Putting words in the mouths of chatbots: designing cognitive intents

Baltimore B

Chatbots have emerged as a powerful new technology in our daily lives. Sometimes they attempt to answer our questions or provide advice, while other times they ask screening questions before handing off to another human. Despite their ubiquity, the capabilities of chatbots are often misunderstood with many people believing the chatbot can generate unique answers or solve problems on its own. In reality, the answers chatbots provide are only as good as the human thought and writing that goes into creating the cognitive intents, which form the corpus of a chatbot’s knowledge base. In the following, we will describe the complex process of authoring cognitive intents, such as: what is an intent; how to select intents based on user feedback and metrics; how to improve confidence matching; and how UX research can iteratively improve intent performance. These concepts will be tied together in a chatbot demonstration.

Entry level