2

Since computing power has increased very much lately, there is opportunity to simulate very complex systems or at least assist human activity in various fields. E.g. assist medical doctors to quickly analyse medical images

I am wondering if any authority, at least at local level (e.g. mayors) is using some kind of software assistance when taking important decisions. By important I mean decisions that may affect multiple aspects of human life, not just the economical dimension (e.g. electoral, religious or ecological consequences) and their effect is very hard to evaluate.

I am thinking about using complex software simulations like Democracy (maybe adapted at local level which seems more manageable than an entire country).

Question: Are there authorities that use complex simulation software when making decisions?

  • I answered this question based on the premise that it is about any governmental authorities, but I see you've tagged it specifically with "local government". Did you want an answer focused specifically on local governments? – indigochild Jan 25 '18 at 19:17
  • @indigochild - I could not find any other tag. Anything above local community (town halls) level is fine with me. – Alexei Jan 25 '18 at 19:53
  • Not a real answer, but the Dutch water authority (RijksWaterStaat) has fully given over decision power to complex simulation software, for the Maeslantkering. Mathematical estimates of the risk led to the conclusion that humans are just not reliable enough. – MSalters Feb 1 '18 at 0:22
10

Yes, there are - but it isn't fantastically wide spread.

My basis for this answer relies on my professional experience. I work for my state government (in the United States) as an auditor, and my major claim is that I use computer simulation to analyze government programs.

In general? Yes

I have personally observed many different simulations being used. Here is a quick list of some cases where simulation is used in government:

  1. The United States Forest Service has a number of simulation tools for simulation the occurrence and spread of wildfires. These programs are used to plan for wildfires and to help deal with fires that are occurring. For an example, check out the FARSITE software This tools are used at the federal and state level.
  2. There are prison population simulation tools that corrections agencies can use to forecast the prison population in the coming year. This is used for financial and managerial planning. Here is an example of a report from the state of Kansas which uses such a simulation.
  3. Combat modelling is an entire discipline. The Winter Simulation Conference (the premiere venue for simulation research) usually has an entire track of presentations dedicated to simulating combat. Westpoint (a military training institution) has a website that describes some of their educational simulations. Sometimes these tools include more than loss of life indicators, such as environmental damage.
  4. Although I don't have a link for an example, there are an array of ecological simulations used by the United States federal government. They include meteorological, climatological, predator/prey, ecological, and other models. These are often used to analyze trends in the natural environment, which is valuable for agencies that work in those fields. For example, the Environmental Protection Agency uses simple Monte Carlo methods enough that they have published guidance on the subject.
  5. You specifically excluded economic analyses from the question, but of course economic simulations are relatively common. For example, the Congressional Budget Office used a simulation when analyzing the Affordable Care Act in December.
  6. Governments at all levels use simulation to analyze traffic patterns. This is one case where simulation is actually common. City planners and traffic engineers often use simulation to understand how proposed changes to roads (including adding roads, changing stop light configurations, etc.) will impact traffic patterns. It's so common that the Federal Highway Administration even published some guidance on simulation projects.
  7. The European Union has developed a simulation tool for evaluating the safety of nuclear reactors (thanks to ohwilleke for pointing out this one).
  8. The National Aeronautics and Space Administration uses simulated extraterrestial environments to help develop technology for their missions. (thanks again to ohwilleke)

Like Democracy? No.

Democracy attempts to analyze an entire country in a relatively holistic way. That is too general to be useful to good government in most instances. They typically want to deal with a particular program with particular goals. For this purpose, you should really have a simulation tailored to the program.

Simulations also get exponentially more complex as you add more concerns. A simulation which tried to be that holistic would likely be a very large project. Not impossible - but difficult.

Why not more?

I don't know of any specific religious simulations (which was part of the question). Most simulations are focused on "hard" problems which have measurable outcomes.

Why not use simulation more? And why not include more subjects?

  • Simulation requires specialist knowledge. Not just anyone can create a simulation with a reasonable degree of validity. You can't even hire a computer programmer or statistician to do it - it really requires a unique combination of skills and experience unique to computer simulation. These people can be hard to find sometimes.
  • Simulation is expensive. Creating a simulation typically requires contracting experts - and they aren't cheap. The project is also going to be far longer than an ordinary statistical analysis. It's more of an engineering project than pure analysis. Expect to spend several months (at least) on a simulation of even relatively small scale. Meanwhile, if you are tracking that data in your agency's systems you could perform a reasonable statistical test far quicker. Or you could make a judgment call and accept the risk in no time at all.
  • Simulation is still fairly obscure. It comes from the world of operations research and industrial engineering; fields which don't generally penetrate public administration. Public administration professionals aren't taught about these methods, so they are largely unknown. Professionals from those fields are unlikely to move to public service, so there is no way for this approach to really get to the public service world.

Some people in industry are trying to create tools specifically for public policy use. One such company is Forio. Although I am unfamiliar with their products, they claim to have several customers in the public policy space.

