BigBlueBot: teaching strategies for successful human-agent interactions
summary
A web experience called BeTheBot/BigBlueBot teaches people to simplify queries and recover from breakdowns by making them play two roles in scripted banking and shopping conversations using fixed-response and free-response turns plus an egregiousness classifier. Pre-study log analysis uses Watson Assistant data. Mechanical Turk evaluation ( N=88 , 15–20 minute sessions) finds 83% simplify as the optimal strategy, ~70% recall multiple breakdown types, and roughly three quarters express desire to use chatbots.
AI-written summary · about
chapters (6)
- 0:00Introduction and motivation
- 2:13Teaching better chatbot interactions
- 4:43Online training experience design
- 8:21Study evaluation and results
- 11:59User feelings and empathy toward chatbots
- 13:49Q&A on study purpose and future work
related talks
14mTrustworthy Robots | Building Customer Relationships in Automated Conversational Experiences
35mDesigning Experiences for Chatbots
22mAll Work and No Play? Conversations with a Question-and-Answer Chatbot in the Wild
45mThe Humans Are Dead, We Are All Bots Now
56mBots - the next UI revolution
35mChat Bot Building Story - The Good, the Bad and the Unexpected
10mEnglishBot: An AI-Powered Conversational System for Second Language Learning
01h 54mBuild your own Generative AI Chatbot with Watsonx Assistant