It seems very challenging to construct an intervention that takes both the macro and micro into account. Although I'm not saying they shouldn't attempt it!
As a transition to that world, I wonder if many agencies would find it easier to have a portfolio of macro and micro focused interventions. Is an implication of this essay that the larger sources of funding (e.g. NIH, etc) have over-indexed on macro-level interventions and should run more experiments to understand how to improve things at a micro scale?
the implication is to study the system as a whole - the parts (micro) and the wholes (macro). identify leverage points and study 2nd-3rd-4th order effects at all scales. design and deploy interventions agnostic of “larger source of funding can only fund macro level intervention”
something about causal emergence 2.0 here too but those thoughts are still half baked
Ahh gotcha! What would be an example of this? Has anyone successfully done anything like this before, or is this sufficiently novel that it hasn't been attempted before?
pieces of this have been done before, but not the full thing as a deliberate, standing method. novelty here is less about single components and more about the integration and the feedback loop across scales.
there are a few important partial precedents:
> Development economics / RCTs (e.g. work by Esther Duflo and Abhijit Banerjee):
This is probably the closest micro → macro pipeline. Small-scale experiments generate insights that later inform large policy decisions. What’s missing is sustained bidirectional feedback — once something scales, the system usually isn’t re-instrumented to study how macro changes reshape micro behavior over time.
> Aviation safety & systems engineering:
These fields explicitly study micro-level failures (human factors, near misses) and redesign macro systems (training, incentives, protocols). This comes close to what I’m describing, but it works because the system is tightly bounded and operationally constrained. Social and scientific funding systems are far more open-ended and adaptive.
> Public health & behavioral policy:
Some interventions combine macro tools (taxes, regulations) with micro tools (nudges, education). But they’re typically treated as separate levers, not as one coupled system whose causal dynamics are studied across multiple orders of effect.
> Complex systems science / agent-based modeling:
The theory is there — emergence, multi-scale causality, feedback loops. But this work mostly lives in simulations or post hoc analysis, not in real-world institutional design where interventions are continuously updated based on cross-scale effects.
sooo this has definitely been attempted in fragments, in specific domains, and often one-directionally. what hasn’t really been done is building institutional infrastructure that treats micro and macro as one coupled system — studied, intervened on, and updated continuously with explicit attention to second-, third-, and higher-order effects.
that’s the gap we’re pointing at. it isn’t “no one’s ever thought of this,” but rather that our institutions are not designed to operate this way, especially in open-ended domains like science funding or knowledge production.
novelty here is not conceptual novelty but methodological and institutional novelty
What would need to be true to build a minimum viable institution (i.e. I guess a project?) that does this at multiple levels? And what specific problems/domains would be well suited for experimentation within this space?
> “Now every neuroscientist puts the word ‘Alzheimer’s’ in their grant proposal no matter what they are actually studying.”
This happens with basically every federal targeted research initiative. New money comes into the politically favored area, and even if that money isn't explicitly taken from other areas, there's a constant downward pressure on general science funding. Therefore, scientist have to rephrase their existing research in terms of the new thing, or their labs will sink. That's why these initiatives are never successful. So far, no one in power has learned this lesson. My own prescription to fix this: targeted initiatives should come with guarantees that each new targeted dollar is matched by an unrestricted dollar.
really proud of this one!
Great observation! I hope someone is listening.
Excellent work, extremely helpful framework.
It seems very challenging to construct an intervention that takes both the macro and micro into account. Although I'm not saying they shouldn't attempt it!
As a transition to that world, I wonder if many agencies would find it easier to have a portfolio of macro and micro focused interventions. Is an implication of this essay that the larger sources of funding (e.g. NIH, etc) have over-indexed on macro-level interventions and should run more experiments to understand how to improve things at a micro scale?
the implication is to study the system as a whole - the parts (micro) and the wholes (macro). identify leverage points and study 2nd-3rd-4th order effects at all scales. design and deploy interventions agnostic of “larger source of funding can only fund macro level intervention”
something about causal emergence 2.0 here too but those thoughts are still half baked
Ahh gotcha! What would be an example of this? Has anyone successfully done anything like this before, or is this sufficiently novel that it hasn't been attempted before?
pieces of this have been done before, but not the full thing as a deliberate, standing method. novelty here is less about single components and more about the integration and the feedback loop across scales.
there are a few important partial precedents:
> Development economics / RCTs (e.g. work by Esther Duflo and Abhijit Banerjee):
This is probably the closest micro → macro pipeline. Small-scale experiments generate insights that later inform large policy decisions. What’s missing is sustained bidirectional feedback — once something scales, the system usually isn’t re-instrumented to study how macro changes reshape micro behavior over time.
> Aviation safety & systems engineering:
These fields explicitly study micro-level failures (human factors, near misses) and redesign macro systems (training, incentives, protocols). This comes close to what I’m describing, but it works because the system is tightly bounded and operationally constrained. Social and scientific funding systems are far more open-ended and adaptive.
> Public health & behavioral policy:
Some interventions combine macro tools (taxes, regulations) with micro tools (nudges, education). But they’re typically treated as separate levers, not as one coupled system whose causal dynamics are studied across multiple orders of effect.
> Complex systems science / agent-based modeling:
The theory is there — emergence, multi-scale causality, feedback loops. But this work mostly lives in simulations or post hoc analysis, not in real-world institutional design where interventions are continuously updated based on cross-scale effects.
sooo this has definitely been attempted in fragments, in specific domains, and often one-directionally. what hasn’t really been done is building institutional infrastructure that treats micro and macro as one coupled system — studied, intervened on, and updated continuously with explicit attention to second-, third-, and higher-order effects.
that’s the gap we’re pointing at. it isn’t “no one’s ever thought of this,” but rather that our institutions are not designed to operate this way, especially in open-ended domains like science funding or knowledge production.
novelty here is not conceptual novelty but methodological and institutional novelty
The stuff you're describing here also reminds me of one of Kegan's books - An Everyone Culture. IIUC he and his team are also based in NYC.
ohhh gotcha!!! This is all very cool.
What would need to be true to build a minimum viable institution (i.e. I guess a project?) that does this at multiple levels? And what specific problems/domains would be well suited for experimentation within this space?
i will text you the idea doc! not wanting to share it publicly yet
This was a fun read and I found it insightful!
> “Now every neuroscientist puts the word ‘Alzheimer’s’ in their grant proposal no matter what they are actually studying.”
This happens with basically every federal targeted research initiative. New money comes into the politically favored area, and even if that money isn't explicitly taken from other areas, there's a constant downward pressure on general science funding. Therefore, scientist have to rephrase their existing research in terms of the new thing, or their labs will sink. That's why these initiatives are never successful. So far, no one in power has learned this lesson. My own prescription to fix this: targeted initiatives should come with guarantees that each new targeted dollar is matched by an unrestricted dollar.