Conversations with Computers & Systems
Dialog based interaction with computers should feel comfortable to users and fit into the habits that we have practiced for centuries. Dialog based interactions with a “conversational interface” present us with challenges when technical feasibility meets the hidden complexity of natural conversations. With dialog based systems, however, the user no longer has to get used to input methods such as terminal commands, mouse or touchscreen: The system or the computer must learn to adapt to the user – and that is exactly what conversational design is all about.
What is Conversational Design
Conversational design does not simply mean writing down texts with a dialogue structure. Precisely because we use messenger apps every day (studies show that, on average, we have 3 messenger apps installed on our mobile phone and send 3 messages per hour), we expect a certain naturalness in dialogs – regardless of whether we interact with a chatbot, voicebot or human. And it is precisely this expectation that formulates the demand for good conversational design.
Reading Between the Lines
Even if we have the text in front of us in black and white, most of the processing and, above all, interpretation takes place in our heads. We can even decipher short messages without any problems; subtle, linguistic cues are linked to context; we believe we recognize personality and humor and look for a common thread, a coherent story.
This is precisely why it is anything but easy to be consciously “conversational” – that is, in dialogue – with technical systems.
Multidisciplinary Approach
Regardless of whether it is about dialog structures for a specific project or a holistic, conversational design strategy, conversational design uses concepts from various disciplines: It is a combination of copywriting (texts) via UX design, interaction design, visual design and audio design and understanding of language.
Google compares the task of a conversational designer with an architect, who in turn determines what users can do in an area, taking into account user experience, needs and technical limitations.
NLU + Content = Conversational Design
Good conversational design does not just need dialogs, but also logically coherent dialog structures and an iteratively optimized user experience that takes the respective context of the user into account. The necessary basics for this can be found in linguistics, UX design and current best practices.
Essentially, conversational design encompasses the subjects of Natural Language Understanding (interpreting user inquiries correctly) and Dialog Management (providing the appropriate content and answers). One cannot do without the other.

3 Pillars of a Conversation
Before we create our first dialog, we should take a quick look at 3 general principles that distinguish successful conversations and dialogs.
Cooperative Principle
In linguistics, the cooperation principle is used to describe how communication participants normally behave – in other words, how we can expect the users to act when they use a conversational interface. From a scientific point of view, the principle is described as follows:
“Make your contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged. – Grice (1975) in “Logic and conversation”
Put more simply, we assume that a user wants to lead the conversation to their goal with their messages – and the user also assumes that the digital assistant is also goal-oriented. Every message is interpreted as relevant in this context and thus harbors the chance that the conversation can be much more efficient, but also the risk that misunderstandings arise.
Turn-Taking
As a rule, dialogues need at least two partners who mutually contribute – which is called turn-taking in linguistics. In natural conversations, we don’t worry about handing over the speaking role. We almost automatically take over after questions, speaking pauses or filler words.
When designing the dialogue, however, we have to use this very consciously in order to put the user in the role of the speaker. Usually the user can only give one answer, then the system takes over again. This form of dialog guidance – short answers from the user followed by an interpretation and answer by the digital assistant – is particularly challenging if, for example, the user was interrupted too early or sent a message incorrectly. In dialogue design, correction options and “going back” must be taken into account in advance.
Context
Even when it comes to voice input, we now know exactly what the user said – thanks to Advanced Speech Recognition. But interpreting what it means is still a challenge. This interpretation is also difficult because the relative context of the conversation has to be taken into account.
It is challenging because the same words in different scenarios can have different meanings. For example, the word “tape” could either refer to recording equipment, packaging material or the action of recording something or gluing something together. The context is decisive.
But not only the meaning of words, but that of entire statements is context-sensitive, i.e. dependent on the environment or situation of the user. This is precisely why it is important in conversational design to always develop dialogues and logic processes as close to the user’s reality as possible.
Success Criteria for Digital Assistants
The three principles are – admittedly – very generalistic and intended more as a meta-guideline. The question that automatically arises is: What makes a digital assistant “good”? Derived from this, individual dialogues can be examined for their success in order to enable a uniform assessment of the quality.
In conversational design it is important to master three central challenges, which are also used as success criteria:

- Navigability: Can users navigate easily?
- Discoverability: Can users simply explore and use the functions?
- Usability: Can users achieve their goals?
A well-designed digital assistant enables users to easily use the various functions, guides them through complex structures and thus helps them to achieve their goal.
Guidelines: Do’s & Don’ts
Tying in with the success criteria, we can derive some concrete guidelines and best practices that can be used for the development of your digital assistant. The following order does not reflect the importance of the individual aspects.
Greeting

The first impression matters- but the thousandth one also does. With the greeting, the digital assistant has the chance to set the context and expectations of the user. So a greeting can be very versatile:
- First use? Why not give some simple instructions?
- Returning user? How about an individualised summary?
Response to Mistakes

It is inevitable that the digital assistant can’t answer every question under the sun. That makes it crucial to deliver a clear answer to an unrecognized request instead of just offering a generalized answer.
Character & Personality

No matter if consciously or unconsciously: Users will interpret the answers of the digital assistant and characterise it. Part of Conversational Design is to decide when and how much personality you want to show.
Taking Context into Account

By recognizing and storing slots or extracting entities, user inquiries can be processed directly and precisely. For many users, it is particularly annoying when an answer has to be given several times, although the information could actually already have been known.
Being Precise

Since misunderstandings are inevitable, it is helpful to suggest to the user that they have been understood correctly and that the digital assistant answers the correct question. Simple repetitions referencing the question in the answer are sufficient. This way, the user can be sure that they have been understood.
Reacting to Mistakes

Not only user inquiries can lead to errors – information coming from third-party systems and interfaces can also be incorrect, empty or not available. Therefore, small logic tests of the responses from APIs help to catch an error in the interface in a user-friendly manner.
Using “Real” Language

The great advantage of digital assistants lies in their natural language understanding. It is therefore particularly unfortunate when we use cryptic codes or unnatural commands (e.g. “/ help” instead of simply “help”). A digital assistant should try to be helpful – understand and speak the language of the users.
Conversational Design needs to be User Oriented
The assistance function is already included in the name of the “digital assistant” – and an assistant helps others to achieve their goal better and more effectively. This is precisely why it is essential that a large part of the work in the field of conversational design takes place in dialogue with users. Frequent tests, fast iteration cycles and the consideration of feedback are the real success factors in conversational design (and user experience design in general). In addition to all the principles and guidelines from this article, one thing should remain above all: The focus on the user and their needs.
Leave a Reply