Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 1 is a system claim thus it falls into one of the four categories of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding independent claim 1, following limitations recite a judicial exception:
“generates synthetic datasets from the training set wherein sensitive protected attributes are simulated”
[Mental Process] – generating datasets from the training set requires comparing and analyzing the training set which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
“runs the initial AI model against synthetic datasets to gage robustness of the initial AI model by scoring the initial AI model for each test set and exposing the sensitive protected attributes to target for bias mitigation”
[Mental Process] – running against the synthetic datasets to gage robustness by scoring the test set requires comparing and analyzing the datasets to yield the scores of the dataset and compare them to the threshold to determine if the scores actually exceed the threshold involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 1, the claim recites additional elements of
“a memory that stores computer executable components”
The memory is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“a processor that executes the computer executable components stored in the memory”
The processor is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“executable components”, “construction components”, “generation components”
These components are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“constructs an initial artificial intelligence (AI) model using a structured data set with continuous, binary or multi-class prediction labels”
Constructing AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional elements [1,2,3] are considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional elements [4,5] are considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 2
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 2 is a dependent claim of 1, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 2 does not have any abstract idea by itself, thus uses all the limitations of Claim 1.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
The claim 2 does not recite any additional elements other than abstract ideas, so it does not integrate into a practical application. Thus, this claim is directed to the abstract idea.
Regarding Claim 3
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 3 is a dependent claim of 1, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 3 does not have any abstract idea by itself, thus uses all the limitations of Claim 1.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 3, the claim recites additional elements of
“synthetic data is generated using techniques comprising at least one of Random Sampling, Data Augmentation, Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), Copula Models, Markov Chain Monte Carlo (MCMC) Methods and Rule-based Models”
The listed techniques are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 4
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 4 is a dependent claim of 1, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 4 does not have any abstract idea by itself, thus uses all the limitations of Claim 1.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
The claim 4 does not recite any additional elements other than abstract ideas, so it does not integrate into a practical application. Thus, this claim is directed to the abstract idea.
Regarding Claim 5
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 5 is a dependent claim of 5, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 5 does not have any abstract idea by itself, thus uses all the limitations of Claim 1.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
The claim 5 does not recite any additional elements other than abstract ideas, so it does not integrate into a practical application. Thus, this claim is directed to the abstract idea.
Regarding Claim 6
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 6 is a dependent claim of 1, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 6, following limitations recite a judicial exception:
“analyzing indirect bias, test sets are run through a model pipeline to associate the sensitive protected attributes to data points”
[Mental Process] – analyzing indirect bias of the test sets require comparing and analyzing them which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 6, the claim recites additional elements of
“model pipeline”
The model pipeline is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 7
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 7 is a dependent claim of 1, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 7, following limitations recite a judicial exception:
“a score is determined for the initial AI model using each individual test set”
[Mental Process] – determining a score using the test set requires comparing and analyzing the set to the ground truth answers which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 7, the claim recites additional elements of
“the initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 8
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 8 is a dependent claim of 7, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 8, following limitations recite a judicial exception:
“the determined score of test sets are analyzed using fairness statistics”
[Mathematical Calculations] – Analyzing the scores using fairness statistics requires mathematical computation which recites to an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
The claim 8 does not recite any additional elements other than abstract ideas, so it does not integrate into a practical application. Thus, this claim is directed to the abstract idea.
