Bot Basics, Chatbot Development

Survey Hints at the Importance of Chatbots’ Expertise for Gaining Users’ Trust

In order for customer service chatbots to be effective, it is important for human users to feel that they can trust them and that the information they provide is reliable

By Ingrid Fadelli
September 24, 2019

Several companies and institutions worldwide have recently started using chatbots to provide information or guidance to customers and others using their services. Although these customer service applications are becoming increasingly popular, only a few studies have investigated the perceptions of the people using them.

In order for customer service chatbots to be effective, it is important for human users to feel that they can trust them and that the information they provide is reliable. So far, however, the key factors and qualities that affect people’s trust in artificial agents have remained somewhat unclear.

Filling the gap

A recent study carried out by researchers at the University of Oslo and the Norwegian research Institute SINTEF could help to fill this knowledge gap by providing interesting insight into how developers can enhance users’ trust in chatbots. Their paper, “An Initial Model of Trust in Chatbots for Customer Service—Findings from a Questionnaire Study,” published by Oxford University Press, is part of the master’s thesis of Cecilie Bertinussen Nordheim, who was supervised by Senior Researcher Asbjørn Følstad and Professor Cato Alexander Bjørkli.

users expect from chatbots
What do users expect from bots?

“This research came about in response to a lack of knowledge regarding how trust in chatbots for customer service is formed,” Følstad, who also helped carry out the study, explained. “We believe that trust is important for users to engage with chatbots for customer service, hence it is important to know which factors that may determine such trust.”

In the past, researchers have conducted a significant number of studies investigating humans’ trust in technology, particularly in robots. The team at the University of Oslo and SINTEF wanted to add new insight to this pool of knowledge that specifically relates to user trust in customer service chatbots.

“We took as a starting point the factors previously identified as important to trust in technology,” Følstad said. “In addition, we wanted to explore the possibility that other factors may be relevant for chatbots, as we see these as a form of technology that may provide a novel user experience, for example, due to their relative likeness to chatting with real humans.”

Human input

The researchers asked 154 people recruited via four customer service chatbots to complete a questionnaire, in order to better understand their perceptions and observations after interacting with customer service agents. The questionnaire asked users to express their views and rate how much specific factors outlined by previous literature affected their trust in the agent they just communicated with. Finally, those participating in the survey were asked to explain what affected their trust in the chatbots, in their own words.

Do chatbots need a human touch?
Do chatbots need a human touch?

“The single most important takeaway from our study might be the importance of chatbot expertise on trust,” Følstad said. “For users to trust a chatbot, the chatbot needs to correctly interpret the users’ requests, providing relevant and correct answers. This is food for thought, suggesting that launching chatbots with immature capabilities may lead to reduced trust.”

Just like us

In addition to highlighting the importance of a chatbot’s expertise for gaining users’ trust, the findings collected by the team suggest that the “human-likeness” of chatbots for customer service (i.e., how much a chatbot’s communication resembles that of humans) may be somewhat less important for trust than what is often suggested. According to the researchers, this could be explained by the fact that customer service chatbots have only recently been introduced into society, thus they are typically only used for very brief interactions or dialogues.

“Quite possibly, human likeness may be of greater importance to trust in the future, when chatbots have developed to a point where they may engage the user in more long-lasting service processes,” Følstad said.

Essentially, the user feedback collected in this survey suggests that chatbots that are not sufficiently able to answer user queries struggle to gain human trust. In other words, a customer service chatbot’s expertise and its ability to provide answers that are both useful and relevant appear to be crucial for increasing human trust in artificial agents.

The trust factor

The researchers also used their observations to develop a model that outlines how chatbot-related (i.e., perceived expertise and responsiveness), environment-related (risk and brand perceptions), and user-related (propensity to trust technology) factors affect a human user’s trust in chatbots for customer service. They are now planning to continue investigating how users experience chatbots and how specific qualities of artificial agents can help to gain people’s trust. Their work ultimately could help to set the stage for chatbots, informing developers on how to provide the best possible user experiences.

“In our next studies, for example, we will be exploring different approaches to chatbot interaction design, how chatbots and human customer service representatives should interact, and how a chatbot may be designed to become a more important touchpoint in service companies’ key customer journeys,” Følstad said.