General Characteristics and Design Taxonomy of Chatbots for COVID-19: Systematic Review (Preprint)

Abstract

BACKGROUND A conversational agent powered by artificial intelligence, commonly known as a chatbot, is one of the most recent innovations used to provide information and services during the COVID-19 pandemic. However, the multitude of conversational agents explicitly designed during the time of COVID calls for characterization and analysis using rigorous technological frameworks and extensive systematic review. OBJECTIVE This study aimed to describe the general characteristics of COVID-19 chatbots, and examine their system designs using a modified adapted design taxonomy framework METHODS We conducted a systematic review on the general characteristics and design taxonomy of COVID-19 chatbots with 59 studies included in the final analysis. This review used PRISMA in selecting articles published from January 1, 2019 to April 30, 2022 in databases and search engines. RESULTS Results show that 41/59 (70%) of studies on COVID-19 chatbot design and development are implemented in Asia (41%) and Europe (29%), and can be accessed on websites (32%), messaging apps (29%), and Android devices (24%). The COVID-19 chatbots are further classified according to their temporal profiles, appearance, intelligence, interaction, and context. In terms of the temporal profile perspective, almost half of COVID-19 chatbots interact with users for several weeks (49%) for more than a single time (58%) and can remember information from previous user interactions (44%). Regarding appearance, the majority of the chatbots assume the expert role (61%), are task-oriented (63%), and have no visual representation (56%). As for intelligence, almost half of the chatbots are artificially intelligent (48%) and respond to textual input and a set of rules (42%). More than half of these chatbots operate on a structured flow (53%) and do not portray any socio-emotional behavior (56%). In addition, nearly half can process both external data and broadcast resources (37%). Under the interaction perspective, the majority of the chatbots communicate through text (68%), react to user input (53%), and are adaptive (63%). Relative to this, more than half of the chatbots do not require additional human support (75%) and are not gamified (80%). When it comes to context, all of the COVID-19 chatbots are goal-oriented (100%), while the majority fall under the healthcare application domain (79%) and are designed to provide information to the user (61%). CONCLUSIONS The conceptualization, development, implementation, and utilization of these conversational agents emerged to mitigate the effects of the pandemic in societies worldwide. We strongly believe that this study is a starting point to help the developers conveniently choose a future-proof chatbot archetype that would meet the needs of the public amid the growing demand for a more structured pandemic response.