Regarding Claim 9
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 9 is a dependent claim of 7, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 9, following limitations recite a judicial exception:
“the determined score of test sets can determine which of the sensitive protected attributes the initial AI model is most susceptible to”
[Mental Process] – using the determined scores to determine which of the attributes is most susceptible to requires comparing and analyzing the scores to the attributes a certain circumstance which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 9, the claim recites additional elements of
“the initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 10
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 10 is a dependent claim of 8, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 10, following limitations recite a judicial exception:
“analysis of fairness statistics can identify which sensitive protected attributes are most sensitive to bias within the initial AI model”
[Mental Process] – using the fairness statistic to identify the most sensitive attributes to bias requires comparing and analyzing each attribute which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 10, the claim recites additional elements of
“the initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 11
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 11 is a method claim thus it falls into one of the four categories of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding independent claim 11, following limitations recite a judicial exception:
“generating by the system, the synthetic datasets from a training set wherein the sensitive protected attributes are simulated”
[Mental Process] – generating datasets from the training set requires comparing and analyzing the training set which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
“executing by the system, the initial AI model against the synthetic data sets to gage robustness of the initial AI model by scoring the initial AI model for each synthetic data set and exposing the sensitive protected attributes to target for bias mitigation”
[Mental Process] – executing against the synthetic datasets to gage robustness by scoring the test set requires comparing and analyzing the datasets to yield the scores of the dataset and compare them to the threshold to determine if the scores actually exceed the threshold involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 11, the claim recites additional elements of
“by a system operatively coupled to a processor”
The system and the processor are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“constructing, by a system operatively coupled to a processor, the initial AI model using a structured data set with continuous, binary or multi-class prediction labels”
Constructing AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional elements [2,3] are considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 12
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 12 is a dependent claim of 11, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 12, following limitations recite a judicial exception:
“generating, by the system, the synthetic datasets from the training set in which single or multiple sensitive protected attributes can be simulated”
[Mental Process] – generating datasets from the training set requires comparing and analyzing the training set which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 12, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 13
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 13 is a dependent claim of 12, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 13, following limitations recite a judicial exception:
“creating, by the system, test samples from the synthetic datasets that shows bias against an unprivileged group”
[Mental Process] – creating the test samples from the synthetic datasets requires comparing and analyzing the set which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 13, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 14
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 14 is a dependent claim of 11, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 14, following limitations recite a judicial exception:
“analyzing, by the system, for indirect bias, the synthetic datasets that are run through a model pipeline to associate sensitive attributes to data points”
[Mental Process] – analyzing indirect bias of the test sets require comparing and analyzing them which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 14, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“model pipeline”
The model pipeline is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional element [2] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 15
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 15 is a dependent claim of 11, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 15, following limitations recite a judicial exception:
“determining, by the system, a score for the model based on using each of the synthetic datasets and analysis of the model using fairness statistics”
[Mathematical Calculations] – determining the score using fairness statistics requires mathematical computation which recites to an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 15, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“the model”
The model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional element [2] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 16
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 16 is a dependent claim of 15, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 16, following limitations recite a judicial exception:
“determining, by the system, which sensitive protected attributes the initial AI model is most susceptible to”
[Mental Process] – determining which attributes is most susceptible to requires comparing and analyzing each attribute a certain circumstance which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 16, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“the initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional element [2] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 17
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 17 is a dependent claim of 15, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 17, following limitations recite a judicial exception:
“identifying, by the system, which tests were performing below expectations and identifying the sensitive protected attributes contributing to the below expectations”
[Mental Process] – identifying which tests and attributes performed poorly requires comparing and analyzing the set which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 17, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 18
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 18 is a product claim thus it falls into one of the four categories of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding independent claim 18, following limitations recite a judicial exception:
“generate, by the processor, synthetic datasets from a training set in wherein sensitive protected attributes are simulated”
[Mental Process] – generating datasets from the training set requires comparing and analyzing the training set which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
“execute, by the processor, the initial model against synthetic data sets to gage robustness of the initial model and scoring the initial AI model for each dataset and exposing the sensitive protected attributes to target for bias mitigation”
[Mental Process] – executing against the synthetic datasets to gage robustness by scoring the test set requires comparing and analyzing the datasets to yield the scores of the dataset and compare them to the threshold to determine if the scores actually exceed the threshold involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 18, the claim recites additional elements of
“by the processor”
The processor is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“constructing, by a system operatively coupled to a processor, the initial AI model using a structured data set with continuous, binary or multi-class prediction labels”
Constructing AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“initial AI model”
The initial AI model is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional elements [2,3] are considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 19
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 19 is a dependent claim of 18, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 19, following limitations recite a judicial exception:
“generate, by the processor, synthetic data using techniques comprising at least one of Random Sampling, Data Augmentation, Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), Copula Models, Markov Chain Monte Carlo (MCMC) Methods and Rule-based Models”
[Mental Process] – generating synthetic data requires comparing and analyzing an original dataset which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 19, the claim recites additional elements of
“by the processor”
The processor is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
“techniques comprising at least one of Random Sampling, Data Augmentation, Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), Copula Models, Markov Chain Monte Carlo (MCMC) Methods and Rule-based Models”
The listed techniques are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
The additional element [2] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
These limitations remain a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional elements represent a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 20
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 20 is a dependent claim of 18, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 20, following limitations recite a judicial exception:
“create, by the processor, test samples from synthetic datasets that shows bias against an unprivileged group”
[Mental Process] – creating test samples that show bias requires comparing and analyzing the set to intentionally come up with such bias which involves observations, evaluations, judgments, and opinions that is capable of being performed in the human mind with the assistance of paper and pen
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 20, the claim recites additional elements of
“by the system”
The system is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered merely computer components that are just to store and execute code-based instructions which are considered a mere instruction to apply an exception and amount to storing and receiving information in memory, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless – (a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1-10 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Saha et al. (Saha), Non-Patent Literature listed in IDS filed on June 28 2023, “Data Synthesis for Testing Black-Box Machine Learning Models”, Published in November 2021, 11 Pages.
