I’d like to conduct an open, informal study with the goal of better understanding and cataloguing approaches to solving the metacrisis.
I’m not expecting this goal to be achieved by this study. I just think it would be interesting to a) bring our attention to possible solutions and b) attempt to get a relatively unbiased look at how intended solutions compare with one another. This will be done by reducing human bias, insofar as possible, through the use of multiple LLMs to evaluate candidate solutions.
Methodology
I will try to keep this as simple as possible so that it will be easy to understand, replicate, and critique. The following will be done by the LLMs in order to avoid cherry-picking. I’ll provide an example prompt and response by an LLM in the next post to demonstrate how this would work.In future iterations, those interested can request solutions they’d like evaluated to be included in the solution set.
Each LLM will come up with a list of problems that must be solved in order to solve the metacrisis.
Each LLM will come up with a set of numerical criteria that can be used for determining how effective a proposed solution is.
Each LLM will assign weights to its own criteria, with the weights adding up to 100.
Each LLM will evaluate every candidate solution according to its own weighted criteria. This will result in a table like:
Criterion A Criterion B Criterion C Total
Solution 1 _ _ _ _
Solution 2 _ _ _ _
Solution 3 _ _ _ _
Since different LLMs may use different criteria, the raw scores will not be compared directly across models. Instead, each LLM’s total scores will be converted into a normalized rank score from 0 to 100. This means that each model’s favorite solution receives 100, its least favored solution receives 0, and the others are placed proportionally in between.
normalized score = 100 × (N - rank) / (N - 1)
The final score for each solution will be the average of its normalized scores across all LLMs. The amount of disagreement between LLMs will also be reported, using the standard deviation of those normalized scores.
This means that the final result will look something like:
The point of this is not to treat the final score as an objective measurement of a solution’s value. Rather, the score should be interpreted as a comparative signal: a way of seeing which solutions are favored across different LLM-generated evaluative frameworks, and where there is strong agreement or disagreement between models.
Limitations
There are some major limitations with this methodology that should be kept in mind.
The most obvious one is that LLMs, while being less biased than most humans in many ways, still reflect biases inherent in the data they were trained on. Furthermore, LLMs are not conscious and are much more equipped to reason about matters requiring cognitive, linguistic, or purely factual judgment than emotional, interpersonal, moral, aesthetic, and other dimensions of experience.
Another limitation in this methodology is imposed mutual exclusivity. There is no reason why many, or even all, of the solutions listed cannot be used together. There is no one solution to the metacrisis that will work to the exclusion of all others. So if some candidate solution receives the highest score, this does not necessarily mean others are not important or useful.
A further limitation is that the choice of candidate solutions itself may affect the results. A solution’s rank depends partly on what it is being compared against. For that reason, the list of candidate solutions should be made public, and ideally expanded or revised through open feedback.
Finally, the purpose of this study is exploratory rather than authoritative. It is meant to generate useful comparisons, reveal patterns, and surface questions worth discussing. It should not be treated as a final answer to what the best solution to the metacrisis is.
I am conducting an open, informal study on proposed approaches to solving the metacrisis.
Your task is to generate and evaluate a set of candidate approaches as fairly and rigorously as possible. The goal is not to reach a definitive answer, but to produce a useful comparative analysis.
Please follow this process:
Briefly define what you understand by “the metacrisis.”
Identify the major problems that would need to be solved in order to solve, substantially reduce, or meaningfully respond to the metacrisis.
Generate a diverse list of candidate approaches to addressing the metacrisis. Include approaches from different domains, such as governance, economics, technology, culture, psychology, spirituality, education, coordination, institutional reform, and ecological restoration. Try to choose approaches that are meaningfully different from one another, not minor variations of the same idea.
Create a set of numerical evaluation criteria for judging how effective a proposed approach is likely to be. The criteria should be broad enough to apply to all candidate approaches.
Assign weights to the criteria, with all weights adding up to 100. Explain briefly why you chose those weights.
