Responsible AI

Lapetus Solutions is an AI company focused on developing de novo AI solutions for life event spaces–life insurance, financial services, annuity risk management, and more—while maintaining a commitment to the development of technologies that make the world better by creating solutions that works for everyone. At Lapetus we do not boil the AI ocean, instead we develop very specific AI solutions that focus on data that our scientists have been researching for more than 20 years, e.g. national health data sets and more.
Lapetus has from its inception, April 2015, has engaged with insurance regulators from around the US to discuss how AI could be used responsibly with in the life insurance space. We have developed our model for Responsible AI based on the feedback from regulators in New York, Florida, Maryland, Utah, and many more states. This two-way exchange has led to the Total Transparency Policy which demystifies the “black-box” of our AI technology. We see the users of our technology as partners not clients, we share critical insights on data used to develop our AI solutions including demographics. All of our AI technologies are built as discrete solutions allowing our partners (clients) and regulators to critical examine the inputs and outputs.

At Lapetus Solutions we follow the recommended best practices for AI development and deployment from Google, Microsoft, Amazon, and others on Fairness, Interpretability, Privacy, and Security.

Fairness

Interpretability

Privacy

Security

Lapetus Solutions Follows Best Practices

Fairness

  • Models are designed with concrete goals for fairness and inclusion
  • Models are built (train and tested) with representative datasets
  • Models are evaluated for bias (differential performance) across all critical covariates, subgroup analysis

Interpretability

  • Work with our partners (clients) to understand interpretability needs
  • Build models to be interpretable
  • Choose metrics to ensure the end goal and desired performance across subgroups
  • Understand the trained model
  • Test models often especially in the deployment phase

Privacy

  • Responsible collection and handling of all data
  • Secure and safeguard the model(s)

Security

  • Identify threats to system
  • Combat threats with practical countermeasures
  • Stay abreast of new threat vectors through