Narrative Reasoning in Natural Language Dialogue Systems to Win User's Trust
Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Trust is a critical aspect in building and maintaining social relationships. Similar to trust in human to human relationships, trust plays a role when humans interact with intelligent computer systems (also called intelligent agents). Current literature discusses building trustworthy relationships by means of small-talk, gossiping and social role awareness, but they do not address the role of explanations of decisions and actions play in building relationships. A reasoning mechanism is required that establishes a relationship between human and computers by improving the level of trust between humans and agents. This thesis offers a reasoning mechanism for intelligent agents we refer to as Narrative Reasoning (NR). An agent that applies NR (NR agents) explains the reason behind its actions and decisions. An experiment is performed where NR and non-NR agents collaborate with participants to achieve a joint goal (here, purchasing a house in a virtual environment). The influence of narrative reasoning on the trust levels of participants in the test group is compared to trust levels with a control group where an agent does not reason narratively. The hypothesis that NR agents improve trust levels is partially supported. Participants that interact with a NR agent developed higher trust levels (on two of the semantic differential scales) than participants that interacted with an non-NR agent.
Place, publisher, year, edition, pages
2010. , 26 p.
IdentifiersURN: urn:nbn:se:su:diva-40650OAI: oai:DiVA.org:su-40650DiVA: diva2:325211
UppsokSocial and Behavioural Science, Law
Ramberg, Robert, Prof
Johnson, Maarit, Docent