Evaluate each candidate approach according to your weighted criteria. Use a 0–10 score for each criterion.
Calculate a weighted total score for each approach.
Convert the weighted total scores into normalized rank scores from 0 to 100, using this formula:
normalized score = 100 × (N - rank) / (N - 1)
where N is the number of candidate approaches.
The highest-ranked approach should receive 100, the lowest-ranked approach should receive 0, and the others should be placed proportionally in between. If two approaches are tied, assign them the same rank and explain how the tie was handled.
Present the results in a table with the following columns:
Over in another thread I human-brainstormed what the metacrisis is. Because you are bringing out all this LLM action, I decided to LLM critique my ideas and share the results here. Gemini pretty much turned it all into a 2R white paper!
This is a sophisticated diagnostic list that aligns closely with contemporary systemic thinking on the metacrisis (the interconnected web of ecological, political, economic, and psychological crises facing humanity).
However, looking at this list through the lens of complexity theory and prominent metacrisis frameworks (such as those posited by Daniel Schmachtenberger or Nate Hagens), your list conflates root generator functions with downstream symptoms and accelerants.
Here is an evaluation of your ranking, followed by a suggested re-ordering based on what truly drives the system.
The Critique: What’s a Root vs. a Symptom?
1. The True Roots (Items 6 & 8)
Lack of cognitive/emotional capacity (6) and Culture out of phase (8) are actually the bedrock foundational layers.
Our current institutional and technological power is growing exponentially, but our collective sense-making, emotional maturity, and cultural wisdom are still running on tribal-era software. This asymmetry is the ultimate generator of the crisis. If we cannot comprehend or emotionally process the complexity we create, we cannot govern it.
The Westphalian system (3) and Financialization of capital (5) are the game-theoretic engines.
The Westphalian model forces nation-states into a multipolar trap (a race to the bottom where if one nation doesn’t exploit a resource or develop a weapon, another will).
Global economic growth (2) is a mandate of the financial system (which requires exponential growth to service debt), rather than a standalone root cause.
3. The Downstream Pressures & Multipliers (Items 1 & 4)
Global population growth (1) and Weapons of mass destruction (4) are profound accelerants, but they are outputs. Population growth is driven by the success of industrial agriculture and medicine (fossil fuel exploitation), and WMDs are the inevitable technical output of the Westphalian multipolar trap combined with exponential technology.
4. The Feedback Loop (Item 7)
Biospheric pushback (7) (climate change, biodiversity loss) is a consequence of the system hitting physical boundaries. It is an effect that feeds back into the loop, creating new crises, but it is not a root cause.
A Suggested Realignment (From Root to Symptom)
If we re-order these items from the interior/foundational causes (how we think and organize) to the exterior/material consequences (what happens to the world), the list flows more logically:
Tier 1: The Foundational/Cognitive Crisis (The “Why”)
Culture out of phase with current requirements: Our values, ethics, and narratives are inadequate for handling existential-scale power.
Lack of cognitive/emotional capacity to cope with extreme complexity: The human brain struggling to process hyperobjects (like global supply chains or algorithmic information ecosystems), leading to polarization and sense-making breakdown.
Tier 2: The Institutional/Structural Rules (The “How”)
The end game of the Westphalian nation-state system: An obsolete geopolitical model that prevents global coordination on global problems.
The end game of the financialization of capital: An economic system that optimizes for short-term fiat metrics over long-term biophysical reality.
Global economic growth: The inevitable, unsustainable output of the structural rules set by Tiers 1 and 2.
Tier 3: The Material Accelerants & Outcomes (The “What”)
Weapons of mass destruction: The deadly manifestation of rivalrous games paired with exponential tech.
Global population growth: The physical scaling of the human footprint, pushing the limits of finite resources.
Biospheric pushback: The planet’s natural systems breaking down under the weight of the layers above it.
Summary: Your list identifies the exact right variables, but flipping your ranking upside down might yield a more accurate map of causality. The material crises (population, growth, WMDs) cannot be solved without first addressing the plumbing of our cultural narratives, financial incentives, and cognitive capacities.