As to independent Claim 1,
Saha teaches a computer implemented system, comprising:
a memory that stores computer executable components; and a processor that executes the computer executable components stored in the memory (Saha, Pg6, Right Column, Section 4, Lines6-9, "All the experiments are performed in a machine running macOS 10.14, having 16GB RAM, 2.7Ghz CPU running Intel Core i7 running Python 3.7", wherein Saha discloses that all the experiments were performed on the macOS that comprises a memory and a processor where all the components are included (hereinafter execution component will be referring to them)), wherein the computer executable components comprise:
a construction component that constructs an initial artificial intelligence (AI) model using a structured data set with continuous, binary or multi-class prediction labels (Saha, Pg1, Right Column, Paragraph5, Lines1-2, "The current scope of our system is classification models for tabular data",
Pg2, Figure1,
PNG
media_image1.png
286
572
media_image1.png
Greyscale
,
Pg1, Left Column, Introduction, Paragraph2, Lines3-5, "The data scientists use the technique of splitting the cleaned data to get train and test set in order to build and select the best model in terms of accuracy",
Pg2, Left Column, Paragraph3, Lines5-10, "Each sample/individual is denoted as (X, Y, Z) where X are all attributes used in the prediction, Y is the corresponding ground-truths for the samples in X, and Z is the binary protected attribute which may be included in X. A classifier is a mapping h: X → [0, 1]. The final prediction is denoted by Yˆ where Yˆ = 1 <--> h(X) > 𝜎",
Pg6, Right Column, Section4, Lines9-10, "For each benchmark, we have generated target model (accuracy>85%) using default configuration as in scikit-learn",
Pg7, Left Column, Paragraph5, Lines1-2, "Model Accuracy: We divide the input data in d_train and d_test using the chosen split, and train a model M_1 using d_train", wherein Saha discloses about training or constructing the target model, M_1 (the corresponding initial AI model) as shown in Figure1 that has multiple construction components using tabular data (the corresponding structured data) where the data is in binary and continuous format as each sample is denoted as (X,Y,Z), rendering it functionally equivalent to the claimed invention);
a generation component that generates synthetic datasets from the training set wherein sensitive protected attributes are simulated (Saha, Pg1, Right Column, First Bullet Point, "First, to address the limited test data, we develop a technique to synthetically generate realistic test data... Specifically, the generated data (without any customization) has the similar statistical characteristics of the training data",
Pg1, Right Column, Second Bullet Point, Lines4-9, "To cater to the testing for a different geographic region that has a different Male:Female (say 1:1), it is important to test the fairness metric with such synthetic data. AITEST allows to incorporate user-defined constraints (UDC) to add and/or update the data constraints",
Pg2, Figure1,
PNG
media_image1.png
286
572
media_image1.png
Greyscale
,
Pg2, Right Column, Section3 Data Synthesis, Lines3-6, "Using such constraints, AITEST can generate synthetic data using a new constraint solver, after optionally merging user-defined constraints and path constraints",
Pg6, Right Column, Section3.4.2, Lines1-7, "Specifically just for the group fairness use case, we unconditionally make the protected attribute independent by removing all those associations from the data constraints where it is mentioned as the target. This is done to ensure that the protected attribute in resultant synthetic test inputs follow the same frequency distribution as that of the input training data set", wherein Saha discloses about generating synthetic datasets that simulates the sensitive protected attribute based on the user-defined constraints and data constraints as shown in Figure1 that the training data is used to generate such data, rendering it functionally equivalent to the claimed invention); and
an execution component that runs the initial AI model against synthetic datasets to gage robustness of the initial AI model by scoring the initial AI model for each test set and exposing the sensitive protected attributes to target for bias mitigation (Saha, Pg1, Left Column, Fourth Bullet Point, "Use of synthetic test data to test fairness [18,44], robustness [7,28] properties in AI model which is majorly skipped by current industrial practices"
Pg6, Left Column, Section3.4, Lines1-3, "We further intend to generate realistic test cases to test an input AI model for a number of properties, namely Individual Fairness, Group Fairness and Robustness"
Pg8, Left Column, 3rd,4th,5th Bullet Points, "We use Robustness Score (i.e. RS = #Succ/#Gen) as an evaluation metric... We use Success Score (i.e. SS = #Disc/#Gen) as an appropriate metric to evaluate individual discrimination... We use Disparate Impact(DI) which is a well known metric to evaluate group fairness"
Pg6, Right Column, Section 3.4.1, Lines8-10, "Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood using the below perturbation function p", wherein Saha discloses inputting the synthetic data set to the model to get scorings of each test set (Individual, Group, and Robustness) to find a discriminatory sample that comprises the protected attributes (the corresponding sensitive protected attribute) to mitigate the bias, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 2,
Saha teaches, as mentioned above, all the limitations of Claim 1. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented system of claim 1, wherein
sensitive protected attribute definitions are provided for the sensitive protected attributes including protected classes, privileged groups and unprivileged groups, and favorable labels and unfavorable labels (Saha, Pg2, Left Column, Paragraph3, Lines1-3, "Fairness. A classifier tries not to discriminate individuals or groups defined by the protected attribute (like race, gender, caste, and religion)"
Pg2. Left Column, Paragraph4, Group Fairness, Lines4-6, "the probability of the favorable outcome of the unprivileged and privileged group should be more than a particular threshold"
Pg2, Left Column, Paragraph3, Lines9-12, "The final
prediction is denoted by Yˆ where Yˆ = 1 <--> h(X) > 𝜎. We will use P(Y^=1 | Z=1) as the probability of a favorable outcome (Y^=1) for the privileged group(Z=1)", wherein Saha explicitly discloses the definition of components that makes the sensitive protected attributes as protected classes, privileged groups, and favorable/unfavorable labels, which is identical to the claimed invention.)
