DETAILED ACTION
1. This office action is in response to Application No. 18666471 filed on 05/16/2024. Claims 1-20 are presented for examination and are currently pending.
Notice of Pre-AIA or AIA Status
2. 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.
3. Claims 1-20 are rejected under 35 U.S.C 101 because the claimed invention is directed towards an abstract idea without significantly more.
Step 1
Independent claim 1 is directed to a method and falls into one of the four statutory categories.
Step 2A, Prong 1
Claim 1 recites the following abstract ideas:
initializing a genetic algorithm with a plurality of points (Mental process directed to initializing a genetic algorithm with a plurality of points. This can be done with a pen and paper),
the plurality of points contains at least one point selected from a group consisting of: a point that specifies multipliers for the plurality of protected groups that maximize a fairness for the plurality of protected groups (Mental process directed to plurality of points that specifies multipliers for the protected groups that maximize a fairness which can be done by observing the point and make a judgment on which point maximize the fairness), and
a point that specifies multipliers for the plurality of protected groups that maximize an accuracy of a machine learning model (Mental process directed to a point that specifies multipliers for the plurality of protected groups that maximize an accuracy which can be done by observing the multipliers for the plurality of protected groups and making a judgement on the groups that maximize accuracy);
generating, by the genetic algorithm, a new plurality of multipliers of the feature (Mental process directed to generating multipliers which can be done with the use of a pen and paper);
inferring, from an input that contains a value of the feature, a probability of a class (Mental process directed to inferring a probability class which can be done with the aid of pen and paper);
selecting based on the value of the feature in the input, from the new plurality of multipliers of the feature, a multiplier that is specific to both of the feature and the value of the feature (Mental process directed to selecting a multiplier based on the value of the feature in the input. This can be done by observing the input feature and making a judgement on the selection); and
classifying the input based on a multiplicative product of the probability of the class and the multiplier that is specific to both of the feature and the value of the feature (Mathematical concepts directed to classifying an input based on a multiplicative product of the probability of the class and the multiplier);
Step 2A, Prong 2
Claim 1 recite the following additional elements:
wherein: each point in the plurality of points specifies, for each protected group of a plurality of protected groups, a multiplier of a feature (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (h)), and
wherein the method is performed by one or more computers (This limitation is directed to a generic computer component to execute the method. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (f)).
Step 2B
Claim 1 recite the following additional elements:
wherein: each point in the plurality of points specifies, for each protected group of a plurality of protected groups, a multiplier of a feature (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not amount to significantly more than the judicial exception. see MPEP 2106.05 (h)), and
wherein the method is performed by one or more computers (This limitation is directed to a generic computer component to execute the method. This limitation does not amount to significantly more than the judicial exception. see MPEP 2106.05 (f)).
4. Dependent claim 2 is directed to a method and falls into one of the four statutory categories.
Claim 2 recites the following abstract ideas:
a point that specifies a multiplicative identity for the multiplier of each protected group of the plurality of protected groups (Mathematical concepts directed to a point that specifies a multiplicative identity for the multiplier),
a point that specifies multipliers for the plurality of protected groups that maximizes an equality of opportunity of the plurality of protected groups (Mental process directed to a point that specifies multipliers for the plurality of protected groups that maximizes an equality of opportunity of the protected groups which can be done by observing the point and make a judgment on which point maximize the equality of opportunity),
a point that specifies multipliers for the plurality of protected groups that minimizes a disparate outcome of the plurality of protected groups (Mental process directed to a point that specifies multipliers for the plurality of protected groups that minimizes an equality of opportunity of the protected groups which can be done by observing the point and make a judgment on which point minimizes the equality of opportunity), and
a point that specifies multipliers for the plurality of protected groups that minimizes an outcome rate decrease of the plurality of protected groups (Mental process directed to a point that specifies multipliers for the plurality of protected groups that minimizes an outcome rate decrease of the protected groups which can be done by observing the point and make a judgment on which point minimizes an outcome rate decrease).
Claim 2 recites the following additional elements:
wherein the plurality of points contains at least one point selected from a group consisting of (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (h)):
Claim 2 recites the following additional elements:
wherein the plurality of points contains at least one point selected from a group consisting of (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not amount to significantly more than the judicial exception. see MPEP 2106.05 (h)):
5. Dependent claim 3 is directed to a method and falls into one of the four statutory categories.
Claim 3 does not recite any abstract ideas.