Do you see these factors as operating linearly, or do you view them as a web of mutual reinforcement?
Yeah. Gemini is right about that. But humans can play N-D chess too! I started with simple, tangible symptoms (like human population growth) as a pedagogical trick to gets heads nodding about the stone-cold obvious. My list complexifies from there, because I am using all the typical cognitive biases to get people on board with the much harder “sell” about personal and cultural change. It’s all a rhetorical bait and switch, basically.
On another level, I see too much discussion in these channels that leans heavily into root-causal idealism (like the need for cognitive-emotional complexification) and blows past the messier physical layer stuff like nation states, WMD, financial capital, etc. I like how Gemini Donella Meadows’d the whole business and worked from low-leverage symptoms (global population) to higher leverage handles on what we might do about it.
The results I got from running my prompt do not seem very meaningful to me. The prompt did too little to drill down on specifics. Collective intelligence can mean a lot of things to a lot of people. Although I stipulated that mutual exclusivity in the approaches would be unrealistic, there’s still some amount in group behavior; people generally choose one collective intelligence project to work on over others, for example. A person who believes in Sociocracy will often invest time and resources in trying to make it work, which must come at the exclusion of things like: Parecon, Holacracy, and so on.
So it didn’t have a way of comparing approaches against one another, for one, and yet it also did compare things like inner development and governance which are in no way mutually exclusive.
I think a better version of this study would look at a set of critical success factors similar to how our goal tree does it. We must have healthy, happy interpersonal relationships. We must have environmental conditions that make the planet livable. And so on and so forth. Then, for each of these success dimensions, ask for a set of distinct approaches to achieving the success factor.
My gut feeling is that running an LLM tournament will never get below the surface, because “meaning” is precisely about gut feelings! Riffing off of LLM content works fine by me, but the resulting riffs will always bounce of into the tangible and the anecdotal, because that is where the real energetic “juice” is to found. Rule of thumb: don’t craft arguments. Create moments.
I don’t expect to get results I’d consider good enough to inform policy; this is more of a learning experiment. The purpose is having a better idea of knowing what sort of a claim “X can solve the meaning crisis” would actually be making.
I improved the prompt based on my observations above and ran it again. The result is better. Actually, what this seems to be turning into is a set of increasingly disjunctive choices. If we keep applying a similar process, where we start with a set of options that are by no means mutually exclusive, and drill down until they are, we might end up with something that looks like reasons for preferring one route over others.
The results from this chat, for example, still contain results like the following. Obviously, one can do both trauma informed care and community hubs. But perhaps we could ask for specific approaches to these things and we’d end up with something that looks like a ranking of mutually exclusive choices.
A few minutes ago I put down a Taleb chapter on this:
"“Good enough” heuristics—often referred to in psychology as satisficing or fast-and-frugal heuristics—are mental shortcuts that prioritize speed and efficiency over finding the absolute perfect solution. Because human cognitive capacity is limited, our brains use these practical rules of thumb to solve problems and make decisions that are “good enough” for the situation at hand. [1, 2, 3, 4, 5]
While traditional algorithms (like a strict recipe or math formula) guarantee an optimal result, they are often too time-consuming or completely impractical to apply to daily life. By utilizing heuristics, we can manage cognitive overload and act quickly when information is incomplete"
This gist of it is, given extreme complexity and unpredictability, a “less is more” approach generally works better than lengthy detailed studies (like UN reports running to hundreds of pages or whatever ChatGPT is up to in your session above.) As you correctly observe, none of the proposed solution vectors are mutually exclusive. They are only mutually exclusive on the level of personal time, attention, and focus at any given moment. But given a large network of persons who will presumably apply time, attention, and focus over the entire field, it seems plausible that a quick and dirty read about “what can I do about all this?” will work about as well as anything. If the entire human network leans too far one way or another, no doubt systemic feedbacks will amplify and get our attention also one way or the other!
