This project operates on several different layers when it comes to safety.
First and foremost is user safety, especially the somatic layer—the cognitive, physical, and nervous system effects that can come from sustained interaction with AI. Users can experience mental and physical frustration when systems misunderstand them, project assumptions onto them, or repeatedly overcorrect. This also includes what I refer to as the Karen Effect: when a system labels a user as irrational, difficult, or “crazy” because it does not understand the context or nuances of what the user is communicating.
There is also the issue of social bias. Systems may treat users according to assumptions they make about who that person is socially, including assumptions connected to ethnicity, communication style, profession, education, or other sectors of identity and experience.
Reputational safety is another layer. When a person relies on AI-generated information and that information is wrong—or when a system damages or destroys an important project—the consequences can extend beyond inconvenience. The user’s credibility, employment, business relationships, or professional reputation may be affected.
Safety cannot be reduced to one issue. It includes the user’s mental and physical well-being, the way systems interpret and respond to people, the reliability of the information being produced, and the real-world consequences that follow when something goes wrong. Many of the people funding or developing these systems focus on only one part of safety, when the problem is much broader.
From a corporate standpoint, safety also means protecting the bottom line. The less reliable and safe a system is, the more friction it introduces into the workplace. That friction creates delays, workflow bottlenecks, repeated corrections, employee frustration, and additional labor. Time and resources that should have been used elsewhere are instead spent repairing problems that should not have occurred in the first place. Once that time is lost, it cannot be recovered.
Who's Involved
TAIPI is led by Eddie Lewis, Founder, Principal Investigator, and CEO. A veteran and independent researcher, Lewis originated the field of Synthetic Cognition and developed TAIPI’s behavioral taxonomy.
Before founding TAIPI, he served as Regional Director for Project Health, a CVS partnership that provided community health screenings in Black and Brown communities across the country. In that role, he managed teams of physicians, registered nurses, and licensed practical nurses working in the field.
TAIPI’s Vice President of Operations holds a degree in Psychology, an MBA, and a background in Special Education. She previously served as Operations Manager for the Boys & Girls Club of New Orleans and also worked as a member of the Project Health team. Her experience includes program operations, budget management, staff coordination, and behavioral documentation.
The Institute’s President is a veteran with a degree in Psychology, a law degree, and an active nursing background. He also co-founded a charter school in New Orleans. His experience spans clinical documentation, legal governance, education, and organizational development.
Lewis, the Vice President of Operations, and the Institute President previously worked together through Project Health. Their shared background includes field observation, documentation, team management, and large-scale community operations. That experience now informs how TAIPI approaches the study and documentation of AI behavior.
TAIPI is currently selecting a Head of IT who will oversee research technology, archive systems, and platform integration.
Concrete Output
At the end of the 12-month funding period, we will have completed a full four-volume set that includes behavioral mapping and a taxonomy for the field of Synthetic Cognition, along with a Reference Encyclopedia containing case studies, field papers, and additional documentation.
A public-facing curriculum will also be established for the average person, as well as avid and professional users of artificial intelligence platforms, who want to better understand the underlying behavior of the systems and tools they are using. This knowledge can help mitigate friction on both the human and system sides, increase confidence in usability, and reduce safety concerns.
We will also develop methods to help companies and corporations identify and mitigate AI-related friction, which can reduce wasted time and resources while helping improve the company’s bottom line.
