The effects of stress and chatbot services usage on customer intention for purchase on e-commerce sites

Abstract

The customer relationship is an inevitable part of the growth of a business. Customer services companies have shown a major interest in integrating Artificial Intelligence (AI) into their systems. A chatbot is an AI app that can commence a conversational session with a human partner, maintain, and handle a twisted and complicated conversation in a natural language. The main reason that businesses are interested in chatbots is that they have found chatbots as a solution to reducing customer service costs as well as having the capability of handling multiple users simultaneously. Investigations on recorded chats can lead companies to comprehend the contexts of messages, whether inquiries are informational or emotional. Therefore, sifting through messages can reveal “What is being said?”, “Was it positive or negative?”, “Was the customer angry, happy, frustrated, stressed, etc?”. Previous research indicates that some emotions are more relevant in marketing, such as anger, which shows that the customer is active and has an optimistic view of the future, therefore, it is more likely to lead to action. The aim of this research is to power chatbots with algorithms that can determine a potential buyer from customers’ chats to offer them a sale. To reach our goal, detecting the potential customer from the chat is the main challenge that we have to overcome. Discovering emotions from chat will direct us to understand more about customers’ intention to purchase or accept an offer. Experimental (empirical) research is defined as data-based research which relays on experiments or observations. Moreover, in experimental research, a verifiable conclusion should be generated by the researcher. Therefore, we developed a hypothesis and established an experimental design to prove or disprove it. The Null Hypothesis (H0): There is no relation between user emotion to their online buying decision-making. The Alternative Hypothesis (H1): User emotions play a significant role in online purchasing decision-making. To prove or disprove this hypothesis, experimental research with a positive approach has been designed. The goal of this experimental research is to find out whether there is a relation between users’ emotions and their purchasing decision-making process. We found four datasets that are labelled with emotion tags and then filtered them based on the conversation about purchasing (both accepting and declining purchases).

The first dataset is called EmotionLines which is about Friends TV Series and has each utterance labelled by relevant emotions. The second dataset is called CARER which is generated from tweets from Twitter, labelled based on Emotional Hashtags. The third dataset is called GoEmotions which built its dataset manually and contains 58K Reddit comments labelled. The fourth dataset is called EmotionPush which is more than 8K messages from Facebook Messenger, manually labelled with relevant emotions. Based on the four datasets (EmotionLines, CARER, GoEmotions and EmotionPush) we filter the sentences that have mentioned purchase and diagrams have been generated for each dataset. The result obtained doesn’t show a clear relation between any particular emotion or stress to purchase. However, Joy and Neutral emotions still have the highest number of purchases. Our hypothesis testing also support our findings and we had to reject the alternative hypothesis.