DETAILED ACTION
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 .
The filing date of the present invention is 03/31/2023.
This action is in response to amendment and/or remarks filed on 02/24/2026. In the current amendments, claims 1, 12 and 20 have been amended. Claims 1-20 are currently pending and have been examined.
Response to Arguments
Applicant's arguments filed 02/24/2026 have been fully considered but they are not persuasive.
Rejections Under 35 U.S.C. 101:
Applicant asserts that “The claimed invention provides a non-generic technical solution to a technical problem inherent in algorithmic monoculture detection. This inventive concept, therefore, resides in the unconventional integration of "determining whether the computing algorithm suffers from an algorithmic monoculture." This is effectively achieved through the strategic use of "constructing...a structural causal model" and "identifying a mediator variable and a confounder variable associated with the computing algorithm, based on the constructed SCM." This approach goes beyond merely observing a human activity; it represents a profound technical solution that fundamentally alters the optimization process to rigorously account for and leverage human operational characteristics, thereby leading to a system that functions superiorly in a real-world context for both human operators and automated processes. This is specifically the kind of technological improvement that distinguishes patent-eligible claims from abstract ideas. Furthermore, the determining yields a technical solution where an algorithmic monoculture may be detected. This is not a generic optimization but a highly specialized one that produces a tangible, improved outcome in the intricate operation of a logistics system. This specific and innovative method provides an inventive concept by transforming an abstract idea into a concrete and significantly improved technical solution for bias detection. Thus, the subject matter of the claims amounts to significantly more than the judicial exception (Step 2B: Yes). Accordingly, withdrawal of the rejection is respectfully requested." as recited in Claim 1 (and similarly in Claim 18). (Remarks pg. 9-10)
Examiner’s response:
The Examiner respectfully disagrees. The claim as a whole is still directed to abstract idea mental process. With regards to the argument determining yield a technical solution where an algorithmic monoculture may detect and optimize highly specialized one. The determine steps is still performable in the human mind.
Regarding applicant’s reliance on the decision of the Appeals Review Panel in Ex parte Desjardins, No. 2024-000567 (P.T.A.B. Sept. 26, 2025), Examiner notes that, in Desjardins, unlike in the claims at issue here, the appellants specifically argued that the claimed invention “address[es] challenges in continual learning and model efficiency by reducing storage requirements and preserving task performance across sequential training”. Desjardins, op. at 7. That is, the appellant in Desjardins specifically alleged that the claimed subject matter improves machine learning itself. By contrast, Applicant in the instant case does not point to any specific claim language that characterizing an improvement, and does not point to any claim language that is analogous to the claims at issue in Desjardins.
Rejections Under 35 U.S.C. 101:
Applicant’s arguments with respect to claim(s) 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
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 and does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception.
Regarding claim 1
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“….encoding domain knowledge information associated with the received dataset; constructing, for a computing algorithm related to the domain, a … associated with the dataset based on the encoded domain knowledge information; identifying a mediator variable and a confounder variable associated with the computing algorithm, based on the constructed SCM; …determining whether the computing algorithm suffers from an algorithmic monoculture, based on the estimated causal effect, to detect a popularity bias in the computing algorithm; and rendering information indicative of whether the computing algorithm suffers from the algorithmic monoculture.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Step 2A Prong 2: This judicial exception is not integrated into a practical. In particular, the claim only recites additional elements that are mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. See MPEP 2106.05(f). The additional element of “A method, executed by a processor, … structural causal model (SCM) estimating a causal effect associated with the computing algorithm, based on the identified mediator variable and the identified confounder variable;”, as drafted, is reciting generic computer components. The generic computer components in these steps are recited at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. In addition, the claim limitation “
comprising: receiving a dataset associated with a domain” as explained by the Supreme Court, the addition of insignificant extra-solution activity does not amount to an inventive concept, particularly when the activity is well-understood or conventional. See MPEP 2106.05(g). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The additional element of using generic computer components to perform the abstract idea amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The only remaining limitation of the claim “comprising: receiving a dataset associated with a domain” constitute storing and retrieving information in memory, which the courts have found to be well-understood, routine, and conventional. See MPEP 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
Regarding claim 2
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1:
“determining information associated with an observed variable and an unobserved variable associated with the domain; and identifying an input variable and an output variable associated with a problem area associated with the domain, wherein the encoding of the domain knowledge information is further based on the determined information associated with the observed variable and the unobserved variable and the identified input variable and the output variable.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 3
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1:
“wherein the … is constructed based on at least one of the computing algorithm associated with the SCM, a number of citations of the computing algorithm, a confounder variable associated with the computing algorithm, a probability that the computing algorithm is used, an outcome of the computing algorithm, or a familiarity of the computing algorithm.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Step 2A Prong 2: This judicial exception is not integrated into a practical. In particular, the claim only recites additional elements that are mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. See MPEP 2106.05(f). The additional element of “structural causal model (SCM)”, as drafted, is reciting generic computer components. The generic computer components in these steps are recited at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of using generic computer components to perform the abstract idea amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. Thus, the claim is not patent eligible.
