We built a knowledge tree of human flourishing from 11 deep research syntheses — each drawing on dozens of meta-analyses and systematic reviews spanning hundreds of thousands of participants — covering sleep, inflammation, cognition, trauma, nutrition, exercise, relationships, psychological flexibility, contemplative practice, and purpose.
When we analyzed the evidence computationally — extracting effect sizes from the research, converting them to comparable units (Cohen's d equivalents), and scoring each dimension by research-grounded importance — a pattern emerged. But it was more nuanced than we initially reported.
What we found — and what we initially overclaimed
Our earlier analysis stated that deficiency-fixing produces "3 to 30 times more human flourishing than optimizing what's working." That framing was based on comparing the most dramatic deficiency effects (iodine deficiency: 13.5 IQ points; financial scarcity: 13 IQ-equivalent points) against modest optimization effects (creatine: SMD 0.34, non-significant).
Those individual comparisons are real. But when we computed research-grounded importance scores across all 198 dimensions — using effect sizes extracted and validated by two independent models, with deterministic conversions and traceability checks — the overall picture was more complex:
| Hierarchy Level | Dimensions with effect data | Mean |d| |
|---|---|---|
| Foundational | 24 | 0.79 |
| Secondary | 16 | 0.77 |
| Refinement | 10 | 0.84 |
The mean effect sizes across levels are essentially equal. There is no clean 3-30x ratio between hierarchy levels. Refinement dimensions actually have the highest mean — largely driven by strong meta-analytic evidence for therapeutic interventions (ACT treatment: d=0.83, psychedelic-assisted therapy: d=0.91).
The honest finding: the asymmetry is real, but it's not a property of hierarchy levels in aggregate. It's a property of specific threshold-type dimensions at the foundational level, where falling below a biological minimum produces catastrophic effects that no amount of higher-level optimization can compensate for.
Where the asymmetry IS dramatic
Certain foundational deficiencies produce effects so large they exist in a different category from any optimization intervention:
| Deficiency | Effect | Source |
|---|---|---|
| Iodine deficiency (gestational) | 13.5 IQ points population-level deficit | Bleichrodt & Born meta-analysis, 18 studies |
| Financial scarcity | 13 IQ-equivalent points cognitive deficit | Mani, Mullainathan, Shafir & Zhao, Princeton/Harvard |
| Sleep deprivation (<7h) | ηp² = 0.41 on selective attention | 24h deprivation studies, multiple replications |
| Iron deficiency | SMD 0.79 intelligence in anemic children | Gutema et al. 2023, meta-analysis of 13 RCTs |
| Anti-inflammatory dietary pattern | d = 1.85 (highest research-grounded importance in the tree) | Meta-analyses of dietary interventions |
| ACE burden (4+ adverse childhood experiences) | d = 1.07 | Meta-analysis of ACE impact studies |
These specific deficiencies genuinely dwarf any optimization-only intervention. The difference between having adequate iron and being anemic (d=0.79) is larger than the difference between moderate and elite cognitive training in healthy adults. The effect of dietary pattern on inflammation (d=1.85) is the single largest research-grounded dimension in the entire tree.
Where optimization is stronger than expected
The data also revealed that some higher-level interventions have substantial effect sizes — larger than we initially credited:
| Intervention (secondary or refinement) | Effect | Source |
|---|---|---|
| Therapeutic access (relationship quality) | d = 2.09 | Spengler 2022 meta-analysis |
| Purpose interventions | d = 0.62 overall, 1.57 mindfulness subgroup | Manco & Hamby 2021, 33 RCTs |
| Contemplative practice specificity | d = 1.16 | LKM meta-analysis, 24 studies |
| Body-based trauma intervention | d = 1.07 | van der Kolk 2014 RCT + meta-analyses |
These are not small effects. They complicate any simple narrative about optimization being inherently low-return.
The real asymmetry: threshold curves vs. diminishing returns
Where the asymmetry is structural is in the shape of utility curves, not in average effect sizes by level:
When we analyzed utility curve shapes across dimensions, we found that 36% of foundational dimensions have threshold-type curves — below a certain level, you're measurably impaired; above it, more doesn't help. Only 2% of secondary dimensions and 0% of refinement dimensions are threshold-based.
This means the biology itself is asymmetric at the threshold. Your brain either has enough glucose or it doesn't (threshold at 2.5 mmol/L, Cohen's d > 0.8 for executive function below this). You're either adequately hydrated or you're not (threshold at 2% body mass loss). You either have sufficient iodine during fetal neurodevelopment or the damage is irreversible.
Above these thresholds, the curves shift to diminishing returns. The foundational dimensions stop producing large effects — but secondary and refinement dimensions start producing theirs. The gating cascade (Insight 06) explains why: foundational adequacy opens the gate that lets higher-level interventions work.
The asymmetry is not "foundations always have bigger effects than higher levels." It's: specific foundational deficiencies produce catastrophic effects that operate in a different category from optimization — and fixing them unlocks the capacity to benefit from everything else. The data supports this because of the gating structure, not because of a blanket difference in effect sizes by level.
What this means in practice
The practical implication survives the correction. If you're building systems to improve human lives — whether that's a health app, a recommendation engine, a coaching practice, or a public health policy — the first question should still be "what's below threshold?"
Not because foundational dimensions always have bigger effect sizes. But because:
- Threshold-type failures can't be compensated for. No supplement, therapy, or purpose practice can substitute for adequate sleep, nutrition, or physical safety. The biology doesn't negotiate.
- Foundational failures gate everything above. The computed scores show this clearly: purpose cultivation scores 0.239 despite having d=1.57 for interventions — because weak foundational evidence gates the entire framework down.
- The cascade compounds. Sleep loss + inflammation + sedentary behavior don't add linearly. They interact synergistically, each making the others harder to address.
The most impactful thing you can do for someone below foundational thresholds is the least glamorous: ensure they sleep enough, eat adequately, move their bodies, feel physically safe, and aren't crushed by financial precarity. Everything else builds on that foundation.
But for someone above foundational thresholds, the evidence says higher-level interventions — psychological flexibility, therapeutic relationships, contemplative practice, purpose cultivation — produce effects just as large as foundational ones. The gate is open. The returns are real.
How we arrived at this — including the correction
This insight emerged from analyzing 19 evidence-grounded frameworks containing 198 dimensions and 91 interactions. The original version of this article was based on narrative synthesis — reading the research and identifying patterns qualitatively.
The revision was driven by a computational pipeline that extracted effect sizes from all 198 basis texts using two independent models (a local Qwen 3.5 4B and Anthropic's Haiku 4.5, achieving 93% agreement), converted them to Cohen's d equivalents using deterministic formulas, and scored each dimension by research-grounded importance. When we compared mean effect sizes across hierarchy levels, the clean 3-30x ratio we had claimed dissolved into a more nuanced picture.
We're publishing the correction because the commitment is to evidence, not to our earlier narrative. The core insight — that foundational deficiencies produce threshold effects that must be addressed first — is strengthened, not weakened, by being precise about what the data actually shows.