The Rise of Bots: a Survey of Conversational Interfaces, Patterns, and Paradigms

Editor's Notes

• #3: Chatbots or chatterbots, in the context of the “Turing test”, try to trick users into considering them to be human. Eliza, for instance, pretends to be a psychologist simply by sending vague and evasive questions back to the user.
• #4: Many messaging platforms have started opening up their programming interfaces (APIs) to third-party developers: they can offer services or provide access to information through bots. Popular example: food ordering or taxi service.
• #5: Many messaging platforms have expanded their features with enhanced UI elements. They help communicating information: Clearer structure, Options and guidance, Higher information density.
• #8: Why are chatbots deemed to be such a big opportunity? The usual argument: people are using chat apps more frequently than they are using social networks. If you have followed the conversation online, this is a graph seen very often: in 2015 messaging apps usage has surpassed social network apps usage. Messaging is the most frequently performed action on a smartphone. Users are naturally already heavily invested in messaging. Hype: potential huge business opportunity. “Go where the users are.”
• #10: Bots are often described as being a stepping stone in the path from messaging to fully automated voice assistants (Siri, Alexa, Cortana, etc). Voice assistants provide a completely new/futuristic computing interface (HAL 9000). They have been described as the future of computing, even if they do not solve any new problems (yet). Speech is one of our first, deeply instinctual, skills. Voice UIs allow us to remain “fully human”. They remove the disconnect between man and machine. Amazon Alexa example: like in the “Amazon shop”, these products are not designed for more intimacy/contact. Computing is breaking out of its physical confines (smartphone, watch), it will become invisible, unremarkable, and familiar.
• #11: Bots are usually seen as “transitional fossils”: from messaging to conversational voice agents. They make use of the same interface patterns: natural language (spoken or written)! We argue: Wrong! Voice interfaces are entirely different from touch/text-based interfaces. Voice assistants: not device-bound, “wide open space” of semantics and interpretation given by voice. Hard to design. Much harder problem (orthogonal to actual services provided). Messaging has smaller scope, one purpose: inherently tied to text input, user in close contact with input device, simple traditional interaction model (of mobile applications) that can be mastered by the user.
• #12: So, spoken/natural language is a red herring. Bots transform messaging platforms into full-fledged software delivery platforms. Bot platforms include UI elements and services that enable complex “application-like” bots, that are effective and efficient on smartphone interfaces with no ambiguity. Smartphones are intimate and use a “traditional” touch-based interface. No speaking required. Example: WeChat in China. One application: full OS, software and service delivery to the final end-user (messaging, payments, social networking, news, reading list, step counter, QR Codes, etc). Justification: Apple advantage is usually given by superior, exclusive software on superior hardware. In China: perception of superior software is hindered by WeChat. The messaging platform is the superior software, the most important layer, and it is not exclusive. 2017 iPhone sales in China have been languishing because Chinese users are less loyal to the Apple SW platform.