Claim 3 recites the following additional elements:
wherein the plurality of points contains at least two points that are not randomly generated (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (h)).
Claim 3 recites the following additional elements:
wherein the plurality of points contains at least two points that are not randomly generated (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not amount to significantly more than the judicial exception. see MPEP 2106.05 (h)).
6. Dependent claim 4 is directed to a method and falls into one of the four statutory categories.
Claim 4 recites the following abstract ideas:
measuring a disparate outcome of the machine learning model for the plurality of protected groups (Mental process directed to measuring a disparate outcome of the machine learning model which can be done by observing the machine learning model and making a judgement about the measurement);
penalizing points having a disparate outcome for the plurality of protected groups that exceeds one tenth of a disparate outcome of the machine learning model (Mental process directed to penalizing points having a disparate outcome of the protected groups that exceeds one tenth of a disparate outcome of the machine learning model which can be done by observing the points and making a judgement on the penalizing points that have a disparate outcome for the protected groups).
Claim 4 do not recite any additional elements.
7. Dependent claim 5 is directed to a method and falls into one of the four statutory categories.
Claim 5 recites the following abstract ideas:
measuring an accuracy of the machine learning model (Mental process directed to measuring the accuracy of the machine learning model which can be done by observing the machine learning model and making a judgement on the measurement);
penalizing points having an accuracy less than 99 percent of the accuracy of the machine learning model (Mental process directed to penalizing points having an accuracy less than 99 percent which can be done by observing the point and making a judgment on when to penalize the points).
Claim 5 do not recite ant additional elements.
8. Dependent claim 6 is directed to a method and falls into one of the four statutory categories.
Claim 6 recites the following abstract ideas:
penalizing points having an outcome rate decrease for the plurality of protected groups that exceeds ten percent (Mental process directed to penalizing points having an outcome rate decrease for the protected groups that exceeds ten percent which can be done by observing the points and making a judgement on penalizing the points that have outcome rate decrease).
Claim 6 do not recite any additional elements.
9. Dependent claim 7 is directed to a method and falls into one of the four statutory categories.
Claim 7 does not recite any abstract ideas.
Claim 7 recites the following additional elements:
the genetic algorithm with at least three distinct validation metrics, including a fitness metric and a fairness metric (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (h)).
Claim 7 recites the following additional elements:
the genetic algorithm with at least three distinct validation metrics, including a fitness metric and a fairness metric (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not amount to significantly more than the judicial exception. see MPEP 2106.05 (h)).
10. Dependent claim 8 is directed to a method and falls into one of the four statutory categories.
Claim 8 does not recite any abstract ideas.
Claim 8 recites the following additional elements:
further comprising configuring the genetic algorithm with at least three distinct validation metrics, including two fairness metrics (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (h)).
Claim 8 recites the following additional elements:
further comprising configuring the genetic algorithm with at least three distinct validation metrics, including two fairness metrics (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not amount to significantly more than the judicial exception. see MPEP 2106.05 (h)).
11. Dependent claim 9 is directed to a method and falls into one of the four statutory categories.
Claim 9 recite the following abstract ideas:
further comprising configuring the genetic algorithm with at least three distinct objectives, including an objective that minimizes an outcome rate decrease of the plurality of protected groups (Mental process directed to objectives that minimizes an outcome rate decrease of the protected group using a genetic algorithm which can be aided with pen and paper).
Claim 9 recites the following additional elements:
12. Dependent claim 10 is directed to a method and falls into one of the four statutory categories.
Claim 10 recite the following abstract ideas:
the genetic algorithm generating multiple pluralities of multipliers of the feature (Mental process directed to generating multiple pluralities of multipliers of the feature using the generative algorithm which can be aided with pen and paper);
detecting a subset of the multiple pluralities of multipliers of the feature that are on a tri-objective Pareto frontier (Mental concepts directed to detecting a subset of the multiple pluralities of multipliers of the feature using feature that are on a tri-objective Pareto frontier).
Claim 10 do not recite any additional elements.
13. Independent claim 11 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 11, it is substantially similar to claim 1 and is rejected in the same manner and reasoning applying.
