Designing conversational user interface experience is complicated because conversation comes with many expectations. When these expectations are met, we feel the interface is natural, but once violated, we feel something is amiss. The last decade witnessed human language technologies and behaviours to enable humans converse with software using spoken dialogue to access, create and process information. Less is known about the practicalities of designing chatbot interactions. In this paper, we introduce the nature of conversational user interfaces (CUIs) and describe the underlying technologies they are based on. Moreover, we define guidelines for designing conversational interfaces in various domains. This paper particularly focuses on classifying the elements and techniques used in CUI design patterns. After concluding certain challenges with CUI, we discuss important features and chatbot states to be considered in CUI design for specific domain. We envisage this study to support CUI researchers to design tailored chatbots applicable into certain domain and improve the current state of research challenges in the field of Artificial Intelligence and conversational agents.
Conversational User Interface (CUI) is a software that runs simple and structurally repetitive tasks inside a messaging application [1]. Conversation is how we share knowledge, emotions and it has been part of our makeup for years. Moreover, messaging applications are used in everyday life and are becoming a layer in our daily life. Chatbots are expected to be the new generation of digital product design, after the evolution from Web to mobile applications. The reason lies behind the simplicity offered by the CUI when performing a task, which otherwise could be tedious or requires more time and effort if performed via web or mobile apps. In addition, users spend more time on their messaging platforms than any mobile apps, this generates a new shift to address user needs. Introducing GUI to the web and mobile makes it less efficient and tedious to perform a task [2]. To illustrate, it takes around 18 clicks to perform an airline reservation, while we're faced with an unwieldy array of buttons, ads, drop-downs, text boxes and more. Simplicity is crucial when communicating with our device. Rather than pulling up an app to search for restaurant, tap to select time and type in number of people, we can simply tell a chatbot to "Book me a table for three at 6 tonight, at the Elzar's place". We could facilitate a conversation between a person and a service to fulfill users' request [2]. Using a chatbot to access a service wont requires users to familiarise themselves with the bot as it is the case with mobile apps. Moreover, developing a bot requires less time and cost to build and deploy than a web or mobile app. Consequently, currently the most popular apps in the world are either social in nature or primary for messaging. Although this new paradigm shift, the current state of conversational interface is limited in terms of established user interface design patterns. The current resources on CUI design and the specific bot behaviour in certain domain is limited. In addition, it's unclear when bots should be text or button and keyboard based. However, the big question remains how to use this new medium to build great user experience. It is important to consider users' point of view, such as their needs, what motivates them to seek your bot and how to create unique CUIs for various users and domains when designing CUIs. We witness a transformation represented by chatbots for messaging apps, similar to what API's were for Web 2.0. This is a new way to build services and create interaction experiences for users. This paper investigates application domains for conversational interfaces and defines techniques and features per domain and describes a way to implement domain specific CUI features. We particularly cover user needs and interaction with the chatbot in a specific domain. We contribute to the definition of domain specific CUI design patterns from the UX point and the existing knowledge on CUI design. Moreover, it will set the roadmap to define best practices and address CUI design challenges for researchers and developers to follow when designing domain specific CUIs. We will investigate the role of design in CUI and what it means to design for chatbot experience.
Although CUIs are rather small sight compared to web and mobile apps, more applications are switching from GUI in favour of conversational interface. The reason is messaging is growing as a platform, bots are easy to install and can deliver unique user experience. Effective bots are context aware and can retrieve past interaction and use the data in the current context. Current CUIs are driven either by artificial intelligence or have a human facilitating some tasks behind the scene. This is to enhance user experience since it feels that the bot is starting to understand them. Bots living on messaging apps don’t need the user to download anything. CUIs are based on text dialogue and buttons, although some platforms have added new services and features to their existing one. For example, the gaming platform offered by Telegram bot API. To create effective interaction design, a balance between the text dialogue and the custom keyboard is necessary as a source of input and part of the CUI [2]. Bots personality is important to understand how it can build empathy and emotional connection with the user (See Table-I for a comparison of the major differences between the native vs bot applications).
To better present the major differences between a mobile and chatbot application, we apply it to an example of providing a service to the user, first with a mobile application, then by using conversational interface. Mobile Application: John would enjoy having a popcorn while enjoying the match in the stadium. He notices the stadium has announced about popcorn ordering online. John unlocks his phone, goes to google play, searches for an app, puts his password, waits for it to download, creates an account, enters his credit card details, figures out where he act
This content is AI-processed based on open access ArXiv data.
