Text interfaces

Rough mental dump, because I was excited by the Quartz news app coming out today.

Two years ago I signed up my office for Slack and have been wanting to build chat bots of all kinds since then.

I’m personally enamored with the chat interface because I’m a child of the eighties. I think this fascination with conversational commands, dialogue trees come from personal robot assistants in science fiction, RPG dialogue systems and text-based adventures.

I wonder if everyone else who is trying to make them work are just the same as me?

Here are some attempts to validate and or justify that text interfaces.

1. No context switching. If you’re in WhatsApp talking to your friend and you want to have dinner together, it’s much easier to start off a restaurant reservation inside the app than to open a new app. If you’re in the team Slack and want to update the website it’s much easier to tell your bot to do it than to log in to the CMS.
2. History. Because of your chat history it’s easy to maintain your place and reference what’s already been done in the transaction. Things like order history and status can be seen right in the previous chat bubbles.
3. Constraints. Not having visual metaphors and cluttered interfaces simplifies the interaction and boils it down to its core.
4. Feedback loops. Sort of related to number 2. In one way communication we often wonder about how our messages are received. Questions like, “What’s the best story to put on the homepage”, “How many times should we post on Twitter”. This information can be derived by analytics and observing user behavior, but direct feedback with the user literally asking for what they want is powerful. Instead of guessing when a post will catch their attention, maybe the best time to post is when they ask for it.