Strategies for Conversational Product Experiences
summary
Argues that conversation-driven products will outpace mobile, differentiates one-shot vs multi-turn interactions and names turns, intents, utterances, prompts and entities as the core mechanics; notes ML needs large, high-quality datasets and that repair happens roughly every 90 seconds. Gives concrete outcomes: an Alexa toast query fails with irrelevant recipes, Alexa enables a playable Jeopardy experience, and using a priming line like "tell me a few words about what's going on" materially increases task completion. Emphasizes multimodal Google Hub use and that missing context kills trust.
AI-written summary · about
chapters (13)
- 0:00Introduction to talk
- 1:46What conversation achieves
- 4:16Talking with AI today
- 10:28How AI processes speech
- 15:05Example of bad conversational AI
- 17:11The role of context in meaning
- 18:55Five strategies for conversational AI
- 21:45Maintenance and repair in conversation
- 24:18Human-like interaction cravings
- 27:09Power of everyday language
- 27:50Treating users with respect
- 29:15Language as a living tool
- 31:23Navigating the Uncanny Valley
related talks
45mKeynote: Conversations with Machines
50mSpeak To Me: Voice App Conversation Practices
54mConversational UX—Human Speech Patterns Meet UX
44mNavigating the Hurdles of Conversation Design
19mAlexa, show me a good voice UI
44mThe Future of the Conversational UIs
55mConversation Is the Next Big Interface
32mParty like it's 1970: Conversational interfaces are back (into your Android app)