As to dependent Claim 3,
Saha teaches, as mentioned above, all the limitations of Claim 1. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented system of claim 1, wherein
synthetic data is generated using techniques comprising at least one of Random Sampling, Data Augmentation, Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), Copula Models, Markov Chain Monte Carlo (MCMC) Methods and Rule-based Models (Saha, Pg1, Left Column, Introduction, Paragraph7, Lines1-3, "Although existing techniques like GAN [14], Variational Autoencoder [16] can generate synthetic realistic data, they are not customizable by user-specification",
Pg9, Right Column, Section5, Paragraph4, "Realistic Data Synthesis. Note that there exist sophisticated techniques, such as GAN and VAE, which can generate realistic synthetic data"
Pg1, Right Column, First Bullet Point, Lines3-4, ""a random generation technique can generated married people younger than 20...",
Pg5, Right Column, Subsection: Path Coverage Constraints, Lines5-6, ""TREPAN (Decision Tree Surrogate) generates random data to augment the input training samples", wherein Saha explicitly discloses such techniques to generate the synthetic dataset, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 4,
Saha teaches, as mentioned above, all the limitations of Claim 1. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented system of claim 1, wherein
for each of the sensitive protected attributes, a test sample is created from the synthetic datasets that shows bias against an unprivileged group (Saha, Pg2, Left Column, Subsection: Individual Fairness, Lines4-9, "Essentially, a test case corresponding to individual fairness consists of a pair of samples where the two samples only differ in protected attribute values - one from the privileged group and the other from the unprivileged group. Formally, h(s) < e and h(s') >= e where s.Z != s'.z and s.y = s'.y ∀ 𝑦 ∈ 𝑌"
Pg6, Right Column, Lines1-9, "We generate a set of synthetic samples using the synthetic test case generation procedure. For each sample, we change the predefined set of protected attribute (like race/gender) to create pair of test cases and which are checked against the model for label match... Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood...", wherein Saha explicitly discloses creating a pair of samples that are identical except the protected attribute Z, (s,s') where the privileged group, s', is designed to achieve advantageous result of the threshold of h(s') >= e while unprivileged group, s or the discriminatory sample (the corresponding test sample showing bias against an unprivileged group), has h(s) < e which is likely to have more biased results, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 5,
Saha teaches, as mentioned above, all the limitations of Claim 1. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented system of claim 1, wherein
for each combination of the sensitive protected attributes, a test set is created from synthetic data that shows bias against an intersection of unprivileged groups (Saha, Pg8, Left Column, Paragraph5, Lines13-16, "Note that we consider only one protected attribute at a time per benchmark. However, the effectiveness of AITEST will not be hampered even by considering multiple protected attributes"
Pg4, Right Column, Second Bullet Point, "For each value-combination of gender, age-grp in source, say {Female, Senior}, the frequency distribution of target education (say primary:secondary:tertiary=1:2:3) is used to generate synthetic values for education for all rows containing {Female, Senior}", wherein Saha explicitly discloses the value-combination (the corresponding intersection of unprivileged groups) such as {Female, Senior} where both Female and Senior features are sensitive protected attributes of gender and age in which such combinations are used to generate synthetic values for all rows containing them, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 6,
Saha teaches, as mentioned above, all the limitations of Claim 1. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented system of claim 1, wherein
for analyzing indirect bias, test sets are run through a model pipeline to associate the sensitive protected attributes to data points (Saha, Pg2, Left Column, Paragraph6, Lines1-3, "Note that just removing the protected attribute from the training data doesn’t ensure fairness in AI models due to the existence of possible indirect bias [4], and therefore such a testing is required"
Pg3, Left Column, Lines3-5, "The cat_num association is defined between a category source and a numeric target column"
