How to design a conversational interface?

Many people who go to the Emergency Department don't actually have an emergency. A few months ago, the Mater Hospital in Dublin asked the NCAD to work on a design solution.

Problem Statement

The A&E of the Mater Hospital is overcrowded.  We worked on a design solution that might help inform patients of the existence of alternative options before they arrive at the ED. 

One of the main issues faced by the end-users was that in an emergency situation, human service agents couldn't process requests quickly enough to satisfy them. This is why our solution included an automated bot called Pat.

The Solution: A rebranding of the Website and a Chatbot

Pat is chatbot that guides website visitors to and through the service they need, as well as helping connect to outside services.

The three goals of this bot are: inform, reassure, convince.

The Pat approach brings information to people rather than having to find each item themselves.

Pat is a reassuring figure. It adds that “personal voice” to the Mater Hospital website that patients have been looking for according to our interviews.

Chatbots and Voice Design

In terms of Interaction Design., chatbots design is a fascinating area, mostly because voice design recycles a large amount of HCI information that is already “known” and re-applies it to contemporary experience design. You can find below the learnings from the project.

First, the technical capability of the bot needs to be considered. According to a recent article by Google AI (link in bio), computer sciences research about chatbots tends to focus on the quality of speech. Modern conversational agents (chatbots) tend to be highly specialized — they perform well as long as users don’t stray too far from their expected usage. To better handle a wide variety of conversational topics, open-domain dialog research explores a complementary approach attempting to develop a chatbot that is not specialized but can still chat about virtually anything a user wants. 

Secondly, reading case studies online enabled us to determine that a good user experience with chatbots requires simplicity, intuitive interfaces, and maximal similarity to a natural human conversation. 

For example, I noticed as a user that often, chatbots design present a green circle as some sort of visual indicator of availability. At first, I found quite weird to indicate that a robot is online and available... However, I tried to design this bot with and without the green circle and I figured out that this visual cue was useful to prompt the conversation. The green dot in this case is to follow the user's mental model and provide them with a convention they are used to on messaging applications. The user will see the green dot and they will see availability, quick response time, an active chat, etc.

Another concern was prototyping our conversation flows and testing multiple responses to see which ones work best. Not only had we to consider how users use chatbots, but also how they use messaging applications. We also needed the testing phase to be aware of the context we were putting the bot in. Should a hospital bot be very friendly or serious (personality of the bot), what are people using it for, etc?

Readings

Here are some excellent books which explain how the principles of HCI and cognitive psychology have been applied to early voice apps and IVRs.