So you want to make a robot that talks to people.
It’s not as simple as you think. It’s hard to imagine the complexity of human language when it comes so naturally to you. Unlike visual design, where it’s pretty easy to self-determine whether or not you have artistic talent and ability (especially if you’re anything like me and struggle to draw even a passable stick figure). Because you have conversations with many different people constantly every day, it can be tempting to think that typing out a conversation with a robot will be easy, or at the very least — easier.
Here’s a thought experiment to illustrate this point.
Imagine you’re the queen and you are negotiating peace with a neighboring king. You’ve been fighting over 100 acres of land on the border of your two kingdoms, and you think you should just split it to avoid bloodshed. Neither you nor the king can travel, so you’re going to send your best adviser to negotiate. What advice do you give him?
The first part is easy – you tell your adviser to offer to split the land. But what if your adviser walks in and the king offers to take 25 acres and give you 75 since you have a bigger military? Or, what if the king wants a different 100 acres of your land in return for the 100 acres being disputed? What if at any point any of the offers being made by the adviser are misinterpreted to the point that war could be declared? This gets complicated quickly, and so far, we didn’t talk about the emotional considerations. What if the king is in a bad mood – should your adviser stall and wait until the king is in a better mood? If the king is in a great mood, should your adviser ask for all the land?
Furthermore, all these decisions and actions are based on assumptions and on the premise that the adviser truly understands that the core of the issue has to do with the dispute over the 100 acres of land. Wouldn’t it be worth the time and effort to do some preliminary work, identify the needs and expectations of each side, before you start coming up with negotiation techniques and words to use?
Just because you have conversations every day doesn’t mean that you have successful ones every day. And even if you do have successful conversations every day, it doesn’t mean that you know why or how. Nor does it mean that you can deconstruct the underlying choices into a strong conversational system architecture that can support all of these scenarios and keep your user engaged, both from a high level game strategy perspective (what actions to take in different scenarios), and from a lower level tactical one (how to adjust your phrasing, depending on the specifics of the situation you’re in, for maximum effect).
The game is to design these conversational systems with enough logic that they can successfully automate more than 80% of the interactions that robots would encounter. It takes expert designers who know the game well.
So, before you send your robots in to interact with your customers, make sure you’ve done the preliminary work of gathering the right information and developing the appropriate “negotiation tactics”. But don’t do it alone. Find conversational design experts to help you through it.