Pg3, Left Column, Paragraph2, Lines1-6, "To capture the dependency between all the constraints, we define a directed graph, G(N,E) and call it the Constraint Dependency Graph (CDG). Each node n ∈ N in this CDG corresponds to a feature/column and is annotated with an associated inference error related to individual feature constraints"
Pg3, Paragraph3, Lines1-3, "Consider a toy dataset with five categorical attributes (gender, education, martial, age-grp, intelligence) and one numeric attribute (salary) with relevant associations between them"
Pg3, Figure2,
PNG
media_image2.png
185
700
media_image2.png
Greyscale
Pg6, Left Column, Section3.4, Paragraph2, "The property-based test input generation starts with removing the feature related to input class label along with all the association constraints where this feature is specified as either a source or target. This ensures that any approximation caused due to constraint inference or synthesis does not seep into the prediction labels for these test inputs", wherein Saha explicitly discloses that just removing the sensitive protected attributes would not make the model fair, but indirect bias must be considered. Also, Saha shows the relationship between columns within the dataset by CDG in Figure2. For example, it maps the conditional distributional association between the sensitive protected attribute, gender, and regular attribute, salary, (the corresponding associating the attributes to data points) in which such datasets are also going through the target model to get labels to analyze the indirect bias (the corresponding model pipeline), rendering it functionally equivalent to the claimed invention)
As to dependent Claim 7,
Saha teaches, as mentioned above, all the limitations of Claim 1. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented system of claim 1, wherein
a score is determined for the initial AI model using each individual test set (Saha, Pg6, Left Column, Section3.4, Lines1-3, "We further intend to generate realistic test cases to test an input AI model for a number of properties, namely Individual Fairness, Group Fairness and Robustness"
Pg8, Right Column, Paragraph1, Lines1-8, "In this experiment, we synthesize test inputs using data constraints along with our user-defined constraints where we override the Male-Female ratio (M:F) for gender (protected attribute) in data constraints by a different one... for this experiment, we compare d_test against the AITEST-generated samples satisfying the input data constraints along with the given UDC"
Pg8, Left Column, 3rd,4th,5th Bullet Points, "We use Robustness Score (i.e. RS = #Succ/#Gen) as an evaluation metric... We use Success Score (i.e. SS = #Disc/#Gen) as an appropriate metric to evaluate individual discrimination... We use Disparate Impact(DI) which is a well known metric to evaluate group fairness"
Pg8 Table5,
PNG
media_image3.png
367
447
media_image3.png
Greyscale
, wherein Saha discloses about giving AI model (the corresponding initial AI model) generated synthetic test inputs using data constraints and user-defined constraints which override the Male:Female ratio to create various individual test sets (the corresponding individual test set) which is shown in Table5 as various M:F ratios that each test set comes with its individual scores, rendering it functionally equivalent to the claimed invention)
As to dependent Claim 8,
Saha teaches, as mentioned above, all the limitations of Claim 7. Saha teaches about various individual test sets are created to be tested through the model and each test has its unique scores.
Saha further teaches about the computer-implemented system of claim 7, wherein
the determined score of test sets are analyzed using fairness statistics (Saha, Pg2, Left Column, Paragraph4, Lines2-7, ""Under the definition of disparate impact, a system is fair if: P(Y^ = 1|Z = 0)/P(Y^ = 1|Z = 1) > 𝜖. In other words, the probability of the favorable outcome of the unprivileged and privileged group should be more than a particular threshold. Typically, based on US Govt. rules, in many scenarios 𝜖 = 0.8"
Pg8, Left Column, 5th Bullet Point, "We use Disparate Impact (DI) [12] which is a well known metric to evaluate group fairness. As per the industry standards, any test suite with DI < 0.8 is treated as the one successful in uncovering group bias", wherein Saha explicitly discloses the disparate impact (DI) which is the well known fairness statistical metric to measure the fairness score according to the value of 𝜖, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 9,
Saha teaches, as mentioned above, all the limitations of Claim 7. Saha teaches about various individual test sets are created to be tested through the model and each test has its unique scores.
