Last updated on 10 January 2023
We often ask people for information. Good conversation design helps us create space for these interactions to be natural and human-friendly.
Our most natural interactions with computers have the same rhythm and flow as conversations with people.
Child developmental science tells us that infants are born equipped with the basics for social communication. Newborns prefer direct eye contact and recognize their parents’ voices, as well as their own native languages. In other words, we’re “hard-coded” for conversation.
We follow certain rules and conventions when interacting with each other. So why shouldn’t computers follow the same rules when they interact with us?
Micro-Level Conversation Design: NS Customer Service
A few weeks ago, I wanted to unsubscribe from the Dutch Railways’ (NS: Nederlandse Spoorwegen) member magazine. On their “Customer Service” page, I met Eva, their chat bot. “She” couldn’t answer my question directly, so she helped me “dictate” an e-mail to their human customer service.
In other words, instead of giving me a contact form to fill out, they let me “chat” with Eva. Lots of websites try this, but only few get it right; here’s a snippet of that conversation (click the image to zoom in).
Underneath the surface, I typed in a search query that returned no results, and then I completed a form. However, the process felt smooth and easy, because it happened in a conversation.
The bot served me each form field as a question, “What is your name?” “What is your title?” “What is your e-mail address?” On the right side, I could watch as the bot ‘dictated’ my e-mail in real-time.
A Diamond in the Rough
As mentioned earlier, lots of websites try this kind of form input, but only few succeed. Mostly, they forget that the fake person has to adhere to some social protocols, or else it will be too fake. Here’s why Eva worked.
She Gave Me Instant feedback
Eva displays the e-mail she is drafting during the conversation, so the human can see everything she does in real-time. That level of transparency is very important when a computer is doing something for you.
She Respected My Privacy
Here is a further snippet from the conversation, where Eva asks me for my date of birth:
As with most forms, there are “required” fields and optional ones. Ideally, the NS shouldn’t be asking for my date of birth here, but they make it optional, at least. Eva does not ask directly “what is your date of birth,” but rather “may I include your date of birth?” This small distinction allows people to reply “no” and skip that question. That’s certainly a good thing.
She Tolerated Me Being a Smart-ass
As the date of birth example shows, a little linguistic nuance makes a lots of difference when interacting in a conversation. When Eva asked me “if I would like to include” my card number, I couldn’t resist replying “yes,” instead of just giving the number directly. With a lesser chat bot, this might have gone wrong, probably with some sort of error message or a standard response that didn’t fit into the conversation at all.
Eva, however, ignored my impoliteness and simply rephrased the question in a direct way, “Ok, would you type” the card number. It’s an extremely subtle conversation trick, but it is also quite powerful. It gave me the impression that Eva is a tolerant, helpful professional that just wants to fill out this form as efficiently as possible, even though “she” is just some lines of code.
Here are Some Different Examples
Natural-Language Interaction
Our pocket computers lend themselves wonderfully to natural-language interaction (NLI, or speech interaction). Of course, no discussion of NLI is complete without mentioning Apple’s Siri, which allows people to talk to their iPhones as they would to personal assistants. This article makes a brief statement on why NLI can be so powerful, and Bill Buxton talks in-depth about how NLI should be combined with non-verbal interaction.
Instant Messaging as a Content Platform
Similar to Eva, the NS chat bot, some instant messaging (IM) services use IM to serve content. MXit, an IM platform that’s popular in Africa, enables people to take quizzes and input data, e.g. with Bsmrt. It all happens in the form of a conversation and seems to work quite well.
Mad Libs
Cheek’d, the dating site with a twist, shows us a brilliant and sexy alternative to boring web forms. Some would call this “fill-in-the-blank,” but Luke Wroblewski and I agree that “Mad Libs” is a much more fitting name. This pattern structures the form fields as a sentence. Doing so gives the impression of a conversation and making the interaction feel much more “human.”
Conversation Design at the Macro-Level
The examples shown here are all conversations in human-computer interfaces, i.e. on the micro level. However, the real #win comes from using conversation design as a holistic model for all human experience design. That’s why conversations provide a great framework for thinking about content strategy, consumer experience, service design, and human-system interaction.
This UX Magazine article includes some great starting points. And for a broader look at how machines interact with humans in human-like ways, check out this fascinating podcast on Radiolab.
