You dug far down the memory hole to bring up Jay!!!! I lived that era.
The Population Bomb!! Club of Rome!!! Limits to Growth!!!
I attended a NASA talk that said India was one bad monsoon from mass starvation. Predictions are hard, esp about the future. Now we have Ai run up the flag pole to see who salutes.
I am too damned old to live to see how we survive. Hopefully late this week we will get a peek.
I see it's on your syllabus, but I'd like to take the opportunity to signal-boost Kevin Baker's excellent (2019) essay "Model metropolis," where I first learned about Forrester and his thinking: https://logicmag.io/play/model-metropolis/
I think there is still value in simulation, and while I disagree with Forrester on his politics, his work on Industrila Dynamics is pretty useful. Problem is more that even if you make cool simulations, the people equipped to make them useful are usually not the ones in the decision seat, and that the mental model part of the situation you talk about is very easily hidden behind the cool moving numbers and animations.
I thought systems thinking for complex problem solving (meaning most real world problem solving) had been killed several times over . Do we need to kill it again? Dave Snowden's Cynefin framework and David Chapman's meta-rational stance are the best approaches I am yet aware of.
‘The piecemeal engineer knows, like Socrates, how little he knows. He knows that we can learn only from our mistakes. Accordingly, he will make his way, step by step, carefully comparing the results expected with the results achieved, and always on the look-out for the unavoidable unwanted consequences…’ Karl Popper’s "Piecemeal Social Engineering” (1944)
I do want to support Forrester's initial claims, "Forrester argued that policymakers made all decisions with models, but his computer models were more explicit and concrete than people’s mental models. Computer models laid all of the assumptions on the table and let stakeholders adjust the parameters until they conformed to reality." Avoiding the simulations wouldn't have avoided the issue of bending results to fit politics. And dark things can lurk if hidden inside people's unpublished mental models (I wrote here: https://www.jmmcd.net/2021/11/25/lets-automate-existing-inequalities.html).
I think this is a very interesting point that is especially relevant given the vast amounts of research going into formulating world models.
However, I question whether such models constructed in 1971 or video games made with the intention to entertain as opposed to make decisions are still valid in comparison to what can be generated today or models in years to come. The still exponential increase in high-fidelity data, computational abilities which are orders of magnitude more powerful than that of 50 years prior, and the ability to integrate inductive biases of human expertise into model architectures I think change the argument.
I don’t deny that skepticism around “how good is good enough to make real world decisions based on computer models” is the right question and should always be asked. However, is it not reasonable to think that perhaps his assertion was unreasonable then, potentially even unreasonable now (in certain domains), but certainly reasonable going into the future? Especially if such models can be made transparent and interpretable as to the assumptions and parameterizations used.
You dug far down the memory hole to bring up Jay!!!! I lived that era.
The Population Bomb!! Club of Rome!!! Limits to Growth!!!
I attended a NASA talk that said India was one bad monsoon from mass starvation. Predictions are hard, esp about the future. Now we have Ai run up the flag pole to see who salutes.
I am too damned old to live to see how we survive. Hopefully late this week we will get a peek.
I see it's on your syllabus, but I'd like to take the opportunity to signal-boost Kevin Baker's excellent (2019) essay "Model metropolis," where I first learned about Forrester and his thinking: https://logicmag.io/play/model-metropolis/
I think there is still value in simulation, and while I disagree with Forrester on his politics, his work on Industrila Dynamics is pretty useful. Problem is more that even if you make cool simulations, the people equipped to make them useful are usually not the ones in the decision seat, and that the mental model part of the situation you talk about is very easily hidden behind the cool moving numbers and animations.
I thought systems thinking for complex problem solving (meaning most real world problem solving) had been killed several times over . Do we need to kill it again? Dave Snowden's Cynefin framework and David Chapman's meta-rational stance are the best approaches I am yet aware of.
‘The piecemeal engineer knows, like Socrates, how little he knows. He knows that we can learn only from our mistakes. Accordingly, he will make his way, step by step, carefully comparing the results expected with the results achieved, and always on the look-out for the unavoidable unwanted consequences…’ Karl Popper’s "Piecemeal Social Engineering” (1944)
I do want to support Forrester's initial claims, "Forrester argued that policymakers made all decisions with models, but his computer models were more explicit and concrete than people’s mental models. Computer models laid all of the assumptions on the table and let stakeholders adjust the parameters until they conformed to reality." Avoiding the simulations wouldn't have avoided the issue of bending results to fit politics. And dark things can lurk if hidden inside people's unpublished mental models (I wrote here: https://www.jmmcd.net/2021/11/25/lets-automate-existing-inequalities.html).
I think this is a very interesting point that is especially relevant given the vast amounts of research going into formulating world models.
However, I question whether such models constructed in 1971 or video games made with the intention to entertain as opposed to make decisions are still valid in comparison to what can be generated today or models in years to come. The still exponential increase in high-fidelity data, computational abilities which are orders of magnitude more powerful than that of 50 years prior, and the ability to integrate inductive biases of human expertise into model architectures I think change the argument.
I don’t deny that skepticism around “how good is good enough to make real world decisions based on computer models” is the right question and should always be asked. However, is it not reasonable to think that perhaps his assertion was unreasonable then, potentially even unreasonable now (in certain domains), but certainly reasonable going into the future? Especially if such models can be made transparent and interpretable as to the assumptions and parameterizations used.