Saha further teaches about the computer-implemented system of claim 7, wherein
the determined score of test sets can determine which of the sensitive protected attributes the initial AI model is most susceptible to (Saha, Pg8, Left Column, 4th Bullet Points, Lines5-6, "Higher the value of SS and RS, better is the test set in uncovering faults for individual fairness and robustness, respectively"
Pg8, Right Column, 1st, 2nd Bullet Points, "With the variation in M:F ratios, an average improvement of ≈5% and ≈2% is recorded in Robustness and Success Score, resp., over the random test-split. 55% models show an improvement of >5% in Individual Discrimination testing, with Adult-8 M:F=1:2 showing the maximum gain of ≈15%"
Pg6, Right Column, Lines8-12, "Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood using the below perturbation function p. We use an off-the-shelf explainer LIME [30] to get a set of top (say 50%) attributes, X' such that X' ⊂ X, contributing to the explanation of test s", wherein Saha discloses that higher the RS and SS, the model is more likely to cause faults in such dataset environment meaning by comparing these scores it is possible to determine which sensitive protected attributes the model is most susceptible to. Saha explicitly shows that M:F ratio of 1:2 shows the most bias or model's maximum discrimination gain (meaning gender attribute plays significant role in causing bias) which such attributes will be extracted by the LIME, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 10,
Saha teaches, as mentioned above, all the limitations of Claim 8. Saha teaches about Disparate Impact (DI) is used as a fairness statistics metric in determining the score of the group fairness.
Saha further teaches about the computer-implemented system of claim 8, wherein
analysis of fairness statistics can identify which sensitive protected attributes are most sensitive to bias within the initial AI model (Saha, Pg8, Left Column, Paragraph5, Lines13-16, "Note that we consider only one protected attribute at a time per benchmark. However, the effectiveness of AITEST will not be hampered even by considering multiple protected attributes"
Pg2, Left Column, Paragraph4, Lines2-7, ""Under the definition of disparate impact, a system is fair if: P(Y^ = 1|Z = 0)/P(Y^ = 1|Z = 1)
> 𝜖. In other words, the probability of the favorable outcome of the unprivileged and privileged group should be more than a particular threshold. Typically, based on US Govt. rules, in many scenarios 𝜖 = 0.8"
Pg6, Right Column, Lines8-12, "Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood using the below perturbation function p. We use an off-the-shelf explainer LIME [30] to get a set of top (say 50%) attributes, X' such that X' ⊂ X, contributing to the explanation of test s", wherein Saha discloses that once a discriminatory sample is found, LIME, the explainer, is called to identify top attributes, X', that contributed the most bias to the model by using DI mentioned above by inserting the attributes into the DI equation and if the attributes' DI score is lower than 0.8 they are considered to be very sensitive, rendering it functionally equivalent to the claimed invention.)
As to independent Claim 11,
it is a method claim that contains similar limitations of Claim 1 and thus rejected under the same rationale.
As to dependent Claim 12,
Saha teaches, as mentioned above, all the limitations of Claim 11. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented method of claim 11, further comprising:
generating, by the system, the synthetic datasets from the training set in which single or multiple sensitive protected attributes can be simulated (Saha, Pg1, Right Column, First Bullet Point, "First, to address the limited test data, we develop a technique to synthetically generate realistic test data... Specifically, the generated data (without any customization) has the similar statistical characteristics of the training data"
Pg6, Right Column, Section 3.4.2, Lines7-10, "or example, if gender in the training data has composition of M:F in the ratio 2:1, then, it is desirable to test the model for group fairness with the test cases having same 2:1 M:F ratio"
Pg8, Left Column, Paragraph5, Lines13-16, "Note that we consider only one protected attribute at a time per benchmark. However, the effectiveness of AITEST will not be hampered even by considering multiple protected attributes"
Pg4, Right Column, Second Bullet Point, "For each value-combination of gender, age-grp in source, say {Female, Senior}, the frequency distribution of target education (say primary:secondary:tertiary=1:2:3) is used to generate synthetic values for education for all rows containing {Female, Senior}", wherein Saha explicitly discloses the synthetic dataset is generated from the training set which comprises sensitive protected attributes that can in either single (gender) or multiple (gender, age together), rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 13,
Saha teaches, as mentioned above, all the limitations of Claim 12. Saha teaches about the generated synthetic datasets comprises single or multiple (combination) of sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented method of claim 12, further comprising:
creating, by the system, test samples from the synthetic datasets that shows bias against an unprivileged group (Saha, Pg2, Left Column, Subsection: Individual Fairness, Lines4-9, "Essentially, a test case corresponding to individual fairness consists of a pair of samples where the two samples only differ in protected attribute values - one from the privileged group and the other from the unprivileged group. Formally, h(s) < e and h(s') >= e where s.Z != s'.z and s.y = s'.y ∀ 𝑦 ∈ 𝑌"
