Applying Machine Learning in Personality-based Persuasion Marketing

Personality based persuasion marketing often use consumers’ psychological characteristics to influence purchasing behaviour. When done in an ethical manner it can bring benefits to all involved parties. Personality can be inferred from consumers’ information available online, such as text in electronic word-of-mouth (eWOM) or likes from social media. This paper utilizes natural language processing to extract consumers’ personality and opinions for eWOM and utilise these to optimize persuasiveness of advertising communication. The approach is based on the self-congruence theory postulating that consumers respond more positively to brands and messages that are in line with their personalities. Such insights can assist marketers in developing effective persuasion marketing strategies. The methodology combines text-based personality extraction with topic modelling, ensemble classification and explainable AI, and is applied on restaurants’ reviews from Tripadvisor. EWOM text is vectorized by expressing each review as a vector of topics’ proportions from the learned topic model. Personalities of consumers are recognised using a deep learning classifier, fine tuned on labeled data from the MBTI personality model. Four binary XGBoost (eXtreme Gradient Boosting) classifiers are trained and validated, one for each personality trait, using reviews’ embeddings (topics) as predictors and personality labels as class variables. The logic of each XGBoost model is extracted using an explainable AI technique, namely, Shapley Additive Explanations (SHAP). Shapley values form SHAP global explanations give rise to eWOM topics that could resonate more with each personality trait. Results can help marketers design advertising messages based on the consumer personality that they target. The findings are compared against the literature of persuasion marketing and consumer behavior.