AI chatbots and customer experience - Master's thesis
This thesis examines what users expect from chatbots and when they choose them for customer service. I designed a Qualtrics survey, then prepared and analyzed the responses in Python.
Study design TAM/UTAUT and Herzberg’s two-factor theory informed hypotheses about perceived usefulness, expectations and satisfaction. The survey collected 98 responses; individual analyses use 85 to 89 observations depending on available data.
Findings In this sample, perceived problem-solving effectiveness and trust are associated with satisfaction. The association between perceived speed and satisfaction is not statistically significant. Respondents show a stronger preference for a human agent when requests are complex.
The study measures self-reported perceptions. Its sample and cross-sectional design limit generalization and do not establish causation.
Questionnaire and data
- Designed and distributed the Qualtrics questionnaire, collecting 98 responses.
- Defined engagement and customer-experience variables, including satisfaction, trust and willingness to recommend.
- Cleaned responses with Pandas and handled missing values for each analysis.
Hypothesis testing
- Ran linear and logistic regressions on associations between perceptions, satisfaction and stated intentions.
- Used Baron and Kenny mediation analysis to examine the role of satisfaction.
- Applied Welch and McNemar tests to compare perceptions and preferences by request complexity.