首页|Patent Issued for Detecting and reducing bias (including discrimination) in an a utomated decision making process (USPTO 11922435)
Patent Issued for Detecting and reducing bias (including discrimination) in an a utomated decision making process (USPTO 11922435)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
News editors obtained the following quote from the background information suppli ed by the inventors: “Field of the Invention“The invention relates to machine learning and more particularly to using machin e learning to detect and reduce bias in an automated decision making process, su ch as, for example, whether to provide a particular amount of credit to an indiv idual. “Description of the Related Art “Corporations, government entities, and the like may continually make decisions that take into account a large number of factors. For example, a university may base the decision on whether to offer admission to a student on factors such as the student's grades at another educational institution, the student's extracur ricular activities, the age of the student, the race of the student, social-econ omic background of the student, and the like. As another example, a lender, such as a bank, may base a decision on whether to extend a particular amount of cred it (e.g., mortgage, car loan, etc.) to a person or a business based on a credit score, prior credit history, income, where the person lives, and the like. Such decisions may include some form of bias (e.g., implicit bias) such that, even if an institution considers their decision-making process to be impartial, the res ults may in fact be biased. Thus, bias can creep into the rules used for approva l or denial of an application, such as a credit application, an education instit ution admissions application, and the like, and the bias may be difficult to det ect. Bias refers to making a decision using a decision-making process where the decision favors (or disfavors) certain traits or characteristics, such as gender, age, ethnicity, geographic location, income, and the like. For example, women typically earn significantly less (e.g., approximately 1/4 to 1/3 less) than the ir male counterparts (e.g., in similar jobs and with similar educational backgro unds and experience). When a decision-making process takes annual income into co nsideration, the resulting decisions may have an implicit bias for males and aga inst females. In some cases, the bias may be sufficiently significant that the b ias is discriminatory and therefore illegal.
BusinessCerebri Ai Inc.CyborgsEmer ging TechnologiesMachine Learning
NETL
NSTL