14. Independent claim 12 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 12, it is substantially similar to claim 2 and is rejected in the same manner and reasoning applying.
15. Independent claim 13 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 13, it is substantially similar to claim 3 and is rejected in the same manner and reasoning applying.
16. Independent claim 14 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 14, it is substantially similar to claim 4 and is rejected in the same manner and reasoning applying.
17. Independent claim 15 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 15, it is substantially similar to claim 5 and is rejected in the same manner and reasoning applying.
18. Independent claim 16 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 16, it is substantially similar to claim 6 and is rejected in the same manner and reasoning applying.
19. Independent claim 17 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 17, it is substantially similar to claim 7 and is rejected in the same manner and reasoning applying.
20. Independent claim 18 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 18, it is substantially similar to claim 8 and is rejected in the same manner and reasoning applying.
21. Independent claim 19 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 19, it is substantially similar to claim 9 and is rejected in the same manner and reasoning applying.
22. Independent claim 20 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 20, it is substantially similar to claim 10 and is rejected in the same manner and reasoning applying.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
23. Claims 1-4, 7-14 and 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al. ("Reducing subgroup differences in personnel selection through the application of machine learning." Personnel Psychology 76.4 (2023): 1125-1159) in view of Brookhouse et al. ("Fair feature selection: a comparison of multi-objective genetic algorithms." arXiv preprint arXiv:2310.02752 (2023)) and further in view of Dalli et al. (US20220012591)
Regarding claim 1, Zhang teaches a method (As a result, there is a growing
need to develop robust methods for making hiring decisions that consider multiple objectives, pg. 1147, first para.; The Examiner notes the instant specification discloses: “In a hiring example application, this means that an unqualified candidate will not be selected for hiring if there exists a more qualified candidate in the same protected group” [0027]) comprising:
initializing an algorithm (we describe the data-generating process for the simulation study, the fairness-aware ML algorithms tested, the simulation conditions, and the simulation results, respectively, pg. 1131, last para.) with a plurality of points (All points on a validity-diversity tradeoff curve were generated over the same simulated dataset, Fig. 2, pg. 1135; The Examiner notes that the instant specification: “Herein, each generated solution is referred to as a point”[0015] ),
wherein: each point in the plurality of points (All points on a validity-diversity tradeoff curve were generated over the same simulated dataset, Fig. 2, pg. 1135; The Examiner notes each point on the validity-diversity tradeoff curve is a point) specifies, for each protected group (Racial minorities, Fig. 2b, pg. 1135) of a plurality of protected groups (selected applicants from legally protected groups (e.g., women, racial/ethnic minorities), pg. 1151, third to the last para.), a multiplier of a feature (When 𝜆 is so large that the prediction target is dominated by its second
term, that is, the likelihood of a candidate being a racial minority, pg. 1134, second para.; where 𝜆, the Lagrange multiplier, pg. 1130, last sentence- pg. 1131; dataset features racial majority and minority candidates, pg. 1135, last para.), and
the plurality of points contains at least one point (All points on a validity-diversity tradeoff curve were generated over the same simulated dataset, Fig. 2, pg. 1135) selected from a group consisting of: a point that specifies multipliers for the plurality of protected groups (individual tasks of approximating a prediction target f𝜆(x) for each candidate, where 𝜆, the Lagrange multiplier, pg. 1130, last sentence-pg. 1131; a candidate being a racial minority, pg. 1134, second para.) that maximize a fairness for the plurality of protected groups (u(S) (mean criterion score u(S) … of the selected candidates S) has to be maximized under a fairness, pg. 1130, third to the last para.), and a point that specifies multipliers for the plurality of protected groups that maximize an accuracy of a machine learning model (The Examiner notes this limitation is not mapped because it is an alternative limitation);
generating (To set 𝜆, we performed an iterative optimization like Google’s