  • 3
    I would add (1) dynamic budget forecasting by the Congressional Budget Office and Treasury department with similar tools used for state level fiscal impact estimates in many states for appropriations as well, (2) NASA mission preparations, (3) weather forecasting, (4) some police departments trying to predict crime hotspots for purposes of allocating resources, (5) cost benefit analysis by the EPA, (6) nuclear regulatory safety evaluations, (7) lots of big transportation (road, bridge, transit) investments are modeled this way, and (8) building code exception decisions e.g. novel skyscrapers. – ohwilleke Jan 26 '18 at 0:03
  • @ohwilleke Are you sure those all involve simulation? I just read the CBO materials on dynamic budget forecasting and there doesn't seem to be any simulation involved. Building code exceptions and safety seem especially unlikely to be simulated - but of course I could be wrong! – indigochild Jan 26 '18 at 1:35
  • @ohwilleke I stand corrected, many of those ended up being entirely accurate (including the nuclear safety model, which I was skeptical about). My apologies! And thank you! – indigochild Jan 26 '18 at 1:51
  • 3
    I was an applied math major in college. We wrote a fair number of simulations along these lines for governments as class projects and studied lots of others. Also, often you will use a mix of analytic models and monte carlo models for the same problem. In these areas the line between what is and isn't a simulation can get grey, the word "simulation" wasn't one that was used very often, "model" was far more common and a simulation is a subtype of a model. Agent based, stochastic, iterative rule based, and input-output models and some linear models all pretty much count as simulations. – ohwilleke Jan 26 '18 at 2:36
  • 1
    I'm in the simulations business. In it, we have many phrases like "In God we trust; all others, bring data" and "All simulations are wrong; some are useful." Unless a field is very well defined, there is always a tremendous amount of art in interpreting the results from the simulation. They're rarely used as the sole decisive factor. – Cort Ammon Jan 26 '18 at 16:30
2

There is a story about the city of Hangzhou that is managed using an AI New Scientist

Using AI to optimise Hangzhou has had many positive effects. Traffic congestion is down, road accidents are automatically detected and responded to faster, and illegal parking is tracked in real time. If someone breaks the law, they too can be tracked throughout the city before being picked up by the police.

In this case the AI Runs numerous simulations with iterated changes and combinations to determine the most effective changes to make to achieve the defined goals. Then provides the best results to the policy makers.

All of the oven nine million people living in Hangzhou is tracked. Their activity on social networks, their purchases, their movements, their commutes and it's fed into the AI's database and it then does real-time decisions like controlling traffic lights, informing authorities about upcoming crisis and providing real-time information about traffic and weather to the citizens.IFL Accidents can be detected within a second and traffic police can be on site within five minues.People's Daily

  • 1
    The article doesn't mention using simulation. – indigochild Jan 26 '18 at 14:19
  • @indigochild - That is effectively what the AI does. It just does the analyzing automatically instead of feeding the results to analysts. – SoylentGray Jan 26 '18 at 15:40
-2

It depends, how you define authorities and description making.

There are some simulations made to show what consequences decision may have, especially in natural science based problems. Best example are the countless simulations about, what will happen if we react upon it and in which way.

But mostly there are problems of simulations. They are models and therefore incomplete but very complex in the same time. Which means that for once, they are unreliable and depended on the creators interests, as well as they don't really explain much to the untrained viewer.

You can see that again with climate change, show these simulation to meteorologists and biologists and they are extremely scared, but show it to the normal person and they just moan about the inconvenience of getting rid of their Pick-up because of two degrees more in the summer.

That said, with learning algorithms we will very likely see more simulations in the future, since they can be far more complete and independent than classical algorithms. Not to mention that they can directly make decisions upon their findings and so control countries policies way more efficient than we ever could.

  • 3
    Answers should be backed up. Do you have any references to cite or any particular relevant experience? For example, your claim that simulations don't explain much to the untrained viewer is unusual: that's typically one reason that stakeholders love simulations. They can just see the results right in front of their eyes (unlike diagnostic statistics which require experts to interpret). – indigochild Jan 25 '18 at 21:55
  • It is difficult to reference, I tried to find something with numbers but, there is just a bit of hot air from a view companies, that sell these kinds of solutions, and not much else. For relevant experiences, I am a computer scientist, and while I didn't make any simulation software in itself, I know how calculations and therefore predictions in computer work. And I know what customers are interested in, when they ask for such stuff. To make it short, if someone wants "simulations" or "statistical predictions", they usually want something shiny that seems to back them up, not real simulations. – Etaila Jan 25 '18 at 22:05
  • Sounds like our professional experience differs (which is entirely reasonable), but I would still recommend backing up your answer by citing your experience. – indigochild Jan 25 '18 at 22:19
  • Citations required. This answer neither explains, nor provides any source for the numerous claims of simulation use, ineffectiveness, and acceptance. Your answer also contains a serious appeal to authority fallacy in the 4th paragraph. In short this is really more of an extended comment than an answer in its current state. – SoylentGray Jan 26 '18 at 15:53
  • @SoylentGray, I am surprised to read that there is a appeal to authority fallacy in the 4th paragraph. Can you explain that to me. – Etaila Jan 26 '18 at 16:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.