Sentiment analysis has been widely used to study the effects of messages on social media. However, prior research has primarily focused on the positive or negative tone, overlooking the rational-emotional dimension embedded within these messages. According to dual-process models of thinking, this overlooked dimension may influence how information is processed. This study introduces the rational-emotional spectrum into sentiment analysis and examines how its interaction with negative-positive intensity and intergroup communication affects the popularity of political posts on social media. We compiled a dataset of approximately 400,000 tweets posted by U.S. congressional politicians between February and August 2021 and employed multiple deep learning models to estimate the negative-positive intensity, rational-emotional spectrum, and intergroup communication score for each tweet. We will apply a Deep Instrumental Variables model to assess the relationships among these factors and expect to find that negative emotional expressions are associated with increased popularity—particularly in intergroup interactions.