TensorFlow Constrained Optimization, pg. 1133, second para.; Note that, the larger 𝜆 is, the more f𝜆 becomes “bended” by the reverse sigmoid function, pg. 1132, Fig. 1. The Examiner notes iterative optimization generates new 𝜆), by the algorithm (we describe the data-generating process for the simulation study, the fairness-aware ML algorithms tested, the simulation conditions, and the simulation results, respectively, pg. 1131, last para.), a new plurality of multipliers of the feature (the method of Lagrange multipliers could assign a different 𝜆 to each constraint, pg. 1137, second para.; dataset features racial majority and minority candidates, pg. 1135, last para.);
inferring, from an input that contains a value of the feature (Input. A fairness-aware ML algorithm takes as input an incumbent dataset—known as training dataset … Output. The algorithm’s output is a prediction model, pg. 1127, last two para.; dataset features racial majority and minority candidates, pg. 1135, last para.), a probability of a class (we had the ML models output class probabilities (i.e., continuous values ranging from 0 to 1), pg. 1143, first para.);
selecting based on the value of the feature in the input (We varied … selection rate s, training dataset size N, pg. 1133, third para.; dataset features racial majority and minority candidates, ), from the new plurality of multipliers of the feature (𝜆 = 0.2, 𝜆 = 0.1, 𝜆 = 0, Fig. 1c, pg. 1132), a multiplier that is specific to both of the feature and the value of the feature (The gray zone represents those candidates who would have been selected if 𝜆 = 0, pg. 1132, Fig. 1; dataset features racial majority and minority candidates, pg. 1135, last para.); and
classifying the input (G = 0 being the racial majority and G = 1 being the racial minority, , pg. 1128, last sentence) based on a multiplicative product of the probability of the class and the multiplier (𝜆 times Pr{G = 1|X = x}, that is, the likelihood for a candidate to be a racial minority given the observed predictors in X., pg. 1131, first para.) that is specific to both of the feature and the value of the feature (dataset features racial majority and minority candidates, pg. 1135, last para.);
Zhang is silent about using a genetic algorithm, wherein the method is performed by one or more computers.
Brookhouse teaches a method comprising: initializing a genetic algorithm with a plurality of points, wherein: each point in the plurality of points (GAs (Genetic Algorithms) were set with the same values, pg. 6, left col., second para.) specifies, for each protected group of a plurality of protected groups (The values of sensitive features can be used to split individuals (records in a dataset) into two groups: protected and unprotected individuals. The protected group contains the individuals that are considered to be subject to unfair bias, pg. 2, left col., first para.)
inferring, from an input that contains a value of the feature, a probability of a class (DP (demographic parity) is a group-level fairness measure that takes the optimal value of 1 if both protected and unprotected groups have an equal probability of being assigned to the positive class, pg. 4, left col., second para.);
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Zhang to incorporate the teachings of Brookhouse for the benefit of selecting a subset of features that optimizes both the predictive accuracy and the fairness of the models (classifiers) learned by the subsequent classification algorithm (Brookhouse, pg. 1, right col., first para.)
Zhang and Brookhouse does not explicitly teach wherein the method is performed by one or more computers.
Dalli teaches wherein the method is performed by one or more computers (In addition, for each of the embodiments described herein, the corresponding form of any such embodiment may be described herein as, for example, “a computer configured to” perform the described action [0027]).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Zhang and Brookhouse to incorporate the teachings of Dalli for the benefit of identifying areas where the samples can be improved to eliminate bias and also reduce the likelihood of both type I and type II errors (Dalli [0081])
Regarding claim 2, Zhang, Brookhouse and Dalli teaches the method of Claim
1, wherein the plurality of points contains at least one point selected from a group consisting of: a point that specifies a multiplicative identity for the multiplier of each protected group of the plurality of protected groups (The Examiner notes this limitation is not mapped because it is an alternative limitation),
Zhang teaches a point that specifies multipliers for the plurality of protected groups that maximizes an equality of opportunity of the plurality of protected groups (The first is when 𝜆 is set to eliminate adverse impact by ensuring an equal
mean of ML-predicted scores, say f𝜆 (x) = F, for both groups, pg. 1131, third para.),
a point that specifies multipliers for the plurality of protected groups that minimizes a disparate outcome of the plurality of protected groups (A key requirement on the output prediction model is … minimizing the adverse impact of selection outcome, pg. 1128, second para.), and
a point that specifies multipliers for the plurality of protected groups that minimizes an outcome rate decrease of the plurality of protected groups (The Examiner notes this limitation is not mapped because it is an alternative limitation).
