Lost in Translation: Why We Don't Speak the Same Language as AI Yet

 Have you ever felt like your AI assistant doesn't get you? You ask for a playlist to unwind, and it plays upbeat dance music. You tell your smart speaker to remind you to buy milk, and it adds everything from cheese to yogurt to your grocery list. The reality is that communication between humans and AI could be better.

Here is the glitch in the system: while AI can be unique at predicting our preferences and completing tasks, it often struggles to grasp the nuances of human intention. On the other hand, we need a more comprehensive understanding of how these complex algorithms work, making it difficult to adjust our expectations.

The secret weapon for smoother human-AI interaction might lie in a well-known psychological concept called the Theory of Mind (Wang &Ashok, 2022). Our ability to understand others' thoughts, feelings, and intentions allows us to navigate conversations and social interactions effectively. 

Imagine conversing with a virtual a
ssistant or chatbot that can answer your questions and understand how you perceive them. Researchers are working on that future, and a new study shows it's closer than we might think!

Wang and Ashok (2022) introduce the framework for a concept called Mutual Theory of Mind (MToM) and state that the mutual theory of mind framework is a way to improve communication between humans and AI systems, and it emphasizes how humans and AI can develop a shared understanding of each other throughout a conversation.  The researcher gave their MToM framework a spin.

The Study: Chatting with Jill Watson

Wang and Ashok (2022) placed a question-answering AI named Jill Watson in an online student discussion forum. Over a 10-week semester, students interacted with Jill Watson, asking questions and getting answers. The researchers tracked how the students perceived Jill Watson through surveys.

The Future of AI Communication

The big takeaway?  Wang and Ashok (2022
) concluded that with their MToM AI, Jill could pick up on clues in our communication to understand how we perceive it.

This study paves the way for AI to understand our questions and requests
and how we communicate them. This can lead to more natural and productive interactions between humans and AI in the future.

References

Wang, Qiaosi & Goel, Ashok. (2022). Mutual Theory of Mind for Human-AI Communication. 10.48550/arXiv.2210.03842