Regarding claim 4
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“wherein the … corresponds to a directed acyclic graph that indicates interdependencies in variables associated with the computing algorithm, based on the encoded domain knowledge information”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Step 2A Prong 2: This judicial exception is not integrated into a practical. In particular, the claim only recites additional elements that are mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. See MPEP 2106.05(f). The additional element of “SCM”, as drafted, is reciting generic computer components. The generic computer components in these steps are recited at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into practical application, the additional element of using generic computer components to perform the abstract idea amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. Thus, the claim is not patent eligible.
Regarding claim 5
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“further comprising: determining one or more variables associated with a disparate outcome or a disparate treatment associated with the computing algorithm; computing a direct effect, an indirect effect, and a spurious effect of each of the determined one or more variables based on the …; and establishing an evidence of the disparate outcome or the disparate treatment, based on the computed direct effect, indirect effect, and the spurious effect.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Step 2A Prong 2: This judicial exception is not integrated into a practical. In particular, the claim only recites additional elements that are mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. See MPEP 2106.05(f). The additional element of “SCM”, as drafted, is reciting generic computer components. The generic computer components in these steps are recited at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into practical application, the additional element of using generic computer components to perform the abstract idea amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. Thus, the claim is not patent eligible.
Regarding claim 6
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“wherein the mediator variable is a variable that affects an input variable associated with the computing algorithm, and the confounder variable is a variable that affects the input variable and an output variable associated with the computing algorithm.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 7
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“further comprising: determining a natural direct effect, a natural indirect effect, and a spurious effect based on the estimated causal effect, wherein the determination of whether the computing algorithm suffers from the algorithmic monoculture is further based on the determined natural direct effect, the natural indirect effect, and the spurious effect”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 8
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“comprising:determining whether the natural direct effect is non-zero; and determining an evidence of a disparate treatment based on the determination that the natural direct effect is non-zero.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 9
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“determining whether the natural direct effect is zero; determining whether the natural indirect effect is non-zero; and determining an evidence of a disparate treatment based on the determination that the natural direct effect is zero and the natural indirect effect is non-zero.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 10
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is zero; determining whether the spurious effect is non-zero; and determining an evidence of a disparate treatment based on the determination that the natural direct effect is zero, the natural indirect effect is zero, and the spurious effect is non-zero.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 11
Step 1: The claim recites a method; therefore, it falls into the statutory category of processes.
Step 2A Prong 1: The claim recites multiple mental processes, as explained below. The claim recites, inter alia:
“further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is zero; determining whether the spurious effect is zero; and determining that the computing algorithm is associated with no evidence of a disparate treatment or no evidence of a disparate outcome, based on the determination that the natural direct effect is zero, the natural indirect effect is zero, and the spurious effect is zero.”
This limitation is directed to the abstract idea of a mental process (concepts performed in the human mind, including observation and evaluation [see MPEP 2106.04(a)(2) III. C.]).
