“What do you do?”: A conversation designer’s most feared question
“What do you do?”
Perhaps the most feared, but the most fun, question for conversation designers.
My initial thought: “Well, how do I phrase this?”
Here’s a typical conversation:
“Oh, jeez…
Well, have you used a chatbot?”
The answer goes: “A chat-what?”
Or in the best case scenario:
“Oh yeah, I used the blah blah airline chatbot to book a flight the other day.” In which case, I breathe a sigh of relief, and say, “Well, I write the copy that you see inside there. Like if you asked ‘when is my flight arriving,” whatever response you get, I determine what that should look like.”
“So I’m basically chatting with you when I talk to blah blah airline.”
“Well yes, if I was the conversation designer for that airline, it would be like chatting with me.”
“Cool.”
In the first scenario where someone doesn’t know what a chatbot is, then the conversation goes on much longer to explain the technology.
In essence, conversation designers formulate the interactions, or say, the script, between machines and humans. A Facebook Messenger chatbot for your airline, which you know is a chatbot because it lets you book a flight without talking to an actual human on the other side, is a scripted machine.
Some designers will steer away from using the term “script” because it makes it sound like a narrow back-and-forth that simply acts as a Q & A. I think it’s fair to say we value ourselves on creating a conversation, emphasis on “conversation,” not a command and answer. Companies are enabling technologies such as natural language processing, which allows the machine to pick up the intended meaning from humans, to become more sophisticated to allow users to be as open-ended as possible in what they say to the chatbot.
On high-level, designing conversations is mostly copy. Because it is a messaging technology that carries out back-and-forth interactions that are either typed out or spoken, it is accurate to say the designer creates copy. Even on voice interfaces like Amazon Alexa, we must create a script of copy that informs what the user can say to Alexa and what copy the Alexa should be programmed to respond with.
However, most people associate copy with articles, social media posts and other stand-alone text. I’ve created those copies too. But a conversation cannot stand alone.
We cannot write something, publish, see engagement and move onto creating another conversation like it. We have to fix that same conversation. It’s much more iterative, and we must watch user behavior very closely to adapt the product.
When I was a copywriter, I published a blog post, tweeted about it, then tracked metrics such as click-through rates, number of times that article was shared, etc. I learned which articles performed the best, and wrote more articles under the same topic, but didn’t tweak that article. I never fixed a sentence that didn’t seem to resonate with someone, or elicited the desired emotional reaction. If I wanted to, I could have tested each sentence with readers, but it’s not time-efficient to do such thing because it’s about the overall impression. Instead, the best-performing article became a case study to emulate in future writing.
When conversation designers designs a chatbot and “publishes” it on a website, which means making it available for public use, they have to track the “reader” (the user) at every point of the “article” which is the conversation.
We must know whether the question asked by the bot is properly understood by the user by seeing what they answered and how long it took them to provide the answer needed to move forward with the next part of the conversation.
It’s fun because it lets me really learn about my user to understand what he or she is thinking at all major points of the conversation.
Our satisfaction comes from getting stuff done for people. Our drug is the user’s trust, knowing the conversation has achieved ethics and efficiency in the experience.
As always, if you have any questions or musings about the future of human-machine conversations, reach out to me here.
