Summary of Key Points on Generative AI Conversations

The article analyzes 425 interactions with generative AI bots like ChatGPT, Bing Chat, and Bard, revealing six distinct types of conversations users engage in, depending on their skill levels and information needs. Understanding these conversation types is crucial for designing effective user interfaces for AI chatbots.

The Six Types of Generative AI Conversations

Conversation Length and Success

Conversation length does not necessarily indicate success or failure. Shorter conversations may signify well-articulated prompts (pinpointing) or search queries, while longer conversations facilitate learning through exploration or chiseling.

Tips for Using Generative AI Chatbots

• Use detailed, pinpointing prompts for specific information needs.
• Add phrases like "Ask me questions if you need additional information" to help the bot articulate constraints.
• For factual queries, search engines may provide better results (for now).

Recommendations for Designing Generative AI UX

• Determine the conversation type early based on the prompt's length, structure, and answer complexity.
• Tailor the bot's behavior and suggested prompts based on the identified conversation type.
• For vague prompts (funneling conversations), ask questions to narrow down the topic.
• For broad conversations (exploring, chiseling), offer depth-focused and breadth-focused suggested prompts to facilitate learning.
• In expanding conversations, provide broader refinement prompts to help users find useful information.

By understanding the different types of generative AI conversations and tailoring the user experience accordingly, designers can create more effective and intuitive interfaces for AI chatbots, enabling users to efficiently address their information needs.