A Survey on University Teachers' Trust and Risk Perception Regarding the Use of the AI Agent "Lobster (OpenClaw)"

This study takes OpenClaw intelligent agents as the research context, constructs an analytical framework based on technology trust theory and risk perception theory, and uses questionnaire survey method to conduct empirical research on 163 teachers from 8 universities across the country.

Research has found that university teachers exhibit a "limited trust" characteristic towards AI agents, with ability trust and goodwill trust significantly higher than integrity trust; Accuracy risk and privacy risk are equally high in risk perception, while teaching autonomy risk perception is actually the lowest; There is no significant negative correlation between trust and risk perception, but rather a relatively independent dual track structure; Trust and risk perception together explain 44.8% of the variance in usage intention, with ability trust and accuracy risk being the two strongest predictors, and accuracy risk positively predicting usage intention.