Regarding claim 3, Zhang, Brookhouse and Dalli teaches the method of Claim 1, Zhang teaches wherein the plurality of points contains at least two points that are not randomly generated (To equalize sample sizes, we multiplied the White N by the desired SR (as determined by the desired AI ratio), pg. 1142, second para. The Examiner notes equal sample sizes are not randomly generated).
Regarding claim 4, Zhang, Brookhouse and Dalli teaches the method of Claim 1, Zhang teaches further comprising: measuring a disparate outcome of the machine learning model for the plurality of protected groups (We start with mathematical analysis showing that, unless a “plain” ML algorithm with no fairness constraint already satisfies the organizational requirement on adverse impact, pg. 1127, second para.);
penalizing points having a disparate outcome for the plurality of protected groups that exceeds one tenth of a disparate outcome of the machine learning model (As a result, these ML predictions tend to unfairly penalize those racial minority candidates who “look like” racial majorities according to the predictor battery. When the mean criterion score of racial minorities is lower than the racial majorities, pg. 1127, second para.).
Regarding claim 7, Zhang, Brookhouse and Dalli teaches the method of Claim 1, Brookhouse teaches further comprising configuring the genetic algorithm (The Lexicographic GA for Fair Feature Selection (LGAFFS) is a recently proposed lexicographic optimisation GA (Genetic Algorithms) that uses the wrapper approach for fair feature selection, pg. 4, right col., section C) with at least three distinct validation metrics, including a fitness metric and a fairness metric (The fitness values of each individual – i.e. the values of GM Sen×Spec and the four fairness measures – are computed by a well-known internal cross-validation applied to the training set only (i.e. not using the test set). This internal cross-validation is used in lines 6 and 7 of Algorithm 1, pg. 5, left col., third para.).
The same motivation to combine independent claim 1 applies here.
Regarding claim 8, Zhang, Brookhouse and Dalli teaches the method of Claim 1, Brookhouse teaches further comprising configuring the genetic algorithm with at least three distinct validation metrics, including two fairness metrics (The GA generates all possible 24 (4!) possible permutations of the four fairness measures, where each permutation specifies a priority order for the four fairness measures to be optimized, pg. 5, right col., last para.).
The same motivation to combine independent claim 1 applies here.
Regarding claim 9, Zhang, Brookhouse and Dalli teaches the method of Claim 1, Zhang teaches further comprising configuring the … algorithm with at least three distinct objectives (Even when the optimization goal is loosely defined (e.g., to improve all three objectives to a reasonable degree), one can still narrow down the solution space based on the goal), including an objective that minimizes an outcome rate decrease of the plurality of protected groups (A key requirement on the output prediction model is to meet an organization’s desired level of validity-diversity tradeoff … minimizing the adverse impact of selection outcome, pg. 1128, second para.).
Brookhouse teaches the genetic algorithm with at least three distinct objectives (GAs (Genetic Algorithms) were set with the same values, pg. 6, left col., second para.; 1) Pareto Multi-Objective Optimisation, pg. 2, right col., last para.)
The same motivation to combine independent claim 1 applies here.
Regarding claim 10, Zhang, Brookhouse and Dalli teaches the method of Claim 1, Zhang teaches further comprising: the … algorithm generating (To set 𝜆, we performed an iterative optimization like Google’s TensorFlow Constrained Optimization, pg. 1133, second para.; Note that, the larger 𝜆 is, the more f𝜆 becomes “bended” by the reverse sigmoid function, pg. 1132, Fig. 1. The Examiner notes iterative optimization generates new 𝜆) multiple pluralities of multipliers of the feature (the method of Lagrange multipliers could assign a different 𝜆 to each constraint, pg. 1137, second para.; dataset features racial majority and minority candidates, pg. 1135, last para.);
detecting a subset of the multiple pluralities of multipliers of the feature that are on a tri-objective Pareto frontier (The goal of MOO is to identify Pareto-optimal solutions. A Pareto-optimal solution optimizes one objective, at a certain level of the other objective(s), pg. 1147, last para.).
Brookhouse teaches genetic algorithm (GAs (Genetic Algorithms) were set with the same values, pg. 6, left col., second para.; The values of sensitive features can be used to split individuals (records in a dataset) into two groups: protected and unprotected individuals, pg. 2, left col., first para.)
