When it comes to human vs. human-AI hybrid in answering knowledge demanding questions, who has the edge? This was the question iSchool PhD student Zihan Gao and Professor Jiepu Jiang asked in their research involving human-AI hybrid chat systems. The study, which was accepted as a full paper to appear at November’s CIKM 2021 Conference, showed that a mix of human and AI inputs in chats improved efficiency and quality when answering a wide range of questions. For online customer service settings, that means fewer keystrokes, faster responses, and the ability to answer questions without having to search for information first.
So what does this look like? Here’s an example from Gao: Let’s say a customer asked a customer support associate a question. The associate can reply using a chatbot response suggestion and make edits if needed. It helps when people lack the required knowledge to answer questions, as otherwise, they need to search online to collect relevant knowledge first, which is time-consuming:
Customer: I’d like to learn about the TV show Grey’s Anatomy!
Support: It’s an American drama centered around the medical field that premiered on ABC.
Customer: Ah, I see, and who’s the main character?
Support: … [Chatbot suggestion: The main characters include Miranda Bailey, Owen Hunt, Tom Koracick, Derek Shepherd, and Richard Webber.]
This study can be applied to many real-world situations. From email composition to academic writing to encyclopedia editing, future human-AI hybrid systems could support any number of text creation scenarios. Gao says, “The work has disclosed that people need knowledge support while composing texts, and we can automate this process to help them write more efficiently and effectively.”