AI agents such as chatbots or voicebots typically have an unemotional and uncaring appearance, rendering them poor substitutes for interpersonal interactions. Marketing scholars suggest dealing with this problem by including social and emotional aspects of the customer experience next to efficiency-driven capabilities. Specifically, it is argued that empathy is essential in designing AI-enabled interactions to activate social responses in consumers and contribute to business effectiveness. As marketers grapple with pressing questions such as “What impact do empathic AI agents have on consumer responses?” and “When does AI empathy prove beneficial?” the guidance from business scholars in addressing these concerns remains limited. This results in an impediment to businesses fully incorporating AI into a more effective brand and shopping experience.

There are profound theoretical and managerial implications associated with the recognition that AI agents can exhibit empathy in service encounters. Although typically a characteristic exclusively attributed to human beings, empathy is now increasingly present in AI agents. It follows that as AI agents such as voice assistants progressively display emotional intelligence traits, consumers are expected to rely on empathy in their judgment of the AI agent’s performance. In service encounters, empathy positively impacts communication and interaction processes between salesperson and customer. However, to date the business-related literature fails to provide any structured guidance in relation to the design of empathic AI agents.

Extant research has focused on two responses associated with empathy. The first is a cognitive reaction, also labeled perspective taking, which refers to the ability to understand another person’s thoughts, feelings, and intentions. This thinking state involves actively placing oneself in the perspective of another individual to understand the perspective of others and to anticipate their reactions. The second is an affective reaction, reflected in the ability to “try on” the emotion of another person and emotionally respond congruently with the perceived welfare of others. This form of empathy enables a person to sense and respond to caring and warmth for another person without experiencing the stimuli that prompted this emotion.

Commercially available AI agents remain somewhat distant from being able to share the perspectives of users (cognitive empathy) in precisely the same way humans would. Empathic AI agents capable of sensing, understanding, and mimicking human emotions, while adapting the behavior to their counterparts, are currently under development but unavailable commercially. As such, any prediction that AI agents, through affective computing, will identify the emotions of the users automatically and resemble human emotional intelligence in large-scale marketing applications can currently be seen only as speculation. In a new study, Alex Mari, Andreina Mandelli and René Algesheimer focused specifically on the affective reaction of the human counterpart to the AI empathy triggered by an AI agent.

A total of 412 families participated in a single-session online experiment to purchase batteries and paracetamol on an ad-hoc app for Alexa that replicated the native voice shopping process and manipulated its level of AI empathy (high vs. standard). The author demonstrated the generalized positive valence of AI empathy on how consumers interact and respond to voice assistants. When the voice assistant appears to be empathic (vs. standard), even without exhibiting the ability to sense and adapt to the interlocutor’s emotions, the spontaneous consumer’s responses in terms of perceptions, beliefs, and downstream behavioral intentions are more favorable.

However, the authors identified a boundary condition indicating that consumers do not indiscriminately adapt to the empathy level of the voice assistant. They revealed that family members shopping together with empathic (vs. standard) voice assistants display significantly higher scores for functional voice assistant characteristics, whereas individual buyers prefer standard Alexa interactions. The emerging theoretical model called the “AI Empathy Response Model” shows the prominent mediating power of discerned empathy, alongside usefulness (functional) and trust (relational), in explaining the systematic consumer responses toward the voice assistant depending on its ability to fulfill functional, relational, and social-emotional needs.

Open Access Article on the Journal of Business Research: https://cutt.ly/AIEmpathy

Citation: Mari, A., Mandelli, A., & Algesheimer, R. (2024). Empathic voice assistants: Enhancing consumer responses in voice commerce. Journal of Business Research, 175, 114566.