 It can be a little intimidating trying to predict what disasters the future might hold, but you should always at least hazard a guess. So, the election. Lots of people very worried about the outcome, lots of hand-reading about what the future of the country looks like, or how life might change for certain people. If you're of a certain frame of mind, like I am, you may have spent some time imagining specific events and phenomena that probably won't happen, or probably won't happen to you, but become uncomfortable possibilities in such a tense and unpredictable political climate, things like power grabs and riots. Engineers in particular tend to develop a sort of reflex for rumination on all the myriad ways that things can go horribly wrong, as it's often a useful habit when your whole job is to ensure that things reliably work the way they ought to. Ideally, engineers anticipate potential points of failure and figure out how to avoid them entirely, but sometimes, well, shit happens. Regardless of how much of a bummer it is to think of your beautiful, finely-tuned design exploding because some idiot decided to do something dumb with it despite it being very clearly labeled, it's important to think about what's going to happen if that situation comes to pass, and what can be done to contain and limit the damage, action plans for how to roll with and respond to problems if and when they do come up. Unfortunately, as anyone who's been alive in 2020 can tell you, the number of things that can go wrong is practically endless. Global pandemics, killer hornets, police brutality, wildfires that blot out the sun for weeks at a time, trying to come up with and implement strategies to deal with all this stuff ahead of time would not only require some serious divination magic, but a whole lot of time and resources. Analyzing, developing and implementing solutions for every single possible problem that crosses your mind isn't just an endless undertaking, it tends towards bike shedding, optimizing solutions to the small problems that are easy to deal with rather than the big hard problems. How can we be sure that we're putting out the closest, most pressing fires first? No pun intended. This sort of thinking has led engineers to create an analytical tool specifically geared towards identifying which risks require the greatest attention and focusing effort to efficiently mitigate them. In the 1940s, the US military developed the failure modes and effects analysis, or FMEA method, a version of which is now used by just about any organization you can think of that does some sort of engineering, as well as several that don't. FMEA turns out to be such a good way of thinking about risk and how to handle it that institutions like libraries and coffee shops use it to reduce risk in their internal process. And if you're looking for a way to organize your thinking about everything that could go wrong in the near future, you could do a lot worse. Now, I'm sure that some of you have started looking around for literally anything to pay attention to that isn't an Excel spreadsheet. But think of it more as a set of prompts, a way to hold everything that could go wrong up to the light and inspect it from every angle to make sure that you're not missing anything. Then to take decisive action to stop disaster. It's not a table, it's a time machine through which your future self can communicate what went wrong and how to prevent it this time. First, we're going to list out all the functions that are critical for things to work smoothly. Ideally as pairs of verbs and nouns, open doors, select floor, close doors, lift passengers. If you're talking about company processes, think about necessary steps in the flow chart, things like receiving orders or paying employees. Importantly, we're operating at kind of a high level here and only considering what actions are necessary for things to keep chugging along. Don't get bogged down with stuff like not break, screw one, not break, screw two. Once you have a solid list of these verb noun functions, a complete process or successful operation from beginning to end, then you can start brainstorming every way that things can go pear shaped, both the mode of failure or how our functions can get screwed up and the likeliest cause. What event is probably going to participate that mode? Cable snaps caused by fatigue or maybe loading the elevator beyond capacity. Doors jam caused by external obstruction. Buttons get stuck, caused by some jerk kid stuffing chewing gum into them. All the horror in mayhem goes here. It may be unintuitive, but we're only going to assume that one function fails at a time. That's not really how the world works, as 2020 has shown, but we're trying to identify the cornerstones of this system, the eigenvectors, the jenga pieces that everything is balanced on, and that requires focusing on each element of the system individually. Now, we consider the likely effects of these failures and their severity, assigning a number that indicates how big of a deal it is if something goes wrong in this particular way, generally from one to five, where one means that things can keep working as intended and five means that multiple deaths may occur. Gum in one of the elevator buttons is annoying and inconvenient, but although it will cause the elevator to repeatedly stop at that floor, things will otherwise keep working as intended. I'll give that a two. On the other hand, the doors getting stuck is going to put the elevator out of operation, which is kind of a bummer, but trapping people inside is also