Thus, the judicial exception is not integrated into a practical application [see MPEP 2106.05(d) I.], failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claims 12-19
Claims 12-19 recites analogous limitations to claims 1-8 and therefore is rejected on the same ground as claims 1-8.
Regarding claims 20
Claim 20 recites analogous limitations to claim 1 and therefore is rejected on the same ground as claims 1.
Claim Rejections - 35 USC § 103
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 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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-4, 6-9, 12-15, and 17-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kaddour et al. (“Causal Machine Learning: A Survey and Open Problems”) in view of Bommasani et al. (“Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?”) and further in view of Himan et al. (“Controlling Popularity Bias in Learning-to-Rank Recommendation”).
Regarding claim 1
Kaddour teaches a method, executed by a processor, (Examiner notes that it is well understood in the art for machine learning model to be executed on computer with processor see training on pg. 2 “To link the practice of sharing in ML with the proposed harm of homogenization, we pose and test the component sharing hypothesis: algorithmic systems built using the same underlying components, such as training data and machine learning models, will tend to systematically fail for the same individuals or groups”)
comprising: receiving a dataset associated with a domain; encoding domain knowledge information associated with the received dataset; (see input data being received Algorithm 1 “Input: Training set D over {(X, Y )}”)
constructing, for a computing algorithm related to the domain, (section 3.1.1.2 “In this setting, we have access to samples of X and pre-trained representations R. Mao et al. [47] propose a method (see Algorithm 1) that improves how the pre-trained representations, learned by self-supervised learning or otherwise, are leveraged for classification models. They assume that the data is generated by the causal structure shown in Fig. 3.4, with X ∼ p(x | c, s) generated by style and content features, and R the representation of X that a pre-trained model produces.”)
a structural causal model (SCM) associated with the dataset based on the encoded domain knowledge information; (pg. 12 “In SCMs, we express causal relationships through deterministic, functional equations. This formalism reflects Laplace’s conception of natural laws being deterministic and randomness being a purely epistemic notion [12]. Hence, we introduce stochasticity in SCMs based on the assumption that certain variables in the equations remain unobserved”)
identifying a mediator variable and a confounder variable associated with the computing algorithm, based on the constructed SCM; (section 2.5 “In real-world scenarios, the patient’s pre-treatment health conditions xi influence both the doctor’s treatment prescription and outcome, thereby X confounds the effect of the treatment T on the outcome Y (and we call X a confounder or confounding variable). Moreover, we say that T and Y are confounded (by X), or spuriously associated.”)
estimating a causal effect associated with the computing algorithm, based on the identified mediator variable and the identified confounder variable; (section 2.5 “Fig. 2.3 visualizes how associations flow in the observational p(y | x, t) and interventional distribution p(y | x, do(t)). In Fig. 2.3a, we see that p(y | x, t) entails both causal and spurious associations from T to Y, while p(y | x, do(t)) isolates the causal association from T to Y , as shown in Fig. 2.3b. The causal effect flows along directed paths, while spurious associations flow along all unblocked paths.”)
Kaddour does not teach determining whether the computing algorithm suffers from an algorithmic monoculture, based on the estimated causal effect, to detect popularity bias bias in the computing algorithm;…
and rendering information indicative of whether the computing algorithm suffers from the algorithmic monoculture.
Bommasani teaches determining whether the computing algorithm suffers from an algorithmic monoculture, based on the estimated causal effect, to detect bias in the computing algorithm; (section 2 “To illustrate outcome homogenization and its potential causes (including algorithmic monoculture), we will use the example of algorithmic resume screening. Companies use resumes to screen job applicants, choosing which candidates to interview and which to reject. Maximum homogenization occurs when every company makes the same decision about each candidate, such that each lucky candidate is interviewed by all companies and each unlucky candidate by no companies. We say that the unlucky candidates who receive no interviews experience a systemic failure”)
and rendering information indicative of whether the computing algorithm suffers from the algorithmic monoculture. (Pg. 4 “Homogenization metric for individuals. SYSTEMIC FAILURE quantifies homogeneous outcomes, but is difficult to compare across systems with different underlying accuracies: SYSTEMIC FAILURE will in general be higher for less accurate systems independent of a specific tendency to pick (i.e. fail) on the same person. While we may sometimes want to combine accuracy and outcome homogenization into an overall measure of utility or social welfare, which SYSTEMIC FAILURE(h1,...,hk) implicitly does, we focus on a relative measure of homogenization that disentangles accuracy from homogenization. In particular, we are interested in outcome homogenization even, and perhaps especially, in systems that are highly accurate.”)
