This thesis has focused on adaptation and long-term real world engagement as steps towards creating more personalized social conversational agents. The work is oriented towards dialogue designers and everyone who is involved with design of conversational agents: programmers, researchers, linguists, user experience experts and so on. We have demonstrated the design of three different agents for connecting with people, with each a special focus: multimodality, dialogue design and long-term interaction. The core of each of these systems is Flipper, a dialogue engine for the design of social conversational agents. We supplied the dialogue engine with design guidelines for social conversational agents. For the multimodal agent we focused on building a software toolkit for dialogue designers, as well as personalizing the dialogue by adapting to a user's emotion. The dialogue design agent consisted of writing dialogues and evaluating them with (end-)users. Lastly, we implemented a personalized question asking method for a higher engagement in long-term interaction in the real world. Though a full evaluative study is necessary, we found signs in our data collection and pilot study that most people enjoyed talking to the agents, regardless of the mistakes the agents made. Our recommendation is thus to focus more on dialogue design and accommodating repair strategies for misunderstanding, before implementing complex interaction models for social conversational agents.
|Qualification||Doctor of Philosophy|
|Award date||2 Dec 2021|
|Place of Publication||Enschede|
|Publication status||Published - 2 Dec 2021|
- conversational agents
- dialogue management
- dialogue design