DHL

For DHL we have evaluated the potential of Digital Humans as chat interface. The focus was on their ability to reduce escalation intent (the number of people who still want to speak to a live agent after the chatbot conversation is closed).  DHL also wanted to determine which type of digital human is most suitable, with special attention to the impact of visual realism (the extent to which an agent looks like a human) and behavioral realism (the extent to which an agent behaves like a human).

For this research, an online experiment was conducted, in which 1,500 consumers were exposed to six different types of conversational agents to test these aspects. The agents varied in their level of visual and behavioral realism.

Different agents that participants were exposed to:

• DHL's then current chatbot in the form of a robot. This agent is displayed with a static image in the chat environment, without movements or expressions.
• The same agent, but animated. The animated agent can talk, shows movements and can express emotions such as happiness.
• Woman (static): An agent with a human identity, represented by a static image of a woman.
• Woman (animated): An agent with a human appearance in the form of a woman, who is animated. This agent can talk, move and show emotions.
• Man (animated): An agent with a human appearance in the form of a man, who, like in the previous condition, is animated and is able to talk, move and show emotions.
• Static Non-Human Agent (Cat): An agent in the form of a cat, represented by a static image.

For the experiment, participants were asked to file a complaint through these agents, after which their experiences were evaluated based on several criteria, including: first impression of the conversation, various aspects of customer satisfaction, escalation intent, communication competence, problem solving ability, perceived credibility, perceived intelligence, liking, disliking, and brand fit.

Findings

The results of this project show that an animated, female Digital Human outperforms all other interfaces. Compared to a static chatbot avatar, the animated female scored significantly better in terms of first impressions, provided solution, problem-solving ability, and determinants of acceptance (likeability and intelligence).

Reported differences between the female Digital Human and the current chatbot ranged from 4.4% to 5.8% in favor of the Digital Human. The need for consumers to speak with a human representative was 4.8% lower after interaction with the animated female than after interaction with the current chatbot.

The full research is available upon request.