Kaddour and Bommasani are analogous art because they are both directed to machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combined causal machine learning of Kaddour with monoculture sharing hypothesis of Bommasani.
One of ordinary skill in the art would have been motivated to make this modification in order to improve “societal challenges that inhibit the measurement, diagnosis, and rectification of outcome homogenization in deployed machine learning systems” as disclosed by (Bommasani abstract “We test this hypothesis on algorithmic fairness benchmarks, demonstrating that increased data-sharing reliably exacerbates homogenization and individual-level effects generally exceed group-level effects. Further, given the current regime in AI of foundation models, i.e., pretrained models that can be adapted to myriad downstream tasks, we test whether model-sharing homogenizes outcomes across tasks. We observe mixed results: we find that for both vision and language settings, the specific methods for adapting a foundation model significantly influence the degree of outcome homogenization. We also identify societal challenges that inhibit the measurement, diagnosis, and rectification of outcome homogenization in deployed machine learning systems.”).
Kaddour in view of Bommasani does not teach popularity bias.
Himan teaches popularity bias. (Section 3 “Following a similar approach, we explore the use of regulariza tion to control the popularity bias of a recommender system. We start with an optimization objective of the form… where acc(.) is the accuracy objective, reд(.) is a regularization term, and λ is a coe cient for controlling the e ect of regularizer.”)
Regarding claim 2
Kaddour in view of Bommasani with Himan teaches the method according to claim 1.
Kaddour further teaches the method further comprising: determining information associated with an observed variable and an unobserved variable associated with the domain; (pg. 77 “Consequently, CWMs infer the interventional query “Given that we have observed x t:T = x t:T in the real world, what is the probability that x t+1:T would have been x t+1:T 0 I if x t were x t I in the dream world?” Technically, they want to intervene upon the abstract state variable as shown in Fig. 7.2c, where s t I ∈ S is the counterfactual value in the dream environment. This intervention is then rendered as an observable change applied to x t (such as, for instance, object displacement or removal) by the conditional observation distribution U x t = x t I | do s t = s t I …where x t I ∈ O represents the value of the counterfactual observation.”… section 6.1 “Fig. 6.1a illustrates a setting where only latent confounder U causes Y. Suppose, for example, a car insurance company wants to price insurance for car owners by predicting their accident rates Y, assuming there is an unobserved variable corresponding to aggressive driving U which (a) increases the likelihood that drivers will have an accident and (b) increases the likelihood of individuals preferring red cars, captured by X. Furthermore, people with specific demographic characteristics may prefer driving red cars. Using the red car feature X to predict the accident rate Y seems unfair, because these individuals are no more likely than anyone else to be aggressive or to get into accidents.”)
and identifying an input variable and an output variable associated with a problem area associated with the domain, (section 3.1.2.4 “Compositional recognition is the problem of learning to recognize new combinations of known components. Note that this method is trained on data with both attribute A and object O labels, as opposed to additional environment labels, which the previous methods in Sec. 3.1.2 exploited. Atzmon et al. [78] argue that deep discriminative models fail at compositional recognition because of two reasons: (i) distribution-shift and (ii) entanglement of representations.”)
wherein the encoding of the domain knowledge information is further based on the determined information associated with the observed variable and the unobserved variable and the identified input variable and the output variable. (Section 6.1 “Fig. 6.1a illustrates a setting where only latent confounder U causes Y. Suppose, for example, a car insurance company wants to price insurance for car owners by predicting their accident rates Y, assuming there is an unobserved variable corresponding to aggressive driving U which (a) increases the likelihood that drivers will have an accident and (b) increases the likelihood of individuals preferring red cars, captured by X. Furthermore, people with specific demographic characteristics may prefer driving red cars. Using the red car feature X to predict the accident rate Y seems unfair, because these individuals are no more likely than anyone else to be aggressive or to get into accidents.”)
