With the development of virtual assistants, like Apple’s Siri, social chatbots such as Woebot, and conversational agents in customer services, people become acquainted with conversational AI to contact companies for their questions and complaints. However, the technology does not live up to its full potential yet. Currently, people perceive their conversations with conversational AI in customer service as unnatural and not cooperative.
Challenges
There are both technological and social challenges in the implementation of conversational AI in people’s everyday life. It is challenging to translate content into a conversational format taking the social context into account, and to optimize AI as valuable co-workers for employees. Also, users face challenges when interacting with AI: they fear that their intent will be misunderstood and evaluate the interaction as impersonal.
Approach
In the Smooth Operator project conversational AI will be developed, evaluated, and tested in real-life customer service settings resulting in the experience of AI as cooperative communication partner by both service employees and customers. A multidisciplinary team of scholars therefore closely collaborates with private partners from industry.
Outcomes
The project leads to cutting-edge scientific-based algorithms, lexica, guidelines, and tailored customers’ profiles for designing and implementing effective personalized conversational AI. Furthermore, the knowledge utilization is strongly focused on making employees AI-savvy and optimizes their abilities in a challenging human-AI environment.
Besides developing, evaluating and examining the effects of personalized conversational agents, we also invest in human capital. The aims of Smooth Operator is threefold.
1. Development and evaluation of AI-agents
The collaboration between scholars, AI-developers, and customer service counters in several organizations enables us to develop high-end algorithms, build upon proven technology, and based on real data. We also put effort into a solid and broad evaluation of the quality of the conversations conducted by our systems.
2. Cognitive, affective and behavioral effects
In order to examine how employees and customers perceive personalized conversational AI, we conduct research in real-life settings. Here, different case studies are used in which the conversational agent performs different tasks in different contexts. This enhances the generalizability of the project’s findings.
3. Professionalize the field
AI-implementation entails new job types for which employees (and employers) need to be prepared. To make them more AI-savvy, private partners are strongly involved as ambassadors of the project, and online teaching packages and guidelines will be developed.
