What can we learn from academic research for creating chatbot experiences?
Manage users’ expectations
Manage users’ expectations by making it clear whether it is a chatbot or a human to avoid unrealistic expectations.
Help users learn the rules of interaction with chatbots.
Use high interactivity for a positive user experience.
- Manage users’ expectations before they use a chatbot and make it clear whether it is a chatbot or a human because this affects expectations.
- Be aware that a chatbot appearing too human can lead to unrealistic expectations.
- Help users to learn the rules of interaction with chatbots: When the chatbot misunderstands the user, the user uses human-human communication repair strategies not adapted to chatbot interaction.
- Design for high interactivity in the chatbot communication because this creates a positive user experience.
Increase trust to increase intention to use
Increase users’ trust and intention to use by humanness without being anthropomorphic to avoid creepiness.
Create a personalised, highly interactive experience and use informal language.
Highlight the novelty and provide efficient task solutions.
- Chatbots should show human-like empathy and social presence, but not be anthropomorphic, to increase users’ trust and willingness to use.
- Highlight the novelty of the technology and enhance the perceived humanness to positively influence users’ performance and effort expectancy of the chatbot.
- Create a personalised experience with an efficient task solution to increase users’ trust and willingness to continue using a chatbot. Be aware that task efficiency can also increase the perceived creepiness of a chatbot.
- Increase users’ intention to use a chatbot by managing their expectations regarding performance and effort, and by increasing interactivity.
- Use informal language in chatbots to increase users’ intent to continue using them.
Give users a feeling of control and efficiency
Give users control, but at the same time avoid too many explanations.
Tailor the chatbot to the internet competency and personality type of the user.
Tailor the chatbot to the task (text, visuals, buttons, dropdowns, etc.).
Use a conversational survey because it is more enjoyable and intuitive.
- We need to look carefully at which cases a chatbot is the best solution and when a menu-based interface is better – or we can look into a combination of both.
- Give users decision control when interacting with AI because this improves their experience.
- Avoid too many explanations about how AI works because this can increase the perceived complexity and deteriorate the user experience.
- Consider the internet competency of your users and try to offer low competency users more task-oriented bots (they need more task-related guidance) and use social-oriented bots for users with higher internet competency.
- Adapt the chatbot conversation style (warm vs. competent) to users’ time orientation (present vs. future) for a more positive user experience.
Create not only smart, but also good chatbots to connect emotionally
Consider using a chatbot because they can elicit similar positive emotions and fewer negative emotions than a chat with a human.
Make the chatbot emotionally responsive and use virtual intimacy to improve the user experience.
Increase the humanness in chat agents by visual cues, identity cues, and conversational cues.
Align the chatbot with the user’s language to make the user feel at ease.
- For certain services, consider using an emotionally responsive chatbot instead of a human because they elicit similar positive emotions as a chat with a human, and fewer negative emotions and less conversational concerns.
- Use the concept of virtual intimacy for the digital agent because this creates positive emotions and a better user experience.
- For more effective conversations, increase the humanness in chat agents by visual cues (e.g., human figure), identity cues (e.g., human-associated names or identity), conversational cues (e.g., mimicking human languages, message interactivity, perceived helpfulness).
- In Human-AI teams (HATs), design the AI agent to be both smart and good, because the competence and warmth of a chatbot are crucial for acceptance by humans.
- A chatbot should try to align with users’ language and content to make the user feel more at ease by signalling liking and creating closeness.
Learn from human-human interaction
Use insights from human-human interaction to improve the chatbot experience:
Establish common ground in chatbot-human conversation.
Use the phases of a human service agent conversation: Open message with personalised greeting, express appreciation for the question or comment, provide information, say thank you.
Create a chatbot that can detect the mental state of the user.
- A chatbot should try to establish “common ground” when starting a chat with a user because human-computer dialogue resembles a dialogue between people from different cultures not sharing common ground. Both in intercultural and human-computer conversations, the common ground must be constructed during the dialogue (emergent).
- A chatbot should model the conversation with the user according to the four recurring moves in human-human customer service dialogues:
- Open a message with a personalised greeting
- Express appreciation for the question or comment
- Provide information
- Say thank you
- Create a chatbot with social cognition: one’s ability to understand and think about one’s and others’ mental states in social situations. The chatbot should be able to infer users’ mental states.