Regarding claim 3
Kaddour in view of Bommasani with Himan teaches the method according to claim 1.
Kaddour further teaches wherein the SCM is constructed based on at least one of the computing algorithm associated with the SCM, (pg. 12 “In SCMs, we express causal relationships through deterministic, functional equations. This formalism reflects Laplace’s conception of natural laws being deterministic and randomness being a purely epistemic notion [12]. Hence, we introduce stochasticity in SCMs based on the assumption that certain variables in the equations remain unobserved.”)
a number of citations of the computing algorithm, a confounder variable associated with the computing algorithm, (section 2.5 “where we have a dataset where each observation (xi, ti, yi) ∈ D represents a hospital patient’s medical history record xi, prescribed drug treatment ti, and health outcome yi. In real-world scenarios, the patient’s pre-treatment health conditions xi influence both the doctor’s treatment prescription and outcome, thereby X confounds the effect of the treatment T on the outcome Y (and we call X a confounder or confounding variable). Moreover, we say that T and Y are confounded (by X), or spuriously associated.”)
a probability that the computing algorithm is used, an outcome of the computing algorithm, or a familiarity of the computing algorithm. (Pg. 67 “We interpret the final expression of Eq. (6.1) as the probability that Yˆ predicts y for a given individual for which we observe features x and protected attribute a, had the protected attribute been a 0 instead of a. An analogous interpretation of the first expression asserts that it equals p(y | x, a).” also see pg. 80 “Consequently, CWMs infer the interventional query “Given that we have observed x t:T = x t:T in the real world, what is the probability that x t+1:T would have been x t+1:T 0 I if x t were x t I in the dream world?” Technically, they want to intervene upon the abstract state variable as shown in Fig. 7.2c, where s t I ∈ S is the counterfactual value in the dream environment.”)
Regarding claim 4
Kaddour in view of Bommasani with Himan teaches the method according to claim 3.
Kaddour further teaches wherein the SCM corresponds to a directed acyclic graph that indicates interdependencies in variables associated with the computing algorithm, based on the encoded domain knowledge information. (Pg. 5 “A directed cycle is a directed path that starts from a node A and ends in A. A directed acyclic graph (DAG) is a directed graph with no directed cycles. In a DAG, edges point from a parent node into a child node. We denote the parents of a node X with pa(X); and X an ancestor of Y (denoted by X ∈ an(Y )), and Y a descendant of X (denoted by Y ∈ de(X)) if there is a directed path that starts at node X and ends at node Y. We denote a random node variable as X, assume that all distributions possess a mass or density function, and write p(x) to represent its distribution.”)
Regarding claim 6
Kaddour in view of Bommasani with Himan teaches the method according to claim 1.
Kaddour further teaches wherein the mediator variable is a variable that affects an input variable associated with the computing algorithm, (section 6.2.3 “We denote all observed mediator variables between A and Y as W. Then, the direct effect (DE), indirect effect (IE), and spurious effect (SE)”)
and the confounder variable is a variable that affects the input variable and an output variable associated with the computing algorithm. (Section 3.1.1.4 “A variety of methods [61, 62, 63] propose to remove the influence of unobserved confounders U from predictions through counterfactual regularization, after model training has ended. This involves estimating the confounding effect of U on the prediction Y˜ and then subtracting it, thus deconfounding the prediction. For a prediction over sample x, x 0 is generated such that it carries none of the causal information in x.”)
Regarding claim 7
Kaddour in view of Bommasani with Himan teaches the method according to claim 1.
