📚 node [[ibm notes on chatbot writing]]
📓
garden/KGBicheno/Artificial Intelligence/Creating Chatbots/Week 1/IBM Notes on Chatbot writing.md by @KGBicheno
IBM notes on Chatbot writing
Go back to [[Building AI-powered Chatbots with Watson]] or the [[Main AI Page]], or consider an alternative take on Chatbot conversation design from the [[Feminist Chatbot course's]] page on [[Conversation Design]]
Tone and personality
Consider the audience and purpose of the chatbot
- Imagine you're training a human agent ^fd84d6
- What is appropriate for humans is appropriate for bots
- Remember to own your bothood
Constraints
- Give your bot a name, anchor it in reality
- Again, own your bothood, never pretend it's a human
-
Always be specific in what the bot is able to do
- Never say "How can I help you?"
-
Always say "I'm here to help you with opening hours, membership queries, and placing orders."
- You can't even promise to train a human agent in all things that could come from the question, "How can I help you?" Let alone a chatbot.
3 Rules
-
Avoid Yes and NO answers
- It's very easy to be wrong
- It's much easier to look bottish
-
Incorporate part of your user's question in your response
- The user feels more understood
- The Chatbot looks more understood
- Using emoticons for empathy is allowable
- Incorporate user input
-
Provide succinct and accurate answers
- Nobody like to read walls of text
- Especially if they're wrong
- If you need to infodumb, link to a page on your site
- You can embed HTML to link
A bot answering with a yes or no response
a bot answering with some empathy and user input
📖 stoas
- public document at doc.anagora.org/ibm-notes-on-chatbot-writing
- video call at meet.jit.si/ibm-notes-on-chatbot-writing
⥱ context
↑ pushing here
(none)
(none)
↓ pulling this
(none)
(none)
🔎 full text search for 'ibm notes on chatbot writing'