| Author: | Xie, Wen |
| Title: | Two studies of human-AI interactions |
| Advisors: | Xu, Xin (MM) |
| Degree: | Ph.D. |
| Year: | 2023 |
| Subject: | Artificial intelligence Human-computer interaction Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Management and Marketing |
| Pages: | vii, 102 pages : color illustrations |
| Language: | English |
| Abstract: | In light of the vital role of artificial intelligence (AI) and the distinctive characteristics of AI (such as anthropomorphism, inexplicability, and natural language), AI can be extremely beneficial to us. However, human-AI interaction inefficiency is very common (Agomuoh, 2023; Cem, 2023). It is necessary for us to better understand how users and AI communicate and how users feel in a world where humans and AI co-exist. This thesis consists of two AI research. This first study discusses how chatbot designs affect the human-chatbot interaction process and, thus, influence human-AI interaction outcomes. Chatbots are gaining momentum in a variety of business functions, such as IT service, human resource management, customer service, and sales. Although chatbot proactivity design and social identity design are popular in the industry, limited existing research on chatbots has investigated chatbot proactivity design and its role as a boundary condition of chatbot social identity design. In order to optimize user evaluation on the service with chatbots, there is thus a need from both the business and the academia aspects to better understand how chatbot social designs affect the human-chatbot conversational process and, in turn, affect user perception and task success from the perspective of uncertainty reduction. The field experiment results reveal that proactivity design will decrease communication inefficiency and, thus, increase customer satisfaction and the probability of task success. Furthermore, chatbot social identity design weakens the negative effect on customer satisfaction made by ineffective communication. The second study investigates how the introduction of a hybrid human-AI service affects user evaluations of the service. AI and humans do have their own advantages when facing different tasks in digital platforms—for example, AI agents are better skilled in reliability, scalability, speed, accuracy, and generalization, while human agents are better skilled in creativity, judgment, and empathy (Rai and Sarker, 2019). A Human-AI hybrid service system can benefit from both AI and Human agents’ advantages, and it is of interest to study the impact of human-AI hybrid services in digital platforms. In particular, our study examines the associations between human-AI hybrid services and user evaluations in the context of a major audio streaming platform. On the platform, the introduction of AI podcasters creates the phenomenon—for the same audio, sometimes the AI voice version is released first, and sometimes the human voice version is released first. This provides us with a great opportunity to study how the introduction of a human-AI hybrid service affects user evaluations of the service from a temporal perspective. The empirical findings show that the user evaluations of those who are first exposed to a human service are improved with the presence of the human-AI hybrid service, and the user evaluations of those who are first exposed to an AI service get worse with the presence of the human-AI hybrid service. Our findings have important implications for both digital platforms and AI developers. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/12764