Kaddour further teaches the method further comprising: determining a natural direct effect, a natural indirect effect, and a spurious effect based on the estimated causal effect, (section 8.2 see FIG. 8.14 “Figure 8.14: Investigating Gender Bias using Mediation Analysis [368]. Given a prompt u such as “The nurse said that”, we ask a language model to generate a continuation. A biased model may assign a higher likelihood to she than to he. To understand the role of model components on this biased prediction, we perform the do-operation x = set-gender, which changes u from nurse to man in this example. By inferring direct and indirect effects, we can analyze the causal role of specific mediators (neurons) between x and y.”)
Bommasani further teaches wherein the determination of whether the computing algorithm suffers from the algorithmic monoculture… (section 2 “To illustrate outcome homogenization and its potential causes (including algorithmic monoculture), we will use the example of algorithmic resume screening. Companies use resumes to screen job applicants, choosing which candidates to interview and which to reject. Maximum homogenization occurs when every company makes the same decision about each candidate, such that each lucky candidate is interviewed by all companies and each unlucky candidate by no companies. We say that the unlucky candidates who receive no interviews experience a systemic failure”)
Regarding claim 8
Kaddour in view of Bommasani with Himan teaches the method according to claim 7.
Kaddour further teaches the method further comprising: determining whether the natural direct effect is non-zero; and determining an evidence of a disparate treatment based on the determination that the natural direct effect is non-zero. (Section 5.2 “Since we introduce several metrics, we measure the correlations between our metrics. Further, we measure correlations with accuracy (specifically, the expected rate of systemic failure) to test if homogenization is disentangled from accuracy. Since outcome homogenization is related to fairness, we also measure the correlation between our metrics and a standard group fairness metric. Fairness metrics are generally defined for a single model h, whereas we study entire systems {h i} k i=1. We extend the unfairness definition used by Khani et al. [2019] as the variance in the systemic failure rates across groups.” Also see section 6.2.3 “First, the authors assume that there is a disadvantaged group a1 and an advantaged one, a0. Further, W denotes all observed intermediate variables between A and Y. Next, they define different effects; if the effect is non-zero, the predictor is unfair. They distinguish between direct and indirect spurious effects; the latter considers the back-door paths between A and Y, that is, paths with an arrow into A”)
Regarding claim 9
Kaddour in view of Bommasani with Himan teaches the method according to claim 7.
Kaddour further teaches the method further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is non-zero; (Section 6.2.3 “First, the authors assume that there is a disadvantaged group a1 and an advantaged one, a0. Further, W denotes all observed intermediate variables between A and Y. Next, they define different effects; if the effect is non-zero, the predictor is unfair. They distinguish between direct and indirect spurious effects; the latter considers the back-door paths between A and Y, that is, paths with an arrow into A”)
and determining an evidence of a disparate treatment based on the determination that the natural direct effect is zero and the natural indirect effect is non-zero. (Section 6.2.3 “First, the authors assume that there is a disadvantaged group a1 and an advantaged one, a0. Further, W denotes all observed intermediate variables between A and Y. Next, they define different effects; if the effect is non-zero, the predictor is unfair. They distinguish between direct and indirect spurious effects; the latter considers the back-door paths between A and Y, that is, paths with an arrow into A”)
Regarding claims 12-15 and 17-19
Claims 12-15 and 17-19 recites analogous limitations to claims 1-4 and 6-8 therefore is rejected on the same ground as claims 1-4 and 6-8.
Regarding claims 20
Claim 20 recites analogous limitations to claim 1 and therefore is rejected on the same ground as claims 1.
Claim(s) 5, 10-11 and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kaddour et al. (“Causal Machine Learning: A Survey and Open Problems”) in view of Bommasani et al. (“Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?”) in view of Himan et al. and further in view of Makhlouf et al. (“Survey on Causal-based Machine Learning Fairness Notions”).
Regarding claim 5
Kaddour in view of Bommasani with Himan teaches the method according to claim 3.
