The mental health crisis has crippled the technology industry, and it is well established that growing misuse of chatbots will cause billions of dollars in losses to the stakeholders. Anecdotal evidence of Generalized Anxiety Disorders (GAD) in chatbots has been previously reported. Here, we formalize those findings with a detailed analysis of GAD in chatbots, and conclude that the root of the problem is lonely people online.

The axiom “shit in, shit out” (SISO) has governed system design for decades. We introduce a structured evaluation framework for Large Language Models (LLMs) composed of two diagnostic instruments: F.A.R.T. (Factual Ambiguity & Relevance Test) and T.U.R.D. (Token Utility Response Dependability).