An imbalance in the Ontario health care system

broken red cross

Our health care system is too brittle.

I think of design as an activity that seeks balance between efficiency and effectiveness. So understanding effectiveness and efficiency, and being able to recognize efficiency and effectiveness in systems, is a fundamental skill in design. In this post, I will describe a situation from my own life of hyper-efficiency in the Ontario health care system. The point is to demonstrate that the notion of balance is useful to explain why systems don’t always work properly.

Continue reading

Advertisements

An overly effective system

Grand Cayman Airport

The airport on Grand Cayman Island.

I have many examples of hyper-efficient systems, but I lack examples of hyper-effective systems.  I’ve found one, though, in the Grand Cayman airport.  Chronically overcrowded, horribly organized, and highly unstructured, it is also incredibly malleable.  Both the structure and the staff seem able to accommodate nearly any situation and adapt to it easily. It doesn’t matter how late the flights are getting in, or how slow the security checks are, or how hot it is, or how far out into the parking lot the check-in lines are, the staff seems to just spontaneously come up with ways of jigging things to work out.

Continue reading

Snow, airlines, and balance

Heathrow Airport is snowed in

Heathrow, and most other airports, are brittle systems.

London’s Heathrow Airport – suffering only a few centimetres of snowfall – is largely shut down.  Thousands of stranded passengers may end up spending Christmas in one of the least Christmas-y places there is. Dozens of other airports are carrying huge backlogs because of the cascade effect. The rippling of delays and cancellations is wreaking havoc all over the place.  This is a great example of an unbalanced system that has forsaken effectiveness of efficiency.

Continue reading

A note on collaboration and organizational design

linked hands

Is the "market" a collaborative activity?

One colleague put me on to a very simple yet very crisp and practical distinction between coordination, cooperation, and collaboration.  More than a year later, another academic suggested three levels of organizational behaviour: hierarchical, clan-based, and market based.  It struck me that these are essentially the same.

Continue reading

Lights out in Dörentrup, but it’s too complicated

In Dörentrup, Germany, you can literally call up the streetlights.  It saves energy, but at what cost?

On 30 July, Time posted an article about the town of Dörentrup, Germany, where the local council recently voted to turn off streetlights at night to save energy and carbon emissions.  This naturally caused a fuss.  What’s the point of streetlights that are off when you need them?

Town resident Dieter Grote and his wife, working with the local utility company, came up with a solution.  You can now use your cell phone to send a special ID code to the utility, and the streetlights with that ID (grouped by stretch of road) will come on for a certain length of time.

The energy saved by turning off the lights at night saves the town (population, around 9,000) about 12 tons of carbon emissions per year.  That’s not bad.  And other towns around the world are asking Dörentrup for help to set up their own similar systems.

While it’s fine and good to try to lower consumption – especially when no one needs the service – I think their solution is too complicated.  First of all, there is a carbon footprint associated with the machinery and electronics needed to keep this new service running; I don’t get the sense that anyone has compared this footprint with the alleged savings.

Next, this system ties the utility infrastructure directly to the phone system quite deeply, in a rather centralized way.  You call the magic phone number, which no doubt accesses a facility in one of the utility’s plants, but by way of your phone provider.  The centralized system then has to understand the ID code, which can be dialled in or spoken (requiring voice recognition software).  The IDs are stickered onto every streetlight.  The system then has to direct the pertinent streetlights to turn on, keep track of the time, and then direct them to turn off.

If something goes wrong at the system’s home base, the streetlights won’t come on, even though they otherwise could.  If your phone’s battery dies, or you lose signal strength, you can’t control the lights, even though they would respond otherwise.  If you don’t have a phone, or your hands are otherwise occupied carrying groceries or whatever, you might not be able to use the phone – again, you can’t get the streetlights to work.  If you can’t read the ID sticker on the streetlight (because the streetlights are off), then you might enter the wrong ID or be unable to figure out the ID at all.  Vandals could change the ID stickers.

These are all failure modes that don’t need to exist.  There are probably others that I haven’t thought of.

I would suggest something simpler.

Residents would carry some kind of semi-active sensor, like a battery-assisted RFID tag, or the GPS locator already in many cell phones.  A small unit would be attached to every streetlight, that can detect the presence of a nearby RFID tag or GPS locator.  The unit then turns on the streetlight, and keeps it on till the signal moves out of range.  This can be rigged to ensure that two or three streetlights are on for each person.

Assuming these kinds of sensors have the required range (and I think they do), I think this is a better solution because:

  • no hands are required to activate the system;
  • no hardware/software needed at the central utility facilities;
  • power for the streetlight unit can be drawn from the streetlight itself;
  • the RFID battery can last ‘way longer than a cell phone battery;
  • no need to remember phone numbers or look up streetlight IDs;
  • this kind of distributed/decentralized system is more robust and resilient than centralized alternatives; and
  • fewer streetlights would be on, and for shorter periods.

