AI-assisted Deceptive Hotel Review Detection

The purpose of this research study is to investigate how human interpret some predictive learned by AI systems when detecting deceptive hotel reviews and how AI explanations can shape human understanding. For instance, the AI system learns that words like "floor" and "bathroom" are highly predictive of genuine reviews and words like "husband" and "luxury" are highly predictive of deceptive reviews. While the AI system highlights these words by learning from the data, the association between these words and the outcome are not easy to comprehend for humans. Thus, our goal of this research is to investigate how participants interact with machine-generated explanations that are unintuitive, and how participants make use of our developed visualizations and tools to make sense of these unintuitive features.