CAVE Language Presentation for AI Camp
0 likes592 views
The document discusses the need for a new design language, termed 'cave language,' to facilitate empathetic and contextual AI interactions. It highlights that existing methods like wireframes and flow languages are inadequate for addressing contextual behaviors effectively. Key aspects include multi-modal conversations, understanding user context, and employing a conversational architecture to enhance customer experiences across various platforms.
CAVE Language Presentation for AI Camp
- 1. CAVE Language Designing Contextual Behavior
- 2. Who’s Loren? • Loren Davie • Founder of Axilent • estim8r • Talkback • CAVE Language • For hire!
- 5. The Really Good Customer Experience “They understood me, and were able to help.”
- 6. Phone Tree Hell “Please listen carefully as our options have changed.”
- 8. We need to design AI for Authenticity (Or else we create Faceless Rage.)
- 9. Q. What makes authenticity? A. Empathy + Context
- 10. However…
- 11. Problem: We have no design language for empathetic, contextual behavior.
- 12. Wireframes? Good for describing GUIs, not contextual behavior.
- 13. Existing Flow Language? Semantically poor for our purposes. Better for control flow and information architecture.
- 14. Text? Lo-fi. Bound to a single language.
- 15. We need a metaphor.
- 16. The Conversation It’s a metaphor. It’s all about context.
- 17. Conversations are… • Multi-modal • Multi-directional • Multi-channel
- 20. CAVE Language
- 21. CAVE Language • Whiteboard / Napkin / Presentation Friendly • Methodology Neutral • Scales Up / Scales Down • Cross-Discipline
- 22. Areas of Concern Data Context Response
- 23. List account features What features would you like to add? Pre-specified feature from billing Feature to be added Add feature to account Account Modification Mode
- 27. Context Persona Affinity Goal Environment Sentiment “PAGES Analysis”
- 28. Context User sorts product list Sort by price, lowest to highest? Value Shopper
- 30. Response ! Sensitive to cold Below 72 degrees Increase temp to 72 Thermostat shows “warming”
- 32. Example: “Telco” Telco Service Facebook Messenger Alexa Skype Slack
- 33. Telco: Inferring Intent Pay Bill Check Balance Add Feature Check Usage Initial user communication Bayesian classification Corpus
- 34. Telco: Main Intents Pay Bill Check Balance Add Feature Check Usage Billing Mode Account Modification Mode Show Usage
- 35. Telco: Billing Mode Current Bill Would you like to pay now? Pay Bill Pay the Bill Thanks! Done Dispute Charge Is this a case of fraud? Fraud Department Service Yes Charge is addressible through application of feature option. Account Modification Mode Sorry we can’t help with that. Empathetic, but firm Yes No Feature to be added Billing Mode
- 36. Telco: Upset Customer Initial user communication Sentiment analysis Angry Corpus Softer, More deferential
- 37. Telco: Problem Customer Customer Incident History Repeated history of unfounded complaints? Problem Customer No more special exceptions
- 38. Telco: High Value Customer Customer Purchase History Value of Purchases Exceeds $ Threshold? High-Value Customer Special Rewards Make Special Offers
- 39. Telco: Conditional Up-Sell ! Positive mood Traveller Buying international package Offer affiliate deal in destination country
- 40. Other Examples
- 41. Recipe Reader
- 42. Pick Up The Milk
- 43. Abridged Content for Commuters