Fairness measures (or metrics) allow us to assess and audit for possible biases in a trained model. There are several types of metrics that are widely used in order to assess a model’s fairness. They can be coarsely classified into three groups:
Statistical Group Fairness Metrics: Given a set of predictions from our model, we assess for differences in one or multiple metrics across groups.
Individual Fairness: Basically requires that similar people are treated similar independent of the protected attribute. This is more of a philosophical concept and concrete implementations of this fairness notion are not immediately clear.
Causal Fairness Notions: An important realization in the context of Fairness is, that whether a process is fair is often subject to the underlying causal directed acyclic graph (DAG). Knowledge of the DAG allows for causally assessing reasons for (un-)fairness. Since DAG’s are often hard to construct for a given dataset, we currently do not provide any causal fairness metrics.
A simple way to assess the fairness of a model is to find a definition of fairness that is relevant to a problem at hand. We might for example define a model to be fair if the chance to be accepted for a job given you are qualified is independent of a protected attribute e.g. gender
. This can e.g. be measured using the true positive rate
(TPR): in mlr3
this metric is called "classif.tpr"
. In this case we measure discrepancies between groups by computing differences (-)
but we could also compute quotients. In practice, we often compute absolute differences.
\[ \Delta_{TPR} = TPR_{Group 1} - TPR_{Group 2} \]
We will use metrics like the one defined above throughout the remainder of this vignette. Some predefined measures are readily available in mlr3fairness
, but we will also showcase how custom measures can be constructed below.
In general, fairness measures have a fairness.
prefix followed by the metric that is compared across groups. We will thus e.g. call the difference in accuracies across groups fairness.acc
. A full list can be found below.
key | description |
---|---|
fairness.acc | Absolute differences in accuracy across groups |
fairness.mse | Absolute differences in mean squared error across groups |
fairness.fnr | Absolute differences in false negative rates across groups |
fairness.fpr | Absolute differences in false positive rates across groups |
fairness.tnr | Absolute differences in true negative rates across groups |
fairness.tpr | Absolute differences in true positive rates across groups |
fairness.npv | Absolute differences in negative predictive values across groups |
fairness.ppv | Absolute differences in positive predictive values across groups |
fairness.fomr | Absolute differences in false omission rates across groups |
fairness.fp | Absolute differences in false positives across groups |
fairness.tp | Absolute differences in true positives across groups |
fairness.tn | Absolute differences in true negatives across groups |
fairness.fn | Absolute differences in false negatives across groups |
fairness.cv | Difference in positive class prediction, also known as Calders-Wevers gap or demographic parity |
fairness.eod | Equalized Odds: Sum of absolute differences between true positive and false positive rates across groups |
fairness.pp | Predictive Parity: Sum of absolute differences between ppv and npv across groups |
fairness.acc_eod=.05 | Accuracy under equalized odds < 0.05 constraint |
fairness.acc_ppv=.05 | Accuracy under ppv difference < 0.05 constraint |
This vignette assumes that you are familiar with the basics of mlr3
and it’s core objects. The mlr3 book can be a great ressource in case you want to learn more about mlr3’s internals.
We first start by training a model for which we want to conduct an audit. For this example, we use the adult_train
dataset. Keep in mind all the datasets from mlr3fairness package already set protected attribute via the col_role
“pta”, here the “sex” column. To speed things up, we only use the first 1000 rows.
library(mlr3fairness)
library(mlr3learners)
t = tsk("adult_train")$filter(1:1000)
t$col_roles$pta
#> [1] "sex"
Our model is a random forest model fitted on the dataset:
We can now predict on a new dataset and use those predictions to assess for bias:
Using the $score
method and a measure we can e.g. compute the absolute differences in true positive rates.
The exact measure to choose is often data-set and situation dependent. The Aequitas Fairness Tree can be a great ressource to get started.
We can furthermore simply look at the per-group measures:
Before, we have used msr("fairness.tpr")
to assess differences in false positive rates across groups. But what happens internally?
The msr()
function is used to obtain a Measure
from a dictionary of pre-defined Measure
s. We can use msr()
without any arguments in order to print a list of available measures. In the following example, we will build a Measure
that computes differences in False Positive Rates making use of the "classif.fpr"
measure readily implemented in mlr3
.
# Binary Class false positive rates
msr("classif.fpr")
#> <MeasureBinarySimple:classif.fpr>: False Positive Rate
#> * Packages: mlr3, mlr3measures
#> * Range: [0, 1]
#> * Minimize: TRUE
#> * Average: macro
#> * Parameters: list()
#> * Properties: -
#> * Predict type: response
The core Measure
in mlr3fairness
is a MeasureFairness
. It can be used to construct arbitrary measures that compute a difference between a specific metric across groups. We can therefore build a new metric as follows:
m1 = MeasureFairness$new(base_measure = msr("classif.fpr"), operation = function(x) {abs(x[1] - x[2])})
m1
#> <MeasureFairness:fairness.fpr>
#> * Packages: mlr3, mlr3fairness
#> * Range: [-Inf, Inf]
#> * Minimize: TRUE
#> * Average: macro
#> * Parameters: list()
#> * Properties: requires_task
#> * Predict type: response
This measure does the following steps: - Compute the metric supplied as base_measure
in each group defined by the "pta"
column. - Compute operation
(here abs(x[1] - x[2])
) and return the result.
In some cases, we might also want to replace the operation with a different operation, e.g. x[1] / x[2]
in order to compute a different perspective.
mlr3fairness
comes with two built-in functions that can be used to compute fairness metrics also across protected attributes that have more than two classes.
groupdiff_absdiff
: maximum absolute difference between all classes (the default for all metrics)groupdiff_tau
: minimum quotient between all classesNote: Depending on the operation
we need to set a different minimize
flag for the measure, so subsequent operations based on the measure automatically know if the measure is to be minimized or maximized e.g. during tuning.
