Under the Hired Brains Advisory brand, Hired Brains will provide advisory services to chief executives, company boards and management teams on the design and development of frameworks and toolkits for ethical artificial intelligence. In addition, the firm will offer education, coaching and mentoring services related to the development of team or individual’s artificial intelligence capabilities.

We address these issues with focused consulting deliverables such as data adequacy, strategy formation research, media, and events. Our services are designed to avoid costly mistakes, uncover hidden risks, inform your team and share best practices with the Hired Brains Ethical AI Community of Interest.

Ethical Challenges Faced by AI

How does this promising technology get it so wrong sometimes? Surprising elements of bias, invasion of privacy, discrimination, breach of regulatory requirements and professional precepts are common. Judicial software biased against people of color, hiring models that favor only males, “chatbots” that are manipulated to speared vile hateful dialogue in a matter of minutes? The answer is actually quite simple. Machine learning models can learn, but they are not very smart. People are much smarter and there are preventative measures and remedies.

Machine learning is based on the concept that large enough volumes of data will yield inferencing algorithms that can then be applied to new data as it arrives. But at its core, it’s just curve fitting. The algorithm tries to fit variables to the desired objective. But if the developer is not careful or skilled enough with the features selected, or the data itself isn’t what it appears to be, things can go haywire. In its attempt to converge on an objective function, a poorly trained model will diverge from its designed features and search for certain “latent vales” that allow it to converge in often bizarre ways. Transparency is weak and when this happens in production, it can fire thousands or millions of times before it is noticed

Until recently, decision-making systems in insurance companies were composed of conventional development techniques and platforms. The scope of the systems ranged from rules-based scoring for underwriting to claims adjudication to valuation models. Often, rules were encoded in logic elucidated so they could be inspected at any time If errant results were detected, code traces, transaction logs and persistent data in databases could be investigated.

No organization is immune. But it doesn’t have to be this way.

Solutions

Insurance companies are in the greatest danger from mishaps with AI. Unlike other industries, everything an insurance company does affects people and is potentially subject to ethical breeches that are often not detected until the damage is evident. They arise from naïve preparation of data, ad hoc development, the team's level of skill, the quality of the data, and a variety of biased outcomes, and, Do-It-Yourself AI becoming more popular.

We offer solutions to the potentially extreme and insidious damage that can be caused by ethical breaches of errant AI and, unfortunately, those who exploit AI.