Counterintuitive Behavior of Social Systems (1995) [pdf]
web.mit.edu
web.mit.edu
"Limits to Growth" made several forecasts for humanity that basically predicted mass negative changes, with best-case scenario being that it would settle out with humanity having an average quality of life equal to that of what eastern europe has (or perhaps had 10-20 years ago).
The forecast gets updated every ten years or so, and I've lost track of what the most recent updates said. They generally claim that the metrics are tracking what the model predicted in its worst-case scenario, which would lead to quite a lot of horrible worldwide societal changes in the 2020-2030 range.
The model has generally been ignored and criticized for being too pessimistic and not based on rigorous data.
When they talk about levers for change, one of the most powerful ones they recommend is changing the goal entirely, in this case removing the emphasis on "growth". It's interesting to think about changing the definition of GDP, since our entire definition of recession and depression is based off of that.
It sounds like some of their prescriptions to avoid the worst outcomes are pretty draconian, like limiting food production to help control population growth.
Not perfect, sure. But also better than any other metric I know of.
Depends on what you want to maximize, though. Income equality? Planet health? Life expectancy? Some sort of happiness index? Probability of becoming an interstellar species?
Maximizing one is not the answer. We want to maximize all of them. And in that sense GDP is a great measure since it impacts all of them greatly.
Even sustainability is linked to GDP growth. Green technologies are typically the most expensive, and we need growth so that we can one day afford them across the board.
To me it seems easier said than done as the process of surfacing all assumptions and accurately reflecting reality in your model - at least to a useful level - strikes me as extraordinarily exhaustive and time-consuming. To the point that it may only be useful if you can devote quite a lot of time and money to it.
http://chicago.curbed.com/2015/1/16/10001642/reallife-simcit...
It's not that I don't think they can't be useful, like all models, but the paper greatly simplifies and brushes over some VERY complex problems and issues with them.
Let's focus on some:
1The paper says the underlying rules governing such systems can be easily translates into computer language and understood. I assure you they cannot and as the model gets bigger, like software it gets harder and harder. If your model is so complex it can accurately describe the things about social systems that fool simple human cognitive systems, the question arises how did YOU discover those rules without the model, and how do you debug something that has effects that are far removed and non intuitive.
The decision as to what the foundations of the model are and what is important are often confounded and become political: fundamentally when it comes to the nature of these complex systems, we don't have a consensus reality, so someone has to make a call, and am the effects that stop you being able to understand complex social systems also feed into who usually makes that call.
And finally, as with all predictive future models and forecasting, often conjoined in this case because of the chaotic and nonlinear effects, how do you judge the falsafiability or fit/error of your model and implementation.
Let's say your model says that humans get wiped out in 50 years, or your policy, which everyone in the room loves, looks like a total failure. Then is the model right? Have the assumptions that gone into it been correct? Is the implementation correct?
So what you'll tend to get in practice are iterative design choices that result in models that confirm the assumptions and beliefs of the most powerful people in the room: because how well the model confirms to their beliefs is the judge of the model fitness, and you quickly devolve into a political echo chamber where modelling becomes marketing/self-confirming...