Towards a rigorous, practical, and faithfully optimistic collapsology
[!info] Edit: Out of Date I was not accepted into programs I applied for, and have chosen not to re-apply for 2025. As a result, this is out of date, but I’m still leaving it here!
I’m considering going to graduate school in the fall of 2024. Towards that end, I’m leaving my current guess for a research question here so it can be easily shared with people that I’m talking with.
If you’re here, you’ve probably been sent this link directly as part of a conversation we’re having. If you just stumbled upon it, please feel free to contact me about it - I’m looking for feedback and direction!
This page is also frequently edited as I learn more. If you are returning to this page and it looks different, don’t be surprised!
I am interested in studying system dynamics/systems modeling/complex systems/economic ecology.
My core goal will be to help us collectively best prepare our mechanisms of economic production and distribution for the climate changes we are likely to see in the 21st century.
I think the following related things are probably true:
- We will need increased supply redundancy, but we know little about what kind and to what degree. As the baseline risk of various large scale natural disasters continues to increase and the baseline productivity of organic resources declines, the probability of moments of sustained supply disruption increases, and we will need more supply redundancy to ensure survival. However, redundancy has significant costs. Therefore, it is useful to have a toolkit for evaluating what specific types of redundancy are worth their use of scarce resources.
- Deducing the economic impact of climate models is a hard collective information and planning problem that the market is not producing good information about, and ecological and economic system models are the best way to answer the above given the complexity. Right now, the main path for climate risk to create redundancy is through increased insurance premiums, which then create higher costs for purchasers who can then diversify. However, this mechanism is unlikely to be enough because:
- there is systemic insurance market distortion, and
- small increases in price that reflect the change in long tail risk are unlikely to force purchaser diversification if the risky supply line is still cheaper
- without antecedent simulation, there is no guarantee that actors who diversify will do so in a way that actually can survive the event
- Changes in social and economic policy mostly come from social movements, technological development, or by becoming the “ideas lying around” when a crisis occurs. Therefore, I should strive to produce research that is relevant to ongoing social movement activity, can contribute to discussions about the relative importance of different streams of technological innovation, and is well-positioned to be deployed in times of crisis.
[!info] Note Right now, I am not so interested in working on “the social cost of carbon” or other questions of mitigation, but instead focused on adaptation strategies.
As a result, I would like to understand and develop the best methodology of:
- collecting the economic information required to assess specific (localized or sectoral) economic impacts of climate change
- simulating the second and third order economic impacts of climate change using that information
- simulating the impacts of specific economic interventions (increased redundancy of specific types) on the systems ability to continue to provide essentials when socks occur
Additionally, I am also specifically interested in some of the following other topics related to system dynamics and complex systems research:
- Using large language models for assisted modeling: I think large language models may be able to assist in system model creation, calibration, simulation, and interpreting the results.
- Training general machine learning models on system model outputs: I think accurate and well calibrated system models can generate synthetic data on the systems behavior, which can be used to train decision support models.
- The potential for multi-modal language and system outcome models, jointly trained on synthetic model output data and synthetic language explanations of the system outcome. If possible, such models may be able to produce both good recommendations for system interventions and explanations of those interventions.
These would be detours I would be excited to go on during my research, especially if they were useful stepping stones for answering my primary questions, but they would not be my primary focus.
Thank you for reading and helping!