Personal Assistant Management System
Chapter One
Introduction
1.1 Background of the study
In
recent years, technical progress has brought us systems that increasingly
reduce the complexity of our everyday lives. Thereby, smart personal assistants
(SPAs), defined as systems that use “input such as the user’s voice and
contextual information to provide assistance by answering questions in natural
language, making recommendations and performing actions” [4, p. 223], have just
conquered a broad consumer market. Recent forecasts predict the worldwide user
count for SPAs such as Amazon Alexa, Apple’s Siri or Microsoft Cortana to
increase from 390 million in 2015 to 1.8 billion in 2021, which results in 2.3
billion USD average sales growth per year [33]. These systems’ success story is
mainly because digital assistants combine the comfort of intuitive natural
language interaction with the utility of personalized and situation-dependent
information and service provision. In practice, SPAs unfold their potential in
various forms and contexts [8], such as on smartphones [38], in smart home
environments [11], in cars [5], in service encounters [43], or as support for
elderly or impaired people [11].
However,
prominent examples such as those mentioned above represent SPAs that are
explicitly developed for a broad consumer market. They thus are only the tip of
the iceberg. Since the idea of information systems (IS) that pervasively assist
humans in conducting certain tasks is by far not new, numerous efforts were
made in IS, computer science and human-computer-interaction research to develop
SPAs as previously defined. Simultaneously, research and practice has often
neglected to ‘stand on the shoulders of giants’ by building up on each other’s
work. This has led to a partly overlapping diversity of concepts and terms for
the developed artifact. For example, while many scholars entitle their SPA as a
conversational agent, others would differ between mainly text-based and
voice-based systems. Still others would label the text-based SPA as chatbot and
the voice-based SPA as smart speaker. This example shows, that the range of
possible terms for different types of SPAs differ heavily due to lacking
conceptual clarity. The interchangeable use of terms has also been observed by
other scholars [e.g., 8]. We, however, argue that conceptual clarity is highly
important, not only for a correct categorization of SPAs to a higher-order
group. It is also important for finding similarities and differences between
systems, identifying design principles, recurring requirements and design
practices (i.e., patterns) and, finally, reline future research and practice
with a reliable structure to allocate SPA-related work. Therefore, this paper
offers a classification approach for SPAs. Based on an exhaustive literature
review, we derived design characteristics of 115 SPAs that were developed
within a research project or for commercial purposes. We further performed a
k-means cluster analysis to yield groups of SPAs which, according to the design
characteristics, have a high internal homogeneity (i.e., most similar items are
within one cluster) and a high external heterogeneity (i.e., each cluster is
highly distinctive to other clusters). An analysis of the clusters, their
similarities and differences, resulted in archetypes of SPAs, which are defined
by the most expressive design characteristics of each cluster. We thus aim to
contribute to research by providing a classification for SPAs that aid future
SPA research to yield more specific and meaningful contributions. We further
contribute to practice by showing design differences between the various SPA
types which may influence development decisions.
Chapters: 1 - 5
Delivery: Email