Pg6, Right Column, Lines1-9, "We generate a set of synthetic samples using the synthetic test case generation procedure. For each sample, we change the predefined set of protected attribute (like race/gender) to create pair of test cases and which are checked against the model for label match... Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood...", wherein Saha explicitly discloses creating a pair of samples, from the synthetic test case generation procedure which generates synthetic datasets by adding variations to the attributes, that are identical except the protected attribute Z, (s,s') where the privileged group, s', is designed to achieve advantageous result of the threshold of h(s') >= e while unprivileged group, s or the discriminatory sample (the corresponding test sample showing bias against an unprivileged group), has h(s) < e which is likely to have more biased results, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 14,
Saha teaches, as mentioned above, all the limitations of Claim 11. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented method of claim 11, further comprising:
analyzing, by the system, for indirect bias, the synthetic datasets that are run through a model pipeline to associate sensitive attributes to data points (Saha, Pg2, Left Column, Paragraph6, Lines1-3, "Note that just removing the protected attribute from the training data doesn’t ensure fairness in AI models due to the existence of possible indirect bias [4], and therefore such a testing is required"
Pg3, Left Column, Lines3-5, "The cat_num association is defined between a category source and a numeric target column"
Pg3, Left Column, Paragraph2, Lines1-6, "To capture the dependency between all the constraints, we define a directed graph, G(N,E) and call it the Constraint Dependency Graph (CDG). Each node n ∈ N in this CDG corresponds to a feature/column and is annotated with an associated inference error related to individual feature constraints"
Pg3, Paragraph3, Lines1-3, "Consider a toy dataset with five categorical attributes (gender, education, martial, age-grp, intelligence) and one numeric attribute (salary) with relevant associations between them"
Pg3, Figure2,
PNG
media_image2.png
185
700
media_image2.png
Greyscale
Pg6, Left Column, Section3.4, Paragraph2, "The property-based test input generation starts with removing the feature related to input class label along with all the association constraints where this feature is specified as either a source or target. This ensures that any approximation caused due to constraint inference or synthesis does not seep into the prediction labels for these test inputs", wherein Saha explicitly discloses that just removing the sensitive protected attributes would not make the model fair, but indirect bias must be considered. Also, Saha shows the relationship between columns within the dataset by CDG in Figure2. For example, it maps the conditional distributional association between the sensitive protected attribute, gender, and regular attribute, salary, (the corresponding associating the attributes to data points) in which such datasets (the corresponding synthetic datasets) are also going through the target model to get labels to analyze the indirect bias (the corresponding model pipeline), rendering it functionally equivalent to the claimed invention)
As to dependent Claim 15,
Saha teaches, as mentioned above, all the limitations of Claim 11. Saha teaches about the overall architecture of determining a sensitive protected attribute that is target for bias mitigation by preparing the initial AI model, generating synthetic datasets from the training dataset that comprises sensitive protected attributes to be simulated.
Saha further teaches about the computer-implemented method of claim 11, further comprising:
determining, by the system, a score for the model based on using each of the synthetic datasets and analysis of the model using fairness statistics (Saha, Pg6, Left Column, Section3.4, Lines1-3, "We further intend to generate realistic test cases to test an input AI model for a number of properties, namely Individual Fairness, Group Fairness and Robustness"
Pg8, Right Column, Paragraph1, Lines1-8, "In this experiment, we synthesize test inputs using data constraints along with our user-defined constraints where we override the Male-Female ratio (M:F) for gender (protected attribute) in data constraints by a different one... for this experiment, we compare d_test against the AITEST-generated samples satisfying the input data constraints along with the given UDC"
Pg8, Left Column, 3rd,4th,5th Bullet Points, "We use Robustness Score (i.e. RS = #Succ/#Gen) as an evaluation metric... We use Success Score (i.e. SS = #Disc/#Gen) as an appropriate metric to evaluate individual discrimination... We use Disparate Impact(DI) which is a well known metric to evaluate group fairness. As per the industry standards, any test suite with DI < 0.8 is treated as the one successful in uncovering group bias"
Pg8 Table5,
PNG
media_image3.png
367
447
media_image3.png
Greyscale
,
Pg2, Left Column, Paragraph4, Lines2-7, ""Under the definition of disparate impact, a system is fair if: P(Y^ = 1|Z = 0)/P(Y^ = 1|Z = 1) > 𝜖. In other words, the probability of the favorable outcome of the unprivileged and privileged group should be more than a particular threshold. Typically, based on US Govt. rules, in many scenarios 𝜖 = 0.8", wherein Saha discloses about giving AI model (the corresponding initial AI model) generated synthetic test inputs using data constraints and user-defined constraints which override the Male:Female ratio to create various individual test sets (the corresponding individual test set) which is shown in Table5 as various M:F ratios that each test set comes with its individual scores including DI score which is a well-known fairness statistics metric, rendering it functionally equivalent to the claimed invention)
As to dependent Claim 16,
Saha teaches, as mentioned above, all the limitations of Claim 15. Saha teaches about various generated synthetic datasets are analyzed to determine each dataset’s score using fairness statistics metric, disparate impact DI.
