Towards a Pattern Language for Smart Personal Assistants

Supporting users in their daily activities, thus, making their lives more comfortable, has long been a goal for consumer-oriented systems development. With the rise of smart personal assistants (SPAs), however, we have reached a new milestone along the path towards this goal. These systems assist their owners by providing personalized and context-dependent information and service. Today's implementations reach from conversational agents, such as Siri, Cortana or Google Assistant, over chatbots, which are primarily text-based, to cognitive assistants, which assist according to a user's current cognitive or emotional state. However, although both research and practice proceed with full pace, recurring design elements of SPAs have not yet been investigated. We hence propose a pattern language for smart personal assistants to guide further empirical and design efforts. Therefore, we review existing information systems, computer science and human-computer interaction literature to find recurring design characteristics among 115 different assistants. The resulting pattern language contains 22 patterns that specify the interaction behavior and the intelligence of smart personal assistants.