Fair AI Tools Action Set

Provides actions for calculating bias metrics and mitigating bias for predictive models.

mitigateBiasGradientBoostTree Action

Mitigate bias for gradient boosting tree models.

CASL Syntax

fairAITools.mitigateBiasGradientBoostTree <result=results> <status=rc> /
required parameter mitigateOptions={
biasMetric="DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY",
bound=double,
event="string",
learningRate=double,
logLevel=integer,
maxIters=integer,
predictedVariables={{
name="string"
}, {...}},
responseLevels={"string-1" <, "string-2", ...>},
seed=double,
required parameter sensitiveVariable={
name="string"
},
tolerance=double,
tuneBound=TRUE | FALSE
},
scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>},
required parameter trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}
;
indicates a required parameter

Parameter Descriptions

* mitigateOptions={fairaitools_mitigateBiasActions_mitigateOptions}

specifies a list of parameters that enable you to adjust how the mitigate algorithm behaves.

The fairaitools_mitigateBiasActions_mitigateOptions value can be one or more of the following:

biasMetric="DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY"

specifies the type of bias measurement.

Default PREDICTIVEPARITY
DEMOGRAPHICPARITY

uses demographic parity as the bias metric.

EQUALIZEDODDS

uses equalized odds as the bias metric.

EQUALOPPORTUNITY

uses equalized opportunity as the bias metric.

PREDICTIVEPARITY

uses predictive parity as the bias metric.

bound=double

specifies the bound value for the exponentiated gradient reduction algorithm.

Default 100
Minimum value 1
event="string"

specifies the formatted value of the response (target) variable that represents the event of interest.

learningRate=double

specifies the step size to use in updating the exponentiated gradient reduction algorithm.

Default 0.01
Range (0, 1)
logLevel=integer

specifies the level of log information to print. Higher levels display more information.

Default 1
Range 0–7
maxIters=integer

specifies the maximum number of iterations to run the exponentiated gradient reduction algorithm.

Alias maxIterations
Default 10
Range 1–50
predictedVariables={{casvardesc-1} <, {casvardesc-2}, ...>}

specifies the list of variables that contain the model's predictions. The order of the variables must match the order that you specify in the responseLevels parameter.

name="string"

specifies the name for the variable.

responseLevels={"string-1" <, "string-2", ...>}

specifies the list of formatted values of the response (target) variable. The order of the variables must match the order that you specify in the predictedVariables parameter. If the event parameter is omitted, the first formatted value is assumed to be the event.

Alias targetLevels
seed=double

specifies the seed for the random number generator. By default, the random number stream is based on the time of day from the computer's clock. A zero or negative seed value also results in random number streams that are based on the time of day from the computer's clock. If you want a reproducible random number sequence between runs, specify a value greater than 0.

Default 0
* sensitiveVariable={casvardesc}

specifies the sensitive variable to use in bias calculations.

name="string"

specifies the name for the variable.

tolerance=double

specifies the parity constraint violation tolerance. If you set this parameter to 0, the action trains for the maximum number of iterations, which you specify in the maxIters parameter.

Default 0.005
Range [0–1)
tuneBound=TRUE | FALSE

when set to True, specifies that the bound value must be tuned.

Default FALSE

scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the score action in the astore action set. For more information, see the parameters for that action.

* trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the gbtreeTrain action in the decisionTree action set. For more information, see the parameters for that action.

mitigateBiasGradientBoostTree Action

Mitigate bias for gradient boosting tree models.

Lua Syntax

results, info = s:fairAITools_mitigateBiasGradientBoostTree{
required parameter mitigateOptions={
biasMetric="DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY",
bound=double,
event="string",
learningRate=double,
logLevel=integer,
maxIters=integer,
predictedVariables={{
name="string"
}, {...}},
responseLevels={"string-1" <, "string-2", ...>},
seed=double,
required parameter sensitiveVariable={
name="string"
},
tolerance=double,
tuneBound=true | false
},
scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>},
required parameter trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}
}
indicates a required parameter

Parameter Descriptions

* mitigateOptions={fairaitools_mitigateBiasActions_mitigateOptions}

specifies a list of parameters that enable you to adjust how the mitigate algorithm behaves.

The fairaitools_mitigateBiasActions_mitigateOptions value can be one or more of the following:

biasMetric="DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY"

specifies the type of bias measurement.

Default PREDICTIVEPARITY
DEMOGRAPHICPARITY

uses demographic parity as the bias metric.

EQUALIZEDODDS

uses equalized odds as the bias metric.

