Tackling Bias in

Machine Learning & Artificial Intelligence

Artificial intelligence (AI) and machine learning (ML) algorithms are used in the decision-making process in many industries.  Nevertheless, AI/ML algorithms have a well-documented history of bias, which fuels systemic or proxy discrimination that can result in disparate impact or treatment of protected classes of people.  Furthermore, their complexity makes understanding and deconstructing their functionality challenging even when full transparency is provided.

The insurance industry aims to classify policyholders into risk-based categories by building actuarial models to relate policyholder characteristics to claim risk that are used to predict the rates at which policyholders will generate claims. Claim rate predictions and amounts (severity) help determine whether coverage is offered to prospective policyholders as well the amount to charge as the insurance premiums.  However, as big data and AI are increasingly being deployed by insurance companies to help classify individuals, bias inherent in AI/ML systems exacerbate the risk of disproportionate harm to members of a protected class.

Auto insurers are known to charge higher average premiums to drivers living in predominantly minority urban neighborhoods than to drivers with similar safety records living in majority white neighborhoods. Such rate disparities are defended as being dependent upon the risk of accidents being greater in those neighborhoods even for motorists that have never had an accident. To determine the degree to which algorithmic bias may exist in AI/ML risk models that are used to set insurance rates, a rigorous examination of an end-to-end data and ML pipeline will be undertaken using auto insurance data to identify different types of biases that may be present.  Furthermore, mitigation strategies including the use of “fair” ML models will be evaluated to determine the most optimal approach in avoiding disparate impact.  Ultimately, an AI algorithmic bias detector will be created and deployed using a suite of data science tools to enable real-time scoring of insurance risk models used to generate insurance premiums. 

Cohort A     I      Cohort B      I      Cohort C

    Tu/Th                    Wed                      Sat

11AM-1PM      6PM-10PM          8AM-12PM

    4 hours per week of live online instruction

LEARNING OUTCOMES

Understanding how ML  & AI are used in the insurance industry

Understanding bias in AI and how proxy discrimination, disparate impact, etc. can be fueled by big data

Examining different ways to eliminate bias in AI/ML systems and why most fail

Why biased AI/ML = BIG  Business + BIG bucks

Designing and developing an automated AI bias detector 

Validating and deploying automated data ingestion and ML pipelines

Creating synthetic datasets, and much more....

OUR FIRM

Artificial Intelligence.  Automation.

Intellectual Property.  

 Machine Learning. Deep Learning.

Big Data.  Analytics.

Chi's Research Corporation focuses on creating, protecting and monetizing products.

Our firm offers services in artificial intelligence, machine learning and intellectual property.

 

We also offer on-site training courses carefully curated to meet rising demands for artificial intelligence skills.

Contact us today to find out more on how to create, brand, protect and monetize AI assets.