In this podcast episode, Edona Elshan tells us about her research into the interaction design of conversational agents and how the topic of chatbots is treated in research.

About Edona Elshan

Edona Elshan is a doctoral student and research assistant at the Institute of Information Systems at the University of St. Gallen. She conducts research in the field of conversation design with conversational agents. In her master’s thesis, she investigated how to design a process method that helps domain experts to create a chatbot. This should efficiently design and stimulate learning processes. She has focused on voice and text bots.

On the one hand, she looked at how the use case analysis should ideally look. In the next step, she investigated how to model the interaction, transfer it into a conversation and finally implement and test it.

Chatbots in research

Research is being conducted at the HSG in various areas of chatbot technology. Edona Elshan is currently working in a research team that is investigating how to conduct large-scale employee evaluations using a chatbot.

In another research group, they are taking a closer look at how chatbots can be used in the area of “learning”. The focus here is on coaching teams working in the field of innovation. Specifically, we are investigating whether it is easier to work in a team when a chatbot is present. The central questions are: How much can a bot intervene in a team constellation? And how do the team members react to this? Sometimes there are people who consider the bot to be unauthorized to give them instructions. According to Edona Elshan, it will be a while before this threshold is overcome.

On the one hand, this is due to initial frustrating experiences with chatbots, which have left many with a negative image of the technology. That’s why you have to try to show the technological progress of chatbots. On the other hand, it is also down to people, who are often rather skeptical about new technologies. In this case, you have to look at how to break through these mental barriers.

Scientific findings on the design and implementation of conversational agents

The use case

The use case has shown that it is very important to think about how and where the technology will be used right from the start. There must be a specific use case. It is important to consider what the capabilities are and what a bot can implement in a company.

Dialogs and content

When it comes to dialogs and content, we are dealing with a whole new class of systems from the perspective of business informatics. The question of what individual dialogs with the user should look like is becoming increasingly important. The bot is often operated by different user groups. You should therefore think about different dialog flows that are tailored to the respective user groups.

A major challenge is to make the dialogs more human, so that the user does not immediately notice that a machine is behind the bot.

Best Practices

When designing chatbots, Edona Elshan and her team use the dialog flow method. With dialog flows, you consider a start and end point for a specific use case. As an example of a starting point, someone wants to report a claim for his or her car on the website of an insurance company. The end point in this case would be that the customer has reported and registered the claim and knows how to proceed from there.

From the starting point, the next step is to draw a tree with various request and response scenarios. Various dialog strands can be built using these cases. If these are programmed or transferred into an interface, the chatbot can be built more efficiently. This method is common in research, as the individual user stories are then evaluated. Dialogue threads are also helpful for those who do not want to spend a lot of money or do not have the time to program themselves or try out a new tool.

Another option is to record the chatbot dialogue in a simple interface such as aiaibot and continue to optimize it.

How do you gain trust in chatbots?

To answer this question, Edona Elshan’s research team is investigating at what point in the online dialog the conversation with chatbots is interrupted. To this end, a study is currently being conducted to examine the limitations of online chatbots and how these can be overcome. The team’s aim is to draw up a recommendation for action for chatbots that could be improved on the basis of existing chatbots.

In order to support the development of chatbots as a user, it is helpful for everyone to leave honest feedback in the user reviews. In this way, users can make an important contribution to the optimization and future of good chatbots.

Learnings

1. people have often had frustrating initial experiences with chatbots that have left them with a negative image of the technology. Or they resist new technologies on principle. Research is therefore being carried out into how these mental barriers can be overcome.

2. when it comes to dialog design, we must increasingly ask ourselves what individual dialogs with customers will look like. You often have to deal with different user groups.

3. in the dialog flow methodology, you consider a start and end point for a specific use case. From the starting point, the next step is to draw a tree with different request and response scenarios, which are used to build different dialog strands.

4. users should be encouraged to leave feedback when interacting with a chatbot by implementing a corresponding function.

It’s best to listen to the podcast episode with Sophie Hundertmark and Edona Elshan for yourself. Have fun!

And if you don’t want to miss any more podcast episodes, then sign up for the mailing list at the bottom of this page.