Tuesday, September 30, 2014

The swiss cheese model of safety

A recent fence jumper at the White House seems to have made it--incredibly--inside the building a fair ways before being taken down. Reading about the multiple failures across various security levels--alarm boxes being silenced, undercover agents outside missing the climber, attack dogs not being released etc.--I am reminded of the 'swiss cheese model of safety'.

In this model, each security layer can be thought of as a slice of cheese. Because all security measures have flaws, the slices contain holes that allow failure opportunities to slip through. Typically these 'holes' don't line up, so a failure at one level is caught by another level. But since modern security systems are complex and dynamic, occasionally a failure can proceed all the way through the system and become realized.

The key insight of this model is that each security layer contributes only in a probabilistic sense to the overall security system. This presents a challenge to managers because employees operating largely within one 'slice' can easily get complacent and start taking shortcuts (like turning off the alarm boxes because they were loud and annoying), increasing the opportunities for overall failure.

Friday, September 26, 2014

Some thoughts on home-cooked meals

Sarah Kliff has an interesting interview about new research into home-cooking habits viz. income. A few thoughts:

1. This type of small-n ethnographic research is not designed to elucidate facts across large populations, but rather develop a rich, thick understanding of specific families' food lives. The incredible local variation in food habits means this approach is often better than large-scale statistical aggregation surveys. In fact, food habit research probably has more in common with ecology's concept of 'patch dynamics' than anything else.

2. Pre-made rotisserie chickens from supermarkets are an underappreciated strategy for quickly supplying a cheap, healthy meal. In fact, the reasons for why these chickens are so cheap remains a somewhat controversial topic in economics.

3. The folk theory that many poor families eat unhealthy takeout food often because of its cheapness and convenience is challenged here, interestingly. Interesting follow-up data for me would be to see how and to what degree families press young children into service preparing and purchasing food, especially across incomes.

4. Much household turbulence seems to stem from the time-consuming nature of preparing good meals. Although the science hasn't been sufficiently popularized yet, I assume that in the near future very-low-carbohydrate diets will be seen as an effective strategy for reducing the social burden of healthy food prep among all socioeconomic cohorts. This is mostly due to the increased ability to go for longer periods without eating on this diet.

5. How could the internet and the 'sharing economy' reduce the difficulty of gaining access to healthy home-cooked meals? An Uber-style distributed peer-to-peer meal preparation network could accomplish this, as well as increasing employment and utilization of underused assets like kitchens and kitchen utilities. I assume the regulatory hurdles could prove difficult to overcome, unfortunately.

Tuesday, September 23, 2014

What is a good age to die?

Ezekiel Emanuel has written a provocative essay explaining why he only wants to live to age 75. This is definitely some good clickbait bioethics, to be sure. But probing a little more into his position, it's actually a great example of a widely-held misconception about death.

Emanuel lists a bunch of reasons why 75 is good age to quit, ranging from physical, cognitive, and social. The subtext to his claim, however, is this: 'given the current state of various medical technologies and social norms regarding age, 75 is a good age to quit'. This is still a perfectly fine attitude to have, but its implications for policy and the social approach towards death are quite a bit different. Namely, these structural factors that determine Emmanuel's 'quit point' can change. And we have the power to change them.

My favorite children's book
Emmanuel chose 75 and not 70 because his personal calculation about the opportunities for flourishing above that age don't outweigh his other concerns. But a few decades ago this number would probably have been lower, because the types of opportunities he sees as still available at age 65 would have already become inaccessible. Medicine and social norms have changed enough such that one's late-60s and early-70s can be pretty great, especially if you're rich and influential.

This brings me to the core point: across most cultures and institutions today, there exists a general lack of appreciation for our ability to change the two structural factors involved with ageing--medicine and norms. Let's put aside norms for the moment, as these tend to be very adaptable, and just focus on medicine. There's a huge consensus that age-related diseases like Alzheimer's and cancer are terrible and should be eliminated via technology and innovation. When you ask people if aging should be eliminated via technology and innovation, however, you get all sorts of confusing and bizarre arguments and rationalizations. Which is understandable given the horror of the topic. But here's the thing: aging and age-related diseases are the same thing.

Aging in the abstract is merely the accumulation of various age-related biological damage, and we're getting better and better at finding ways of repairing this damage. Which is great! But unfortunately the conceptual block between aging and age-related pathology prevents the issue from receiving the attention and funding it deserves (it kills lots of people every year). When a medical innovation comes along and makes being a 70-year old a little better, people cheer (if they notice at all). But they fail to place it into the wider context of history's incremental improvements in medical technology. And so we get powerful, influential people talking about how 75 is a good age to die instead of talking about how new technologies have the potential to make 75, 80--even 100--a great age to live.