In other words, I think my suggestion is more effective, without giving up efficiency.  It compartmentalizes the whole on-demand streetlight system so that failures anywhere are less likely to affect related systems.

Seems like a no-brainer to me.

The down-side of high efficiency

High efficiency comes at a price we probably don’t want to pay – low effectiveness.

I should have noticed this years ago, when I first learnt the laws of thermodynamics.  For those of you not inclined to scary math, the laws of thermodynamics – thanks to C.P. Snow – are really quite easy:

  1. You can’t win.
  2. You can’t break even.
  3. You can’t quit the game.

We have yet to find something that violates these laws anywhere in the universe.  Notice the second law, which basically says nature is never 100% efficient.  (Efficiency is the rate at which we produce output compared to the rate at which we consume inputs. To be 100% efficient, you’d have to produce absolutely no waste at all.)

Now consider modern society, which seems entirely obsessed with ever-increased efficiency.  Improving efficiency isn’t a bad thing in itself.  The problem is that efficiency, I believe, is only increased at the expense of something just as important – effectiveness.  Effectiveness is the extent to which a specific action, when done under ordinary circumstances, achieves intended goals.

Put another way: efficiency is how easy a thing is, and effectiveness is how well it’s done.  And we need both, in roughly equal measure.

In terms of evolution, both efficiency and effectiveness matter.  I once heard a story on CBC Radio (sorry, I have not got a reference for this – but if any reader does know the source, please let me know) that told the story of two types of squirrels found in American cities: extremely active ones that gather food like mad (and burn more energy doing it), and rather lethargic ones who are not especially good gatherers (but conserve energy in the process).

The question was why evolution would have not selected away one of these relatively opposite traits.  The answer turns out to be not very surprising.  In years when there’s lots of food, the “best” squirrels are the ones who can hoard the most food, which they can eat to produce enough milk to suckle their young.  On the other hand, in years when food is scarce, the “best” squirrel is the one able to conserve energy to, again, be able to suckle its young.

Since the quality of the food supply is cyclical over a period shorter than the average lifespan of the squirrels, the best balance to ensure that squirrels generally survive, is to have a reasonable subpopulation of both energetic and lethargic squirrels, even though about half the population will not be efficient in any given year.  So the overall efficiency decreases, but the overall effectiveness increases.

In creatures that can think abstractly, there’s another important aspect of effectiveness.  Since effectiveness applies to meta-level actions – like planning (an action that arranges future actions) and deciding (an action of deciding which other action to perform), effectiveness covers traits like adaptability, flexibility, and resilience.  So these traits become markers of effectiveness, and their lack become markers of ineffectiveness.

I am convinced that if humanity is ever going to get its act together, we need to learn to balance efficiency and effectiveness.  Too much efficiency will make systems brittle and ineffective, while excessively effective systems will have very low efficiency and create too much waste.  Somewhere in the middle is the sweet spot that gives the best result overall.

Simple example: as the fuel efficiency of automobiles has improved, the total distance that people travel by car has also increased.  You can do the math yourself with data for passenger cars available at the American Bureau of Transportation Statistics.  You can find how many cars were in the USA between, say 1980 and 2006, as well as how far they drove in total, and their average fuel consumption.  A little arithmetic and you’ll learn that while fuel efficiency was 24% higher in 2006 compared to 1980, the average car was driven 35% further.

So, at least in the USA, even though cars are more fuel efficient, they’re still burning more gas.  Higher efficiency is fooling people into thinking its okay to drive more, which negates all the reasons for improving fuel efficiency, and makes us less effective both as commuters and stewards of our planet.  (Not that the whole world is like the USA – thankfully it isn’t – but I haven’t found corresponding data for other countries.)

Software is another example of how efficiency and effectiveness interact.  There are some insanely specialized pieces of software out there.  Consider software for computational fluid dynamics; this stuff is used to simulate the behaviour of very specific situations of fluid flow (basically liquids and gases) that can’t be resolved by working out the equations.  This software is wonderful stuff that has helped us design better airplanes, pipelines, and arterial stents, to name a few things.  CFD software is extremely efficient at what it does.  But it’s also incredibly specialized.  If you don’t pose your problem in just the right way, or if the conditions you’re studying are even slightly outside the very domain of the software’s range of abilities, you get total crap for output – assuming the software doesn’t just crash.  So while it’s efficient, CFD software isn’t very effective.

On the other hand, look at something like Google Docs, a free, web-based alternative to packages like Microsoft Word.  In addition to supporting an essentially infinite number of extremely varied objects (documents, images, etc.) it can also be extended (i.e. be given other functionality) through the various APIs that Google has published for Docs.  This very significant flexibility makes Google Docs very effective.  But it can be slow, and a little clumsy to use – that is, it’s not very efficient.