Saha further teaches about the computer-implemented method of claim 15, further comprising:
determining, by the system, which sensitive protected attributes the initial AI model is most susceptible to (Saha, Pg8, Left Column, 4th Bullet Points, Lines5-6, "Higher the value of SS and RS, better is the test set in uncovering faults for individual fairness and robustness, respectively"
Pg8, Right Column, 1st, 2nd Bullet Points, "With the variation in M:F ratios, an average improvement of ≈5% and ≈2% is recorded in Robustness and Success Score, resp., over the random test-split. 55% models show an improvement of >5% in Individual Discrimination testing, with Adult-8 M:F=1:2 showing the maximum gain of ≈15%"
Pg6, Right Column, Lines8-12, "Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood using the below perturbation function p. We use an off-the-shelf explainer LIME [30] to get a set of top (say 50%) attributes, X' such that X' ⊂ X, contributing to the explanation of test s", wherein Saha discloses that higher the RS and SS, the model is more likely to cause faults in such dataset environment meaning by comparing these scores it is possible to determine which sensitive protected attributes the model is most susceptible to. Saha explicitly shows that M:F ratio of 1:2 shows the most bias or model's maximum discrimination gain (meaning gender attribute plays significant role in causing bias) which such attributes will be extracted or determined by the LIME, rendering it functionally equivalent to the claimed invention.)
As to dependent Claim 17,
Saha teaches, as mentioned above, all the limitations of Claim 15. Saha teaches about various generated synthetic datasets are analyzed to determine each dataset’s score using fairness statistics metric, disparate impact DI.
Saha further teaches about the computer-implemented method of claim 15, further comprising:
identifying, by the system, which tests were performing below expectations and identifying the sensitive protected attributes contributing to the below expectations (Saha, Pg8, Left Column, 3,4,5 Bullet Points, Lines2-3, 3-4, 2-4"Succ denotes the subset of the generated test cases (Gen) which fail robustness testing...Disc denotes the subset of the generated test cases (Gen) which results in individual discrimination...As per the industry standards, any test suite with DI<0.8 is treated as one successful in uncovering group bias"
Pg6, Right Column, Lines8-12, "Once a discriminatory sample, s is found, we generate further more test inputs in its neighborhood using the below perturbation function p. We use an off-the-shelf explainer LIME [30] to get a set of top (say 50%) attributes, X' such that X' ⊂ X, contributing to the explanation of test s", wherein Saha explicitly discloses that the system handles subset of generated test cases that fail the robustness testing, result in individual discrimination, and unsuccessful in uncovering group bias (the corresponding tests that were performing below expectations), which such test cases or the samples will be examined and identified by the explainer, LIME, rendering it functionally equivalent to the claimed invention.)
As to independent Claim 18,
it is a product claim that contains similar limitations of Claim 1 and thus rejected under the same rationale.
As to dependent Claim 19,
it is a product claim that contains similar limitations of Claim 3 and thus rejected under the same rationale.
As to dependent Claim 20,
it is a product claim that contains similar limitations of Claim 4 and thus rejected under the same rationale.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Gujar et al., Non-Patent Literature, “GenEthos: A Synthetic Data Generation System With Bias Detection And Mitigation”, Published in 2022, IEEE, 6Pages
Chaudhary et al., Non-Patent Literature, “FairGen: Fair Synthetic Data Generation”, Published in Dec 2022, arXiv, 5Pages
Tiwald et al., Non-Patent Literature, “REPRESENTATIVE & FAIR SYNTHETIC DATA”, Published in Apr 2021, arXiv, 5Pages
Gupta et al., Non-Patent Literature, “TRANSITIONING FROM REAL TO SYNTHETIC DATA: QUANTIFYING THE BIAS IN MODEL”, Published in May 2021, arXiv, 7Pages
Zhang et al., Non-Patent Literature, “Machine Learning Testing: Survey, Landscapes and Horizons”, Published in Dec 2019, arXiv, 37Pages
Tolbert et al., Non-Patent Literature, “Correcting Underrepresentation and intersectional Bais for Fair Classification”, Published in Jun 19, 2023, arXiv, 21Pages
Filippi et al., Non-Patent Literature, “Intersectional Fairness: A Fractal Approach”, Published in Feb 2023, arXiv, 18Pages
Zhao et al., US Patent Application # US 2023/0154235-A1, Published in May 2023
Venkataraman et al., US Patent Application # US 2023/0008904-A1, Published in Jan 2023
Bhide et al., US Patent Application # US 2021/0406712-A1, Published in Dec 2021
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DONG YOON JUNG whose telephone number is (571)270-0198. The examiner can normally be reached 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Cesar Paula can be reached at (571) 272-4128. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/DONG YOON JUNG/Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights