Do Anthropomorphic Chatbots Influence our Financial Decision-Making?

Conversational agents (CAs) are now used almost everywhere, using text, speech, and other channels to communicate with users. But how do CAs influence decision-making when they show features that usually only humans have?

When we search for information, browse the Internet, or use entertainment platforms, we use CAs consciously and almost naturally. Besides conscious usage, we also interact with CAs in places where we don’t expect them. In call centers, for example, voice-based CAs capable of processing and using natural language are deployed for routing users to the right contact. Another application of CAs is chatbots welcoming visitors on websites, providing initial assistance, or guiding them through the web offering.

We know from different studies that human decisionmakers tend to attribute human characteristics (for example, a consciousness or emotions) to objects if they have human-like features, such as a voice or a name. This phenomenon, known as anthropomorphism, has consequences. The anthropomorphic design of CAs may influence behavior in human-machine interactions, for example, by increasing trust and emotional attachment, and thus, increasing for instance the willingness to pay

• A conversational agent increases service satisfaction with a website.
• CAs have implied social presence to a certain degree, and this can have both positive and negative effects.
• Social presence has a positive effect because it increases trust in the CA.
• However, the social presence of CAs can also have a negative effect if users feel observed by the CA.
• The presence of a voice at the CA has no effect.

In different experimental studies, NIM currently investigates the extent to which CAs are perceived as anthropomorphic, depending on their human characteristics, especially when using speech. The central research question is whether CAs are accepted in different financial decision-making contexts. In our studies, participants make investment decisions either on a standard microcredit platform (rational decision) or on a prosocial microcredit platform (intrinsically motivated emotional decision). In both cases, a CA with different anthropomorphic features serves as a filter and decision support system that guides through the preference selection process.