Trust is Not a Marketing Issue
Recently, I had a conversation with a very smart person who was quite enthusiastic about OpenAI's efforts to incorporate friendly voices and a more human style of interaction into their AI systems. He believed that this approach would lead to increased interaction and usage, and subsequently, a greater tendency for people to trust these systems.
However, this excitement overlooks a significant danger. The way conversational systems are marketed—with friendly voices, apologetic tones when mistakes are made, and overall persuasive and convincing behavior—should not replace genuine trustworthiness. It’s essential to understand that these features are designed to make the system more user-friendly, but they risk sacrificing true reliability and accuracy.
There is a difference between making AI systems seem trustworthy and ensuring they are genuinely trustworthy. Trust should stem from the accuracy and honesty of the information conveyed, not from the system's likability or charm. This distinction is crucial because a system designed to be lovable and engaging can easily manipulate users. We must ask ourselves why trustworthiness is being turned into a marketing ploy rather than an issue of the integrity of the information provided. This trend is perilously misleading.
We want to trust AI systems because they are supposed to work in our best interests, provide truthful information, and align with our goals. Trust should not be based on superficial qualities like a sweet voice or appealing demeanor.
While making AI systems more user-friendly and engaging might seem beneficial, it is imperative that we remember that there is a difference between charming and trustworthiness. We must demand that these systems prioritize accuracy and integrity over charm and persuasion. By doing so, we can ensure that AI serves our best interests without leading us astray through deceptive practices.
Kristian Hammond
Bill and Cathy Osborn Professor of Computer Science
Director of the Center for Advancing Safety of Machine Intelligence (CASMI)
Director of the Master of Science in Artificial Intelligence (MSAI) Program