuclear Power, Manned Spaceflight, and similar domains have one critical attribute - They Are Fault Tolerant. Take a look at "Fault-Tolerant System Reliability in the Presence of Imperfect Diagnostic Coverage."
This means when the low probability event happens, the "fail safe" aspects of the systems are invoked. For emergency shutdown systems this means "fail to danger" is once every 100 million times using the SINTEF guide. These system are software based, along with some hardware, they are active software, fault tolerant software. fault tolerant in the same way flight control systems are. One example which I worked on (SW Dev Manager) is the Tricon. The referenced paper above is the theory and practice for the continuous diagnostics embedded in the Tricon to control things like the Westinghouse Frame IV gas turbine, pictured at the top of the post.