a big deal. What if someone was on their way to the bathroom? What if someone's claustrophobic or has some sort of medical emergency while they're locked in a steel box? I'm going to give the doors being stuck a four out of five because if it happens, it might well end up killing someone. The cable snapping, yeah, that's an easy five. Next, we're going to consider the probability of these sorts of things happening, again, from one to five. One would be a freak occurrence. Five means it's a foregone conclusion that is going to happen at least once. Finally, we're going to assign one more number that indicates how detectable the failure is, how quickly someone will know about it if it happens. This is important because sometimes it's not clear that anything is wrong until it's too late to prevent the worst possible outcome. If someone cuts your car's brake lines, the danger isn't really in the failure itself, it's in the way that you find out about it. We multiply our three numbers and get a series of weighted values for each failure mode. The higher numbers indicate things that are deadly, probable, and hard to spot. From here, you'll have some clear candidates for highest possible risk. Obviously, if you have the resources to head against everything, great. But taking care of the largest probability mass of disaster first is really your priority, and the FMEA method can highlight some potentially unintuitive aspects of where that might be. For our elevator example, you might have thought the cable snapping would be the top of the list for things to worry about, but its weighted number is actually lower than the one that we came up with for doors getting stuck. With modern materials and quality control measures, it's actually a pretty rare occurrence, while people are constantly dropping stuff into door mechanisms. Armed with that information, you can work your way down the list, figuring out how you're going to lower some of these numbers with additional features or strategies. It can be helpful to think of decreasing each of these assigned values directly, making the effects of failure less severe, lowering the probability that they'll happen, or finding ways to discover them before they're a real issue. To mitigate the risk of doors getting stuck, we could add an escape hatch and evacuation routes if people really need to get out. Add protection to door opening mechanisms, so it's less likely that someone will drop their keys into something important. Mandate routine inspections and maintenance to make sure that they're not getting worn out or gummed up with dust and debris. When you're talking about machines, these sorts of measures are often simple changes that drastically reduce risk in obvious ways. Problem X, do Y, now you basically don't have to worry about it. When you're talking about reducing risk in life, although it can be nice to simply change a thing and eliminate some unnecessary hazard, like keeping the fire extinguisher somewhere that's not likely to be on fire, often the only thing that you can do is to have a detailed plan in place, a strategy that you can pull the ripcord on and execute to come out in the best possible position when some risk beyond your control comes to pass. The best contingency plans have two parts, a discrete easily determined trigger, some metric or event that you can evaluate quickly which will prompt you to get moving, and a detailed step-by-step set of actions that don't require a lot of interpretation or improvisation to satisfy, hopefully with some redundancy or alternate paths forward in the event of a roadblock. If things get dangerous, go somewhere safe is a bad plan. If there's a wildfire within 25 miles of home, put these items in the car. If an evacuation order is given, grab the dog and drive to this location, unless it's also on fire, in which case drive to that location. That's better. Of course, a lot of the stuff that's made 2020 such an interesting year has been very hard to predict, let alone to form action plans for. The unknown unknowns are lurking out there ready to plant side you and ruin your day. The same is true of machines and companies. No matter how hard you fill out your FMEA spreadsheet, there's always something unimaginable that you're not accounting for, a monkey wrench ready to drop into your whole apparatus. But staring down the known risks and finding realistic ways to cope with them frees you up to deal with the other stuff, to improvise and adapt where necessary without wasting time or mental energy, trying to react to something that you probably could have seen coming. What comes out is the biggest risk in your FMEA calculations. And why is it heart disease? Are you maybe going to develop a solid plan to exercise a little more and eat a little better? Please, leave a comment below and let me know what you think. Thank you very much for watching. Don't forget to blah blah subscribe, blah share. Before future Josh gets here, if you'd like to maybe avoid doing this next election, please check out electionscience.org. They're a lobbyist group pushing for approval voting, which might break up a little of the insanity caused by our first passive post-elect plural system. Anyway, like I was saying, don't forget to change this man. Yep, it's me, future Josh. Just a reminder for anyone watching this, episode 200 is coming up. I'm going to be rewatching some old episodes to comment on them. If you want me to sit through one in particular, be sure to visit the link that's in the description and vote for your favorites. All of these favorites.