Research On The Proactive Interaction Design Of Conversational Agent For Smart Homes

With the advent of Everything connected and intelligent,Conversational Agent products(Amazon Alexa,Tmall Genie,Xiaodu at home,etc.)have begun to be used in families.The smart home ecosystem centered on conversational agents has become Important trend of development of smart homes.At the same time,with the support of rapidly developing computer intelligence technology(such as artificial intelligence,machine learning,adaptive systems,environmental computing,etc.),proactivity has become one of the important attributes of smart products.Smart products can not only respond to user instructions to make corresponding Action also has the ability to perceive the environment,predict behavior,and take proactive actions to change the state of the environment.Although the capability of proactive of artificial intelligence can bring many conveniences for humans(for example,it can meet the potential needs of humans;reducing efforts for humans to get services through agents,etc.),and has wide application value.But in different scenarios,people may have different attitudes towards the proactive behavior of agents,it may be a convenience or an interruption.The proactive behavior of the agent requires not only elaborated design,but also adaptive strategies based on various factors such as environment and user attributes.Therefore,how to design proactive interaction strategy for smart home conversation agents is a problem that needs to be solved.This article proposes a strategy for active interaction design of conversational agents in a smart home scenario.The main work of this paper includes two aspects:(1)Propose an proactive interaction strategy for conversational agents in smart home scenarios.This article first discusses the existing literature on the proactivity of artificial intelligence agents,and proposed the view that defining the proactivity needs to consider application scenarios,and on this basis,proposed the idea that establishing an proactive interaction strategy from the attributes of home agents,based on right of decisions and communication style,established a theoretical framework of proactive interaction strategies,and then proposed 4 proactive styles(straightforward advisors,straightforward decision makers,reversed advisors,reversed decision makers),as the main research variable.(2)Explore the factors that influence the selection of proactive interaction strategies in smart home scenarios.In this paper,starting from the smart home scene,based on the "scenarios,preferences,intentions and actuators" conceptual model proposed by the design of smart home behaviors,this paper proposes a framework for factors that have influence on proactive strategies selection.Based on this,it proposed four factors that affect the proactive interaction design selection: user activity status,functions of smart home conversational agent,user characteristics,and execution device types.These four factors further subdivided into different dimensions based on literature.As a secondary variable in the study.After establishing the theoretical framework,the author designed and organized the "Wizard of Oz" experiment.A total of 60 participants participated in the experiment,to explored the user’s preference for proactive style and the smart home scene factors which have influence on the user’s proactive style preference.The experiment has three findings:First,Users generally feel more satisfied and comfortable that the CA communicates with them in an explicit manner.In the home scenario,users generally believe that the communication efficiency of CAs is more important than politeness.The second is that prior experience of Smart home products and living state have an impact on users’ preference for proactive style of smart home CAs.Smart home products experienced users are more satisfied with the “adviser” style;family users feel more comfortable with the “adviser” style.Third,the type of conversational agent’s function will affect the user’s preference for the proactive style of the conversational agent.When the conversational agent performs the "safety and security" function,the user thinks that the "decision-maker" style is more comfortable;In the "environment and lighting" function,the user thinks that the "adviser" style is more satisfactory and more comfortable.The researchers explored the reasons behind the user interviews and provided design guidance for the conversational agent’s proactive interaction strategy.