Fairness Audits: Equal Treatment¶

class explanationspace.audits.EqualTreatment(model, gmodel, algorithm: str = 'auto', masker: bool = False, data_masker: DataFrame | None = None, verbose=False)¶

Bases: BaseEstimator, ClassifierMixin

Given a model, a dataset, and the protected attribute, we want to know if the model violates demographic parity or not and what are the features pushing for it. We do this by computing the shap values of the model, and then train a classifier to distinguish the protected attribute.

Example¶

>>> import pandas as pd
>>> import numpy as np
>>> from sklearn.linear_model import LogisticRegression
>>> from xgboost import XGBRegressor
>>> from fairtools.detector import EqualTreatment
>>> N = 5_000
>>> x1 = np.random.normal(1, 1, size=N)
>>> x2 = np.random.normal(1, 1, size=N)
>>> x34 = np.random.multivariate_normal([1, 1], [[1, 0.5], [0.5, 1]], size=N)
>>> x3 = x34[:, 0]
>>> x4 = x34[:, 1]
>>> # Binarize protected attribute
>>> x4 = np.where(x4 > np.mean(x4), 1, 0)
>>> X = pd.DataFrame([x1, x2, x3, x4]).T
>>> X.columns = ["var%d" % (i + 1) for i in range(X.shape[1])]
>>> y = (x1 + x2 + x3) / 3
>>> y = 1 / (1 + np.exp(-y))
>>> detector = EqualTreatment(model=XGBRegressor(), gmodel=LogisticRegression())
>>> detector.fit(X, y, Z="var4")
>>> detector.get_auc_val()

fit(X, y, Z)¶

Automatically fits the whole pipeline

Parameters¶

Xarray-like of shape (n_samples, n_features): The training input samples.
yarray-like of shape (n_samples,): The target values (class labels in classification, real numbers in regression)
Zarray-like of shape (n_samples,): The protected attribute values (class labels in classification, real numbers in regression)

fit_inspector(X, z)¶

Fits the inspector model to the explanations

Parameters¶

Xarray-like of shape (n_samples, n_features): The training input samples.
zarray-like of shape (n_samples,): The protected attribute values (class labels in classification, real numbers in regression).

fit_pipeline(X, y, z)¶

Fits the model F to X and y

2. Call fit_inspector to fit the Equal Treatment Inspector Careful this method trains F and G on the same data

Parameters¶

Xarray-like of shape (n_samples, n_features): The training input samples.
yarray-like of shape (n_samples,): The target values (class labels in classification, real numbers in regression).
Zarray-like of shape (n_samples,): The protected attribute values (class labels in classification, real numbers in regression).

get_explanations(X)¶: Returns the explanations of the model on the data X. Produces a dataframe with the explanations of the model on the data X.

predict(X)¶: Returns the predictions (ID,OOD) of the detector on the data X.

predict_proba(X)¶: Returns the soft predictions (ID,OOD) of the detector on the data X.

set_fit_request(*, Z: bool | None | str = '$UNCHANGED$') → EqualTreatment¶

Request metadata passed to the fit method.

Note that this method is only relevant if enable_metadata_routing=True (see sklearn.set_config()). Please see User Guide on how the routing mechanism works.

The options for each parameter are:

True: metadata is requested, and passed to fit if provided. The request is ignored if metadata is not provided.
False: metadata is not requested and the meta-estimator will not pass it to fit.
None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.
str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

New in version 1.3.

Note

This method is only relevant if this estimator is used as a sub-estimator of a meta-estimator, e.g. used inside a Pipeline. Otherwise it has no effect.

Parameters¶

Zstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for Z parameter in fit.

Returns¶

selfobject: The updated object.

set_score_request(*, sample_weight: bool | None | str = '$UNCHANGED$') → EqualTreatment¶

Request metadata passed to the score method.

Note that this method is only relevant if enable_metadata_routing=True (see sklearn.set_config()). Please see User Guide on how the routing mechanism works.

The options for each parameter are:

True: metadata is requested, and passed to score if provided. The request is ignored if metadata is not provided.
False: metadata is not requested and the meta-estimator will not pass it to score.
None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.
str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

New in version 1.3.

Note

This method is only relevant if this estimator is used as a sub-estimator of a meta-estimator, e.g. used inside a Pipeline. Otherwise it has no effect.

Parameters¶

sample_weightstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for sample_weight parameter in score.

Returns¶

selfobject: The updated object.