This post is an extract from the chapter “Approaches to emotion and sentiment analysis” that I contributed to The SAGE Handbook of Digital and Social Media Marketing, which was edited by Annmarie Hanlon and Tracy L. Tuten.
1. Introduction
Emotions are central to how consumers respond to marketing stimuli, and how they go on to behave, both in the physical and the online environments. For instance, research has shown that warm physical environments elicit positive emotions and that these, in turn, increase customers’ valuation of products; and that when Facebook users were exposed to posts that reflected a negative emotion, they were likely to go on to post content of a negative emotion, and that the opposite happened for positively charged posts. In addition, exposure to emotionally charged content online can also have impact on the offline world, as demonstrated by Bigne et al (2018)’s analysis of the impact of social media reviews on purchase of low-cost airline tickets. Not only do emotions shape consumer behaviour online and offline, but messages with high emotional content (e.g., anger) spread quickly and widely, triggering emotional responses in others, and driving further sharing behaviour.
Given the importance of emotions in the outcome of marketing initiatives, and the ubiquity of digital and social media in everyday life, marketing scholars as well as marketing practitioners are interested in identifying and analysing online displays of consumers’ emotions. This includes ad-hoc expressions of emotion, such as in social media conversations, as well as purposeful ones, such as crowd-sourced customer review platforms. The analysis of sentiment displayed in online content can assist marketers in anticipating consumer behaviour and, where pertinent, take remedial action. The latter goal is particularly relevant in the case of brewing social media crises, as these can have significant short, as well as long, terms effects on the brands at the centre of those conversations.
This chapter outlines techniques for studying sentiment online.
2. Studying emotions vs. studying sentiment
Although the terms “emotion” and “sentiment” are often used interchangeably, they actually refer to different psychological phenomena. The former is a socio-cultural expression of feelings, which occurs at the preconscious level, may be abstract and is fleeting in nature. In contrast, the latter is an emotional disposition, which occurs at the conscious level, is directed at an object, and develops over time. Zwebner et al’s (2013) study of the effect of temperature on consumer behaviour focused on the abstract emotions elicited by warmth; while Bigne et al (2018)’s work considered the verbalised assessments made by customers of their experiences with specific airlines and was, therefore, focusing on sentiments.
But what exactly are we talking about when we talk about emotion or sentiment analysis?
By and large, most studies of consumers’ emotion or sentiment focus simply on measuring the valence of the feelings – i.e., whether customers feel mostly positively or negatively. However, focusing on valence, only, will offer an incomplete understanding of the observed behaviour. For instance, Gelbrich (2010) showed that customers complained when an experience resulted in negatively valanced emotions of anger, but not when it resulted in frustration and helplessness, which are also negative emotions. Conversely, Grappi and colleagues (2015) found that willingness to pay increases with the positively valanced emotion of gratitude, but not with happiness which, too, is a positive emotion. As these examples show, when studying emotions and sentiments, we should pursue a fine-grained understanding of customers, which goes beyond the valence.
As demonstrated by the seminal work of Russell (2003), customer behaviour is influenced not only by whether customers are pleased or displeased, but also how strong that emotion or sentiment are (referred to as “arousal level”). Namely, positive emotions and sentiments with high levels of arousal result in exciting experiences; whereas positive emotions and sentiments with low level of arousal result in relaxing ones. In turn, negative emotions and sentiments with high levels of arousal result in distressing experiences; while negative emotions and sentiments, with low level of arousal, lead to boring ones. See Figure 1.
Figure 1. Typology of consumer responses to emotionally charged events
In summary, emotion analysis provides researchers with a measure of the abstract, fleeting feelings that customers may experience as a result of marketing efforts; while sentiment analysis provides researchers with a measure of the specific, sustained emotional disposition towards a product, event or phenomenon (e.g., vaccination). However, for a complete understanding of consumer behaviour, we should go beyond valence, and also seek to identify whether the emotion or sentiment are active or passive (i.e., high or low arousal levels) and, if possible, the specific emotion or sentiment experienced by customers.
Ana Canhoto 