Clarification on identified estimands when estimate direct and indirect effects

[StCross]

Hello I am using dowhy for my Thesis, however not quite sure about the exact meaning of Estimand expression when estimate nde and nie. Much appriciate if an explanation can be provided.
My SCM and code are as follow,where A is the treatement and I is the outcome, and all A to H are parent nodes for I(actually A to H are features and I is label, I am using dowhy to estimate feature's local nde to label )

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

import dowhy
import dowhy.datasets
from dowhy import CausalModel

df = pd.DataFrame(columns={'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I'})

model=CausalModel(
        data = df,
        treatment='A',
        outcome='I',
        graph="digraph {A -> C;B -> C;D -> A;D -> B; D -> H; E -> A; E -> B; E -> H; F -> A; G -> B;A -> I;B -> I;C -> I;D -> I; E -> I; F -> I; G -> I; H -> I}"
        )
model.view_model()

from IPython.display import Image, display
display(Image(filename="causal_model.png"))

identified_estimand = model.identify_effect(estimand_type='nonparametric-nde', proceed_when_unidentifiable=True)
print(identified_estimand)
identified_estimand = model.identify_effect(estimand_type='nonparametric-nie', proceed_when_unidentifiable=True)
print(identified_estimand)

and output is:

Estimand type: nonparametric-nde

### Estimand : 1
Estimand name: mediation
Estimand expression:
Expectation(Derivative(I|C, [A]))
Estimand assumption 1, Mediation: C intercepts (blocks) all directed paths from A to I except the path {A}→{I}.
Estimand assumption 2, First-stage-unconfoundedness: If U→{A} and U→{C} then P(C|A,U) = P(C|A)
Estimand assumption 3, Second-stage-unconfoundedness: If U→{C} and U→I then P(I|C, A, U) = P(I|C, A)

Estimand type: nonparametric-nie

### Estimand : 1
Estimand name: mediation
Estimand expression:
Expectation(Derivative(I, [C])*Derivative([C], [A]))
Estimand assumption 1, Mediation: C intercepts (blocks) all directed paths from A to I except the path {A}→{I}.
Estimand assumption 2, First-stage-unconfoundedness: If U→{A} and U→{C} then P(C|A,U) = P(C|A)
Estimand assumption 3, Second-stage-unconfoundedness: If U→{C} and U→I then P(I|C, A, U) = P(I|C, A)

[amit-sharma]

the nde and nie represent natural direct effect and natural indirect effect. you can refer to this paper for definitions: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8190553/

[amit-sharma]

Converting this to a discussion.