assembly-economics/simulations/gadm02/grounded_readings.py

"""Key readings under the evidence-grounded parameter sets (ASSEMBLY-WP-0008).

Compares GADM-0.4 defaults with the grounded 'central' and 'slow' adoption
variants on: scenario VFAC and ranking, lowest domination, organism lever
order, P3 (wealth tax), L2/G3 (care relief), and attention levels.

Usage: python3 -B simulations/gadm02/grounded_readings.py
"""

from __future__ import annotations

import itertools
import sys
from pathlib import Path

sys.dont_write_bytecode = True

import gadm04 as g  # noqa: E402
import grounding as gr  # noqa: E402
from scenarios import load_all  # noqa: E402

OUT = Path(__file__).resolve().parent / "results"
GRID = {"lineage": ({"L": 0.1, "lin_open": 0.1}, {"L": 0.8, "lin_open": 0.8}),
        "membrane": ({"m": 0.2}, {"m": 0.8}),
        "attention": ({"x_att": 0.9}, {"x_att": 0.2})}


def sim(sc: dict, prm: dict, ini: dict, extra: dict | None = None) -> dict:
    return g.simulate({**sc["params"], **prm, **(extra or {})}, {**sc["init"], **ini})[-1]


def main() -> None:
    OUT.mkdir(exist_ok=True)
    scs = load_all()
    variants = [("GADM-0.4 defaults", {}, {}), ("grounded, central adoption", *gr.variant("central")),
                ("grounded, slow adoption", *gr.variant("slow"))]
    lines = [f"# {g.MODEL_VERSION}: readings under grounded parameters", "",
             "Generated by `simulations/gadm02/grounded_readings.py`. t = 100 years.",
             "Grounded sets: `simulations/gadm02/grounding.py`; justification in",
             "`models/ParameterGrounding.md`.", "",
             "## Scenarios (VFAC / P / A)", "",
             "| Scenario | " + " | ".join(v[0] for v in variants) + " |", "|---|" + "---|" * len(variants)]
    res = {v[0]: {n: sim(sc, v[1], v[2]) for n, sc in scs.items()} for v in variants}
    for n, sc in scs.items():
        lines.append(f"| {sc['title']} | " + " | ".join(
            f"{res[v[0]][n]['VFAC']:.3f} / {res[v[0]][n]['P']:.2f} / {res[v[0]][n]['A']:.2f}" for v in variants) + " |")
    lines += ["", "## Key readings", "",
              "| Reading | " + " | ".join(v[0] for v in variants) + " |", "|---|" + "---|" * len(variants)]
    names = list(scs)
    ref = sorted(names, key=lambda n: -res[variants[0][0]][n]["VFAC"])
    rows = {k: [] for k in ("best", "worst", "rho", "lowP", "levers", "order", "p3", "care", "kO")}
    for label, prm, ini in variants:
        r = res[label]
        order = sorted(names, key=lambda n: -r[n]["VFAC"])
        rank = {n: i for i, n in enumerate(order)}
        rref = {n: i for i, n in enumerate(ref)}
        k = len(names)
        rows["best"].append(scs[order[0]]["title"])
        rows["worst"].append(scs[order[-1]]["title"])
        rows["rho"].append(f"{1 - 6 * sum((rank[n] - rref[n]) ** 2 for n in names) / (k * (k * k - 1)):.2f}")
        rows["lowP"].append(scs[min(names, key=lambda n: r[n]["P"])]["title"])
        out = {}
        for lv in itertools.product((0, 1), repeat=3):
            ex = {}
            for ax, l in zip(GRID, lv):
                ex.update(GRID[ax][l])
            out[lv] = sim(scs["baseline"], prm, ini, ex)["VFAC"]
        eff = [sum((1 if lv[i] else -1) * v for lv, v in out.items()) / 4 for i in range(3)]
        rows["levers"].append(" / ".join(f"{e:+.3f}" for e in eff))
        rows["order"].append("yes" if eff[0] > eff[1] > eff[2] > 0 else "no")
        mem = scs["generative_concentrated_lineages"]
        dp = sim(mem, prm, ini, {"wtax": 1.0})["P"] - sim(mem, prm, ini, {"wtax": 0.0})["P"]
        rows["p3"].append(f"{dp:+.3f} ({'holds' if -dp < 0.06 else 'fails'})")
        b0 = sim(scs["baseline"], prm, ini, {"cp": 0.0})
        b1 = sim(scs["baseline"], prm, ini, {"cp": 0.30})
        b1c = sim(scs["baseline"], prm, ini, {"cp": 0.30, "care_relief": 0.0})
        gain = b1["Gs"] - b0["Gs"]
        rows["care"].append(f"Gs +{gain:.3f}, care share {((b1['Gs'] - b1c['Gs']) / gain):.0%}" if gain > 0 else "—")
        bl = g.simulate({**scs["baseline"]["params"], **prm}, {**scs["baseline"]["init"], **ini})
        rows["kO"].append(f"{bl[10]['Ko']:.2f} / {bl[25]['Ko']:.2f} / {bl[-1]['Ko']:.2f}")
    labels = {"best": "best scenario", "worst": "worst scenario", "rho": "ranking ρ vs. defaults",
              "lowP": "lowest domination", "levers": "lineage / membrane / attention effect",
              "order": "lever order lineage > membrane > attention", "p3": "P3: wealth tax ΔP",
              "care": "G3: senior agency gain at cp 0.3", "kO": "baseline K_Ω at t 10 / 25 / 100"}
    for k, lab in labels.items():
        lines.append(f"| {lab} | " + " | ".join(rows[k]) + " |")
    (OUT / "grounded_readings.md").write_text("\n".join(lines) + "\n")
    print("\n".join(lines))


if __name__ == "__main__":
    main()