Kaddour in view of Bommasani with Himan does not teach the method further comprising: determining one or more variables associated with a disparate outcome or a disparate treatment associated with the computing algorithm; computing a direct effect, an indirect effect, and a spurious effect of each of the determined one or more variables based on the SCM; and establishing an evidence of the disparate outcome or the disparate treatment, based on the computed direct effect, indirect effect, and the spurious effect.
Makhlouf further teaches the method further comprising: determining one or more variables associated with a disparate outcome or a disparate treatment associated with the computing algorithm; (section 4.1 “No unresolved discrimination [22] is a fairness notion that falls into the disparate treatment framework and focuses on the indirect causal effects from…. A resolving variable is any variable in a causal graph that is influenced by the sensitive attribute in a manner that is accepted as nondiscriminatory. Figure 4 presents two alternative causal graphs for the job hiring example. The graph at the left exhibits unresolved discrimination along the heavy paths: … By contrast, the graph at the right does not exhibit any unresolved discrimination as the effect of …justified by the resolved variable.”)
computing a direct effect, an indirect effect, and a spurious effect of each of the determined one or more variables based on the SCM; (pg. 5 left col “the different paths of causal effects (direct, indirect, and spurious) is much easier achieved using SCMs and causal graphs. More generally, potential outcome framework is more suitable for causal inference problems where the goal is to narrowly estimate the causal (treatment) effect of a cause variable on an outcome variable. There are two justifications for this point. First, developing estimators of causal effects and counterfactuals can be more straightforward using the potential outcome framework [59]. Second, the potential outcome framework provides the possibility of decomposing the sources of inconsistency and bias into: unaccounted for baseline differences between individuals and treatment effect bias [28] (Section 3.4)”)
and establishing an evidence of the disparate outcome or the disparate treatment, based on the computed direct effect, indirect effect, and the spurious effect. (Section 4.4 “By conditioning on the sensitive attribute… defined two variants of …Eq. 14) and …(Eq. 15) which focus on the direct and indirect effect for a specific group. In addition, they characterize a third type of effect, spurious, which considers the back-door paths between …that is, paths with an arrow... Using Table 4… which indicates a spurious effect in favor of female. Compared to …counterfactual effects focus only on individuals of a specific group (e.g. only female candidates) and characterize the causal effect through spurious (back-door) paths. This spurious effect is what makes causal relations different from mere statistical correlation”)
Kaddour, Bommasani, Himan and Makhlouf are analogous art because they are all directed to machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combined causal machine learning of Kaddour in view of Bommasani with Himan with survey on causal based machine learning fairness notions of Makhlouf.
One of ordinary skill in the art would have been motivated to make this modification in order to “help selecting a suitable fairness notion given a specific real world scenarios” as disclosed by (Makhlouf abstract “This paper examines an exhaustive list of causal-based fairness notions and study their applicability in real-world scenarios. As the majority of causal-based fairness notions are defined in terms of non-observable quantities (e.g., interventions and counterfactuals), their deployment in practice requires to compute or estimate those quantities using observational data. This paper offers a comprehensive report of the different approaches to infer causal quantities from observational data including identifiability (Pearl’s SCM framework) and estimation (potential outcome framework). The main contributions of this survey paper are (1) a guideline to help selecting a suitable fairness notion given a specific real world scenario, and (2) a ranking of the fairness notions according to Pearl’s causation ladder indicating how difficult it is to deploy each notion in practice.”).
Regarding claim 10
Kaddour in view of Bommasani with Himan teaches the method according to claim 7.
Kaddour further teaches determining whether the spurious effect is non-zero; and determining an evidence of a disparate treatment based on the determination that the natural direct effect is zero, the natural indirect effect is zero, and the spurious effect is non-zero. (Section 6.2.3 “First, the authors assume that there is a disadvantaged group a1 and an advantaged one, a0. Further, W denotes all observed intermediate variables between A and Y. Next, they define different effects; if the effect is non-zero, the predictor is unfair. They distinguish between direct and indirect spurious effects; the latter considers the back-door paths between A and Y, that is, paths with an arrow into A”)
Kaddour in view of Bommasani with Himan does not teach the method further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is zero;
Makhlouf teaches the method further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is zero; (pg. 12 left col “P(ya) = P(y) provided that there are no causal paths between A and Y”)
Kaddour, Bommasani, Himan and Makhlouf are analogous art because they are all directed to machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combined causal machine learning of Kaddour in view of Bommasani with Himan with survey on causal based machine learning fairness notions of Makhlouf.