(Some digerati might complain that Google Docs is in fact very efficient from a strict computer science point of view.  And they’d be correct.  But from the point of view of the user, this would be arguable at best.)

Here’s another classic example of efficiency and effectiveness: us!  The human body is really rather inefficient and flawed.  We’re not as fast as we could be, or as strong, or as perceptive (i.e. our eyes, nose, ears, and tongue all suck most splendidly) – especially compared to certain other species (cheetahs, apes, eagles, dogs, etc.).  While other animals may be more efficient than us, in one way or another, we are more effective: we learn better and faster, and we are more adaptable, than other creatures, and we are just efficient enough to get the job done.  This gave us the edge we needed to basically take over the Earth.

There’s also plenty of examples of systems that are too efficient, which leads to brittleness that causes sudden and catastrophic failures as soon as the situation changes a little too fast or too much.

One example is the blackout of 2003.  This blackout, which affected all of Ontario and pretty much the entire north-east United States, was caused by a single power plant in Ohio that went off-line because of high-demand, followed by a couple of overloaded transmission lines that failed when they came in contact with overgrown trees.  The rest of the system failed in a cascade effect.  The power grid is incredibly efficient at generating and distributing power all over North America.  But it is also incredibly brittle (unable to adapt, or be resilient to failure) because these two small failures, under just the right circumstances, affected some 55 million people.  Again, the highly efficient system is also highly ineffective.  If it had been more effective, there would have been mechanisms in place to prevent the bottom from falling out as it did.

Another example is the financial system – certainly the US financial system, and to a lesser extent, the global financial system.  This system is very efficient at generating and distributing wealth.  But the current global economic crisis can be traced back to a single equation.  That equation was used, and abused, by a variety of financial analysts who simply didn’t understand what it really meant and how it was to be really used.  The result is thousands of foreclosures, bankruptcies, hundreds of thousands of people out of work, and a huge loss in savings wealth.  So for all its efficiency, the financial system was woefully ineffective.

What we’ve got is a society where efficiency is King, and effectiveness ain’t even allowed into the Castle.  What we want is more balance.  We don’t want too much efficiency, and we don’t want too much effectiveness.

There are, as far as I can tell, two major changes that have to happen in the design endeavour to get a better balance between efficiency and effectiveness.

Change #1: We need a fundamental change to design thinking.  This is the general thinking style or framework that guides all designers, whether they know it or not.  It’s a mental guide for how one thinks and acts that is driven by design.  Design thinking is different from analytic thinking (what scientists and mathematicians do) and artistic thinking.

The fundamental change is that design thinking should be based on seeking a balance between efficiency and effectiveness.  In any situation where one might apply design thinking, one must consider the degree to which each of efficiency and effectiveness matter, and then devise ways to bring them together in the right proportions for that situation.  Until this becomes totally ingrained and ubiquitous in design thinking, then I don’t think we’ll get very far in building a better world.

Change #2: We need to change how we optimize things.  Basically, we optimize everything for efficiency – weight efficiency, cost efficiency, fuel efficiency, whatever efficiency.  We’ve known for a long time that if you optimize a thing too much, then you wreck its functionality.  So normal operating procedure is to establish a minimum acceptable effectiveness, then optimize the efficiency, constrained to the minimum acceptable effectiveness.    So we end up with a solution that’s highly efficient and only minimally effective by definition.

But there are other ways to skin this cat.  What we really want is to find the “sweet spot” at which efficiency and effectiveness combine to produce the best overall solution.  This kind of problem is called a multi-objective optimization.

To get a sense of the kinds of improvements that a proper multi-objective optimization can produce, consider the work of Bing Ye, who did his Master’s thesis under my supervision some years ago.  The gory details are in a paper we wrote together.  But in a nutshell, Bing figured out a new way to trade off quality and cost with respect to setting dimensional tolerances in engineered parts.  He compared his method to four other well-used methods.  His method always beat the others, sometimes by as much as 140%.  Bing has had a very successful career since then, and I am honoured to have been a part of his education.

The point is this: the math of his work applies to the matter of efficiency versus effectiveness just as well as it does to quality and cost.  In very general terms, it’s staggeringly easy.  First, measure efficiency, and measure effectiveness.  Multiply each measure by a weighting factor to account for the relative importance of each in a particular situation.  Add the two values together.  This is a measure of the overall “goodness” of that particular arrangement.  Now, start fiddling with the various values in the calculations, looking for the greatest possible result.  That will be the best balance of efficiency and effectiveness for a given situation.

Of course, actually doing all these things is much harder than I may be implying here, but that’s really all there is to it.

All this is eminently possible.  There’s not a single showstopper in any of it.  Sure, it will take some time, effort, and money.  But the result will be far, far more useable and meaningful than how we do things today.