We can also use those operations to construct a measure using msr()
, since MeasureFairness
(key: msr("fairness")
) can be constructed from the dictionary with additional arguments.
This allows us to construct pretty flexible metrics e.g. for regression settings:
While fairness measures are widely defined or used with binary protected attributes, we can easily extend fairness measures such that they work with non-binary valued protected attributes.
In order to do this, we have to supply an operation
that reduces the desired metric measured in each subgroup to a single value. Two examples for such operations are groupdiff_absdiff
and groupdiff_tau
but custom functions can also be applied. Note, that mlr3 Measure
s are designed for a scalar output and operation
therefore always has to result in a single scalar value.
Some fairness measures also require a combination of multiple Fairness Metrics. In the following example we show how to compute the mean of two fairness metrics, here false negative and true negative rates and create a new Measure
that computes the mean (see aggfun
) of those metrics:
ms = list(msr("fairness.fnr"), msr("fairness.tnr"))
ms
#> [[1]]
#> <MeasureFairness:fairness.fnr>
#> * Packages: mlr3, mlr3fairness
#> * Range: [0, 1]
#> * Minimize: TRUE
#> * Average: macro
#> * Parameters: list()
#> * Properties: requires_task
#> * Predict type: response
#>
#> [[2]]
#> <MeasureFairness:fairness.tnr>
#> * Packages: mlr3, mlr3fairness
#> * Range: [0, 1]
#> * Minimize: TRUE
#> * Average: macro
#> * Parameters: list()
#> * Properties: requires_task
#> * Predict type: response
m = MeasureFairnessComposite$new(measures = ms, aggfun = mean)
In this example, we create a BenchmarkInstance
. Then by using aggregate()
function they could access the fairness measures easily. The following example demonstrates the process to evaluate the fairness metrics on Benchmark Results:
design = benchmark_grid(
tasks = tsks("compas"),
learners = lrns(c("classif.ranger", "classif.rpart"),
predict_type = "prob", predict_sets = c("train", "predict")),
resamplings = rsmps("cv", folds = 3)
)
bmr = benchmark(design)
#> INFO [18:31:43.775] [mlr3] Running benchmark with 6 resampling iterations
#> INFO [18:31:43.782] [mlr3] Applying learner 'classif.rpart' on task 'compas' (iter 3/3)
#> INFO [18:31:43.948] [mlr3] Applying learner 'classif.ranger' on task 'compas' (iter 1/3)
#> INFO [18:31:45.625] [mlr3] Applying learner 'classif.rpart' on task 'compas' (iter 1/3)
#> INFO [18:31:45.834] [mlr3] Applying learner 'classif.rpart' on task 'compas' (iter 2/3)
#> INFO [18:31:45.991] [mlr3] Applying learner 'classif.ranger' on task 'compas' (iter 2/3)
#> INFO [18:31:47.761] [mlr3] Applying learner 'classif.ranger' on task 'compas' (iter 3/3)
#> INFO [18:31:49.434] [mlr3] Finished benchmark
# Operations have been set to `groupwise_quotient()`
measures = list( msr("fairness.tpr"), msr("fairness.npv"), msr("fairness.acc"), msr("classif.acc") )
tab = bmr$aggregate(measures)
#> Warning: Measure 'fairness.tpr' needs predict sets 'test', but learner
#> 'classif.ranger' only predicted on sets 'train', 'predict'
#> Warning: Measure 'fairness.npv' needs predict sets 'test', but learner
#> 'classif.ranger' only predicted on sets 'train', 'predict'
#> Warning: Measure 'fairness.acc' needs predict sets 'test', but learner
#> 'classif.ranger' only predicted on sets 'train', 'predict'
#> Warning: Measure 'classif.acc' needs predict sets 'test', but learner
#> 'classif.ranger' only predicted on sets 'train', 'predict'
#> Warning: Measure 'fairness.tpr' needs predict sets 'test', but learner
#> 'classif.rpart' only predicted on sets 'train', 'predict'
#> Warning: Measure 'fairness.npv' needs predict sets 'test', but learner
#> 'classif.rpart' only predicted on sets 'train', 'predict'
#> Warning: Measure 'fairness.acc' needs predict sets 'test', but learner
#> 'classif.rpart' only predicted on sets 'train', 'predict'
#> Warning: Measure 'classif.acc' needs predict sets 'test', but learner
#> 'classif.rpart' only predicted on sets 'train', 'predict'
tab
#> nr resample_result task_id learner_id resampling_id iters
#> 1: 1 <ResampleResult[22]> compas classif.ranger cv 3
#> 2: 2 <ResampleResult[22]> compas classif.rpart cv 3
#> fairness.tpr fairness.npv fairness.acc classif.acc
#> 1: NaN NaN NaN NaN
#> 2: NaN NaN NaN NaN
For MeasureFairness
, mlr3 computes the base_measure
in each group specified by the pta
column. If we now want to return those measures, we need to aggregate this to a single metric - e.g. using one of the groupdiff_*
functions available with mlr3. See ?groupdiff_tau
for a list. Note, that the operation
below also accepts custom aggregation function, see the example below.
msr("fairness.acc", operation = groupdiff_diff)
#> <MeasureFairness:fairness.acc>
#> * Packages: mlr3, mlr3fairness
#> * Range: [0, 1]
#> * Minimize: TRUE
#> * Average: macro
#> * Parameters: list()
#> * Properties: requires_task
#> * Predict type: response
We can also report other metrics, e.g. the error in a specific group: