In our era of intelligent machines, conversational interface design (e.g. chatbots, voice interfaces) has become a growing sub-field of human-computer interaction, where the medium changes from graphical elements (buttons and links) to human-like conversation (emotions and natural language).
Many conversational interfaces use advanced NLP (Natural Language Processing) in the background, while others are based on simple decision tree logic. Regardless of the type of the backend technology, there are certain points that conversational interface designers should keep in mind in order to optimize the users’ experience and maximize the value that the conversational interfaces that they design bring to the users.
Below are some steps that conversations designers should go through and some tools that they can use to facilitate their processes at each step.
The first thing designers should do when embarking on a chatbot design project is to identify the value the bot will bring to the end user.
Is the conversation interface for recommending content? Ordering food? Having a friendly chat?
Designers need to set a purpose to which they can refer throughout their design process. At this stage, user-centered design techniques and tools can help a lot in framing the scope of the project.
One tool that designers can use at this stage is contextual inquiry. It is a highly qualitative user research methodology that does not require a big sample; it helps observe what users do/how they do it in a given context and to discover flaws that might stem from bad design. Thus, contextual inquiry can be an insightful way to understand real conversations between agents and users to help define the purpose of the chatbot.
In order for a chatbot to be well-received, it should have a distinct personality. Moreover, its personality should be well-aligned with the values and the behavior of its intended users. For this, designers should first conduct a through user research and define the conversational interface’s tone of voice.
Which words can the bot say and which ones it should avoid? What type of emojis and gifs can it send?
Choosing a clearly defined tone of voice from the very beginning helps designers not only personalize the interactions and create intimate relationships between the designed interface and its user, but also it helps them harmonize different conversation flows. This is essential for helping the users overcome the anxiety that they might have when during an interaction with an intelligent yet non-human agent. At the same time. it helps cultivate a sense of trust among the users.
In this stage, personality cards are one method that can help articulate the nuances of a chatbot’s tone of voice by providing consistency.
Conversation possibilities are endless. When users initiate a chat with a bot, they can say or ask anything they want. This can cause problems for advancing a dialog using predetermined responses. Designers must take charge and design conversation flows that will lead users through the intended conversation.
In this stage, going through the heuristic principles of user interface design can be handy for designers. One of these principles is to provide enough guidance for users to know where they are in the system, and what is expected of them. This is particularly relevant to conversational interface design as during a conversation, each question must be very clear so that the users can understand what type of information needs to be entered. Following this principle, designers minimize the chances of confusing the users and prompting them to initiate flows for which they have not planned.
“Designers should assume users are unable to understand technical terminology, therefore, error messages should almost always be expressed in plain language to ensure nothing gets lost in translation.”
— Jakob Nielsen
One of the most painful aspects in conversational interface design is the issue of misunderstandings. Chatbot technology is not yet capable of understanding every user response well enough in order to reply in a meaningful way — no matter how perfectly the script is written, under challenging situations, it will most likely fail.
Within this context, it’s important that designers anticipate some of the mistakes that the users might commit.
Supervised testing with interactive prototypes is a method that can be utterly handy for identifying some unintended flows. This is an observational testing technique where users are asked to interact with interactive mockups, complete tasks and share some thoughts about their experience and the purpose of the interfaces which they used.
Similar to apps or websites, chatbots need to be monitored and analyzed in order to iteratively improve.
One way to track the data is by using analytic platforms and/or APIc for chatbots (e.g. Botanalytics). Such platforms provide information on how the chatbots are used, where they fail, how the users interact with it. These platforms usually also include the total number of users, user retention, most used flows, words from users that the conversational agent cannot understand, and much more.
Another way of analysis is to gather data on user satisfaction through success surveys that can be applied to chatbots. For example, when users reach the end of a conversation, they can be presented with a simple survey question asking whether their perception of the interaction was satisfactory or not.
The core element of conversational interfaces is language, which is much more subjective and difficult to work with than typical visual interaction design elements. Within this context, internationalization of conversational interfaces (a.k.a. supporting multiple languages) is an aspect to approach with care and plan for from the very beginning.
Some best-practices that designers should consider when designing for multiple languages are the following:
- Maintain consistency of intents
When it comes to conversational interfaces, the king is the purpose. Strive to keep the same intents and conversation flows for different languages. Consider the culture-specific behavioral patterns and adapt the flows to make sure that the intent is maintained.
- Keep in mind formal and informal language variations
Unlike English, many languages offer formal and informal ways to address a person. This can have sensitive cultural implications and be a determinant factor for the success or failure of the conversational interface. When doing this, it is also important to adapt and get the ‘tone’ of your bot right, which also relates to the personality of your bot.
- Do not assemble sentences from fragments
When creating the language files, the DRY (don’t repeat yourself) principle of reusing phrases as fragments to assemble sentences should be avoided. Fragments do not translate naturally and their assembly into sentences is most often language specific (e.g. different grammatical ordering).
- Design for different languages from the very beginning
Entanglement of the conversation’s core logic with locale or language specific aspects can be confusing. To avoid this, whenever possible, it is useful to design and develop for more than one language from the very beginning instead of translating before. This provides a guiding force for clean separation of concerns.
To conclude, conversational interface design is a broad yet complex and nuanced field that requires a lot of contextual and technical knowledge that designers should accumulate through careful studies and lots of practice. However, there are some considerations, practices and tools which are applicable and useful for all types of conversational design. Mastering the basics is a good start for all designers interested in the field.