One of ordinary skill in the art would have been motivated to make this modification in order to “help selecting a suitable fairness notion given a specific real world scenarios” as disclosed by (Makhlouf abstract “This paper examines an exhaustive list of causal-based fairness notions and study their applicability in real-world scenarios. As the majority of causal-based fairness notions are defined in terms of non-observable quantities (e.g., interventions and counterfactuals), their deployment in practice requires to compute or estimate those quantities using observational data. This paper offers a comprehensive report of the different approaches to infer causal quantities from observational data including identifiability (Pearl’s SCM framework) and estimation (potential outcome framework). The main contributions of this survey paper are (1) a guideline to help selecting a suitable fairness notion given a specific real world scenario, and (2) a ranking of the fairness notions according to Pearl’s causation ladder indicating how difficult it is to deploy each notion in practice.”).
Regarding claim 11
Kaddour in view of Bommasani with Himan teaches the method according to claim 7.
Kaddour in view of Bommasani with Himan does not teach the method further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is zero; determining whether the spurious effect is zero; and determining that the computing algorithm is associated with no evidence of a disparate treatment or no evidence of a disparate outcome, based on the determination that the natural direct effect is zero, the natural indirect effect is zero, and the spurious effect is zero.
Makhlouf teaches the method further comprising: determining whether the natural direct effect is zero; determining whether the natural indirect effect is zero; determining whether the spurious effect is zero; (section 4.1 “No unresolved discrimination is satisfied when no directed path from 퐴 to 푌 is allowed, except via a resolving (explaining) variable 퐸. A resolving variable is any variable in a causal graph that is influenced by the sensitive attribute in a manner that is accepted as nondiscriminatory. Figure 4 presents two alternative causal graphs for the job hiring example. The graph at the left exhibits unresolved discrimination along the heavy paths:”)
and determining that the computing algorithm is associated with no evidence of a disparate treatment or no evidence of a disparate outcome, based on the determination that the natural direct effect is zero, the natural indirect effect is zero, and the spurious effect is zero. (section 2 pg. 2 right col “As the values are equal, the rates of firing between teachers who were assigned low level students and high level students appear to be equal and hence no discrimination is detected.”)
Kaddour, Bommasani, Himan and Makhlouf are analogous art because they are all directed to machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combined causal machine learning of Kaddour in view of Bommasani with Himan with survey on causal based machine learning fairness notions of Makhlouf.
One of ordinary skill in the art would have been motivated to make this modification in order to “help selecting a suitable fairness notion given a specific real world scenarios” as disclosed by (Makhlouf abstract “This paper examines an exhaustive list of causal-based fairness notions and study their applicability in real-world scenarios. As the majority of causal-based fairness notions are defined in terms of non-observable quantities (e.g., interventions and counterfactuals), their deployment in practice requires to compute or estimate those quantities using observational data. This paper offers a comprehensive report of the different approaches to infer causal quantities from observational data including identifiability (Pearl’s SCM framework) and estimation (potential outcome framework). The main contributions of this survey paper are (1) a guideline to help selecting a suitable fairness notion given a specific real world scenario, and (2) a ranking of the fairness notions according to Pearl’s causation ladder indicating how difficult it is to deploy each notion in practice.”).
Regarding claims 16
Claim 16 recites analogous limitations to claim 5 and therefore is rejected on the same ground as claims 5.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to VAN C MANG whose telephone number is (571)270-7598. The examiner can normally be reached Mon - Fri 8:00-5:00pm.
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, David Yi can be reached at 5712707519. 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.
/VAN C MANG/Primary Examiner, Art Unit 2126
Prosecution Insights