Navigating the Hurdles of Conversation Design
summary
Conversation design argues that real dialogue is messy and must handle four concrete failure modes: recognition errors, speaker‑identification, contextual and system errors. Examples include an Amtrak prototype that failed to capture flexible dates until researchers heard customers say “that depends on price,” a Google Assistant bug that renamed a user after “call me by your name trailer” (now fixed), and Google research showing 54% anger at “sorry I didn’t understand.” Prototyping uses Wizard of Oz studies, Dialogflow/Actions and field UX vans; fixes favor guardrails, front‑loading and multimodal recovery.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:00What is Conversation Design
- 9:58Sample Dialogues
- 13:18The Nature of Conversation
- 22:02Preventing Misalignment
- 26:20Know How and When to Ask
- 31:29Dont Give Up
- 39:31Questions
related talks
33mStrategies for Conversational Product Experiences
32mParty like it's 1970: Conversational interfaces are back (into your Android app)
38mPontus Wärnestål | Compass UX Session 2025
28mHuman Computer Interaction: The UX of AI
43mHi, Want to Chat? When Conversation Is the Interface, the Challenge Is Designing Personalities
03mDesigning for humanlike interaction
20mEmbracing the Voice Revolution
10mThe Role of the UX Designer in Creating Conversational Interfaces