Information Nexus

1. E.O. Wilson waxes philosophical and part 2
2. Scaling up urban agriculture
3. Vox's asteroid coverage continues
4. A meal at Noma
5. Brain Pickings review of Sam Harris' new book

Monday, September 22, 2014

Engineering a Safer Central Park

It's chaos out there... but why?
Photo Credit: NY Magazine
There's been a lot of commentary in urbanist circles the past few days about the death of Jill Tarlov in Central Park, who was struck by a spandexed road cyclist on the primary loop road. While its obvious that the cyclist was 100% responsible for the tragic crash, an interesting question concerns the specific conditions that lead repeatedly to these sorts of accidents (and countless near-misses). How does the texture and design of the Central Park loop encourage or discourage safe behavior by both cyclists and pedestrians?

I propose a somewhat counterintuitive theory: that the existence of signalized intersections (of which there are many) actually contribute to the dangerous pedestrian-cyclist interactions that occur on a daily basis in Central Park. First, consider that spandexed cyclists will always try to conserve their speed (rightly or wrongly). Due to congestion and road disrepair, New York City has few local opportunities for cyclists to achieve sustained high speeds. Rightly or wrongly, this means the Central Park loop will attract many aggressive spandexed cyclists.

Next, consider that the vast majority of pedestrians don't utilize structured intersections to cross the roadway--they don't need to. People can simply cross the gap anywhere (it is a nice green public park, after all). Cyclists, meanwhile, use the ample road width to subtly adjust and filter through pedestrians, joggers, slower cyclists etc. and proceed on their way. This is what happens over 99% of the road space on the loop. The mutual appreciation of shared space forces a level of awareness and emergent cooperation that minimizes the potential for crashes.

Typically when cyclists approach red lights in the loop there are no pedestrians around, so they just roll through. No harm done. Conversely, when a pedestrian faces a 'do not walk' sign, there are often no cyclists around, so they just cross. Again, no harm done. At signalized intersections, however, this all breaks down. What happens at intersections during peak times is that the unofficial norm of mutual filtering comes into direct conflict with the explicit directions of the signal lights. This creates chaos in several ways.

Some pedestrians follow the signals, and some follow the norms. Pedestrians who follow the signals and wait for a 'walk' sign clump up and create a large group of people that, when finally crossing, minimizes the gaps available for spandexed cyclists to filter through. This creates a dangerous situation.

Similarly for the cyclists, some stop at red lights, and some don't. When many (typically casual) cyclists stop at red lights, they clump up and create a barrier that minimizes the gaps available for spandexed cyclists to filter through. Signalized intersections, perversely, create the congestion which leads to unsafe pedestrian-cyclist interactions.

Signalized intersections are extremely expensive to maintain, and should be replaced with other passive traffic-calming devices that would better encourage safe interactions between pedestrians and cyclists (and cars, although given their low volume its surprising they're still allowed in the park). Signals were designed for auto-centric roads, and their continued existence on Central Park's pedestrian- and cyclist-dominated loop serves no purpose other than to confuse and muddle what should be pleasantly mutualistic and safe human interactions. So let's save some money, save some lives, and give Central Park back to its people.

Tuesday, September 16, 2014

Sousveillance and ecological change

Most wearable tech incorporating always-on sousveillance devices is currently being used for experimental, largely silly purposes (like life-logging). As these technologies become cheaper and more ubiquitous, they will increasingly be applied to novel situations and processes that produce real value. One potentially useful application of the Narrative Clip is to use it to catalogue environmental changes in ecologically stressed regions. Scientists, historians, and civil bureaucracies like the U.S. Forest Service would greatly benefit from having access to a constant, searchable database of photos covering the same region over time. Such a rich data set (generated by say, lending hikers a Narrative Clip to wear during their trip) would likely enable computational ecologists to understand physical changes in patches of wilderness in a completely revolutionary way. Get out your grant-proposal-writing pens now...

Tuesday, September 9, 2014

Can we just ditch plurality-rule voting systems already?

It's been known for a while now that every voting system has fatal flaws. Most elections in the US use some 'first past the post' method, where the ticket winning a plurality of votes wins. This popular rule allows for some truly perverse outcomes where a block of voters' wishes is sliced up and a niche candidate--opposed by a big majority--wins the election. We see this possibility unfolding tonight in the New York lieutenant governor's race:
"The key problem for the governor is that Cuomo-Wu and Cuomo-Hochul would count as votes for different pairs, and would effectively split Cuomo's vote between two tickets. ... If the Cuomo-Wu and Cuomo-Hochul votes are split enough, though, the Republican nominee Rob Astorino — viewed as a long-shot contender — could theoretically sneak to victory with a small plurality.
This unlikely scenario would be a great example of the winner of an election being determined not by the electorate's vote but by the characteristics of the specific voting system being used. But the disturbing truth is that every election outcome is highly contingent on the voting rule in place, not just the attention-grabbing shockers. The only thing we can do is decide as a polis what voting system we want, and which sorts of groups we want to help and hurt. I for one support instant-runoff voting, or better yet the Borda Count, which both tend to mitigate plurality-rule's twisted logic somewhat and favor more moderate, consensus candidates.