The same motivation to combine independent claim 1 applies here.
Regarding claim 11, claim 11 is similar to claim 1. It is rejected in the same manner and reasoning applying. Zhang and Brookhouse do not explicitly teach one or more non-transitory computer-readable media storing instructions that, when executed by one or more processors,
Dalli teaches one or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause (For example, in an exemplary embodiment, the bias information may be used to focus a Genetic Algorithm (GA) [0060]; the sequence of actions described herein can be embodied entirely within any form of computer-readable storage medium such that execution of the sequence of actions enables the at least one processor to perform the functionality described herein [0027]):
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Zhang and Brookhouse to incorporate the teachings of Dalli for the benefit of identifying areas where the samples can be improved to eliminate bias and also reduce the likelihood of both type I and type II errors (Dalli [0081])
Regarding claim 12, claim 12 is similar to claim 2. It is rejected in the same manner and reasoning applying.
Regarding claim 13, claim 13 is similar to claim 3. It is rejected in the same manner and reasoning applying.
Regarding claim 14, claim 14 is similar to claim 4. It is rejected in the same manner and reasoning applying.
Regarding claim 17, claim 17 is similar to claim 7. It is rejected in the same manner and reasoning applying.
Regarding claim 18, claim 18 is similar to claim 8. It is rejected in the same manner and reasoning applying.
Regarding claim 19, claim 19 is similar to claim 9. It is rejected in the same manner and reasoning applying.
Regarding claim 20, claim 20 is similar to claim 10. It is rejected in the same manner and reasoning applying.
24. Claims 5, 6, 15 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al. ("Reducing subgroup differences in personnel selection through the application of machine learning." Personnel Psychology 76.4 (2023): 1125-1159) in view of Brookhouse et al. ("Fair feature selection: a comparison of multi-objective genetic algorithms." arXiv preprint arXiv:2310.02752 (2023)) in view of Dalli et al. (US20220012591) and further in view of Zafar et al. ("Fairness constraints: A flexible approach for fair classification" Journal of Machine Learning Research 20.75 (2019): 1-42).
Regarding claim 5, Zhang, Brookhouse and Dalli teaches the method of Claim 1, they do not explicitly teach the limitations of claim 5.
Zafar teaches further comprising: measuring an accuracy of the machine learning model (Next, we train logistic regression classifiers optimizing for accuracy on both the datasets, pg. 16, first para.);
penalizing points (it iteratively re-trains the classifier with increasingly higher penalties on this set of data points, pg. 29, first para.) having an accuracy less than 99 percent of the accuracy of the machine learning model (The accuracy of the classifiers in both cases is 0.87, pg. 16, first para. The Examiner notes that 0.87 accuracy (87%) is less than 99%).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Zhang, Brookhouse and Dalli to incorporate the teachings of Zafar for the benefit of successfully limiting unfairness which is often at a small cost in terms of accuracy (Zafar, abstract)
Regarding claim 6, Zhang, Brookhouse and Dalli teaches the method of Claim 1, they do not explicitly teach the limitations of claim 6
Zafar teaches further comprising penalizing points having an outcome rate decrease for the plurality of protected groups (adding a regularization term in the objective that penalizes the mutual information between the sensitive feature and the classifier decisions, pg. 34, second para.) that exceeds ten percent (logistic regression classifier leads to an accuracy of 0.664, pg. 29, second to the last para. The Examiner notes accuracy of 0.664 which 66.4% exceeds 10%).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Zhang, Brookhouse and Dalli to incorporate the teachings of Zafar for the benefit of successfully limiting unfairness which is often at a small cost in terms of accuracy (Zafar, abstract)
Regarding claim 15, claim 15 is similar to claim 5. It is rejected in the same manner and reasoning applying.
Regarding claim 16, claim 16 is similar to claim 6. It is rejected in the same manner and reasoning applying.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MORIAM MOSUNMOLA GODO whose telephone number is (571)272-8670. The examiner can normally be reached Monday-Friday 8:00am-5:00pm EST.
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, Michelle T. Bechtold can be reached on (571) 431-0762. 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.
/M.G./Examiner, Art Unit 2148
Prosecution Insights