EQUALOPPORTUNITY

uses equalized opportunity as the bias metric.

PREDICTIVEPARITY

uses predictive parity as the bias metric.

bound=double

specifies the bound value for the exponentiated gradient reduction algorithm.

Default 100
Minimum value 1
event="string"

specifies the formatted value of the response (target) variable that represents the event of interest.

learningRate=double

specifies the step size to use in updating the exponentiated gradient reduction algorithm.

Default 0.01
Range (0, 1)
logLevel=integer

specifies the level of log information to print. Higher levels display more information.

Default 1
Range 0–7
maxIters=integer

specifies the maximum number of iterations to run the exponentiated gradient reduction algorithm.

Alias maxIterations
Default 10
Range 1–50
predictedVariables={{casvardesc-1} <, {casvardesc-2}, ...>}

specifies the list of variables that contain the model's predictions. The order of the variables must match the order that you specify in the responseLevels parameter.

name="string"

specifies the name for the variable.

responseLevels={"string-1" <, "string-2", ...>}

specifies the list of formatted values of the response (target) variable. The order of the variables must match the order that you specify in the predictedVariables parameter. If the event parameter is omitted, the first formatted value is assumed to be the event.

Alias targetLevels
seed=double

specifies the seed for the random number generator. By default, the random number stream is based on the time of day from the computer's clock. A zero or negative seed value also results in random number streams that are based on the time of day from the computer's clock. If you want a reproducible random number sequence between runs, specify a value greater than 0.

Default 0
* sensitiveVariable={casvardesc}

specifies the sensitive variable to use in bias calculations.

name="string"

specifies the name for the variable.

tolerance=double

specifies the parity constraint violation tolerance. If you set this parameter to 0, the action trains for the maximum number of iterations, which you specify in the maxIters parameter.

Default 0.005
Range [0–1)
tuneBound=true | false

when set to True, specifies that the bound value must be tuned.

Default false

scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the score action in the astore action set. For more information, see the parameters for that action.

* trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the gbtreeTrain action in the decisionTree action set. For more information, see the parameters for that action.

mitigateBiasGradientBoostTree Action

Mitigate bias for gradient boosting tree models.

Python Syntax

results=s.fairAITools.mitigateBiasGradientBoostTree(
required parameter mitigateOptions={
"biasMetric":"DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY",
"bound":double,
"event":"string",
"learningRate":double,
"logLevel":integer,
"maxIters":integer,
"predictedVariables":[{
"name":"string"
}<, {...}>],
"responseLevels":["string-1" <, "string-2", ...>],
"seed":double,
required parameter "sensitiveVariable":{
"name":"string"
},
"tolerance":double,
"tuneBound":True | False
},
scoreOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>},
required parameter trainOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>}
)
indicates a required parameter

Parameter Descriptions

* mitigateOptions={fairaitools_mitigateBiasActions_mitigateOptions}

specifies a list of parameters that enable you to adjust how the mitigate algorithm behaves.

The fairaitools_mitigateBiasActions_mitigateOptions value can be one or more of the following:

"biasMetric":"DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY"

specifies the type of bias measurement.

Default PREDICTIVEPARITY
DEMOGRAPHICPARITY

uses demographic parity as the bias metric.

EQUALIZEDODDS

uses equalized odds as the bias metric.

EQUALOPPORTUNITY

uses equalized opportunity as the bias metric.

PREDICTIVEPARITY

uses predictive parity as the bias metric.

"bound":double

specifies the bound value for the exponentiated gradient reduction algorithm.

Default 100
Minimum value 1
"event":"string"

specifies the formatted value of the response (target) variable that represents the event of interest.

"learningRate":double

specifies the step size to use in updating the exponentiated gradient reduction algorithm.

Default 0.01
Range (0, 1)
"logLevel":integer

specifies the level of log information to print. Higher levels display more information.

Default 1
Range 0–7
"maxIters":integer

specifies the maximum number of iterations to run the exponentiated gradient reduction algorithm.

Alias maxIterations
Default 10
Range 1–50
"predictedVariables":[{casvardesc-1} <, {casvardesc-2}, ...>]

specifies the list of variables that contain the model's predictions. The order of the variables must match the order that you specify in the responseLevels parameter.

"name":"string"

specifies the name for the variable.

"responseLevels":["string-1" <, "string-2", ...>]

specifies the list of formatted values of the response (target) variable. The order of the variables must match the order that you specify in the predictedVariables parameter. If the event parameter is omitted, the first formatted value is assumed to be the event.

Alias targetLevels
"seed":double

specifies the seed for the random number generator. By default, the random number stream is based on the time of day from the computer's clock. A zero or negative seed value also results in random number streams that are based on the time of day from the computer's clock. If you want a reproducible random number sequence between runs, specify a value greater than 0.

Default 0
* "sensitiveVariable":{casvardesc}

specifies the sensitive variable to use in bias calculations.

"name":"string"

specifies the name for the variable.

"tolerance":double

specifies the parity constraint violation tolerance. If you set this parameter to 0, the action trains for the maximum number of iterations, which you specify in the maxIters parameter.

Default 0.005
Range [0–1)
"tuneBound":True | False

when set to True, specifies that the bound value must be tuned.

Default False

scoreOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>}

specifies a list of parameters to be used by the score action in the astore action set. For more information, see the parameters for that action.

* trainOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>}

specifies a list of parameters to be used by the gbtreeTrain action in the decisionTree action set. For more information, see the parameters for that action.

mitigateBiasGradientBoostTree Action

Mitigate bias for gradient boosting tree models.

R Syntax

results <– cas.fairAITools.mitigateBiasGradientBoostTree(s,
required parameter mitigateOptions=list(
biasMetric="DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY",
bound=double,
event="string",
learningRate=double,
logLevel=integer,
maxIters=integer,
predictedVariables=list( list(
name="string"
) <, list(...)>),
responseLevels=list("string-1" <, "string-2", ...>),
seed=double,
required parameter sensitiveVariable=list(
name="string"
),
tolerance=double,
tuneBound=TRUE | FALSE
),
scoreOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>),
required parameter trainOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>)
)
indicates a required parameter

Parameter Descriptions

* mitigateOptions=list(fairaitools_mitigateBiasActions_mitigateOptions)

specifies a list of parameters that enable you to adjust how the mitigate algorithm behaves.

The fairaitools_mitigateBiasActions_mitigateOptions value can be one or more of the following:

biasMetric="DEMOGRAPHICPARITY" | "EQUALIZEDODDS" | "EQUALOPPORTUNITY" | "PREDICTIVEPARITY"

specifies the type of bias measurement.

Default PREDICTIVEPARITY
DEMOGRAPHICPARITY

uses demographic parity as the bias metric.

EQUALIZEDODDS

uses equalized odds as the bias metric.

EQUALOPPORTUNITY

uses equalized opportunity as the bias metric.

PREDICTIVEPARITY

uses predictive parity as the bias metric.

bound=double

specifies the bound value for the exponentiated gradient reduction algorithm.

Default 100
Minimum value 1
event="string"

specifies the formatted value of the response (target) variable that represents the event of interest.

learningRate=double

specifies the step size to use in updating the exponentiated gradient reduction algorithm.

Default 0.01
Range (0, 1)
logLevel=integer

specifies the level of log information to print. Higher levels display more information.

Default 1
Range 0–7
maxIters=integer

specifies the maximum number of iterations to run the exponentiated gradient reduction algorithm.

Alias maxIterations
Default 10
Range 1–50
predictedVariables=list( list(casvardesc-1) <, list(casvardesc-2), ...>)

specifies the list of variables that contain the model's predictions. The order of the variables must match the order that you specify in the responseLevels parameter.

name="string"

specifies the name for the variable.

responseLevels=list("string-1" <, "string-2", ...>)

specifies the list of formatted values of the response (target) variable. The order of the variables must match the order that you specify in the predictedVariables parameter. If the event parameter is omitted, the first formatted value is assumed to be the event.

Alias targetLevels
seed=double

specifies the seed for the random number generator. By default, the random number stream is based on the time of day from the computer's clock. A zero or negative seed value also results in random number streams that are based on the time of day from the computer's clock. If you want a reproducible random number sequence between runs, specify a value greater than 0.

Default 0
* sensitiveVariable=list(casvardesc)

specifies the sensitive variable to use in bias calculations.

name="string"

specifies the name for the variable.

tolerance=double

specifies the parity constraint violation tolerance. If you set this parameter to 0, the action trains for the maximum number of iterations, which you specify in the maxIters parameter.

Default 0.005
Range [0–1)
tuneBound=TRUE | FALSE

when set to True, specifies that the bound value must be tuned.

Default FALSE

scoreOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>)

specifies a list of parameters to be used by the score action in the astore action set. For more information, see the parameters for that action.

* trainOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>)

specifies a list of parameters to be used by the gbtreeTrain action in the decisionTree action set. For more information, see the parameters for that action.

Last updated: November 23, 2025