18-01-2013, 03:09 PM
Measuring Emotions
Measuring Emotions.pdf (Size: 357.61 KB / Downloads: 143)
INTRODUCTION
Our emotions play an important role throughout the span of our lives because they enrich virtually all of our waking moments with either a pleasant or an unpleasant quality. Cacioppo and his colleagues wrote that “emotions guide, enrich an ennoble life; they provide meaning to everyday existence; they render the valuation placed on life and property” (Cacioppo et al. 2001 p. 173), which illustrates that also the relationship with our physical world is emotional. It therefore doesn’t come as a surprise that consumer researchers have found that emotions evoked by products enhance the pleasure of buying, owning, and using them (Hirschman & Holbrook 1982). In addition, it has often been argued that the experiential or emotional quality of products is becoming more and more important for differential advantage in the marketplace because products are now often similar with respect to technical characteristics, quality, and price. In some purchase decisions, emotional responses may even be a decisive factor. Clearly, the ‘fun of use,’ i.e. the fun one experiences from owning or using a product, belongs to this affective rather than to the rational domain. The difficulty in studying affective concepts as ‘joy of use’ or ‘fun of use’ is that they seem to be as intangible as they are appealing. Even more, rather than being an emotion as such, ‘having fun’ is probably the outcome of a wide range of possible emotional responses. Imagine, for example, the fun one has when watching a movie. This person will experience all kinds of emotions, such as fear, amusement, anger, relief, disappointment, hope, etcetera. Instead of one isolated emotion,
P.M.A. Desmet (in press) Measuring Emotions
Delft University of Technology; Department of Industrial Design
it is the combination of those emotions that contributes to the experience of fun. It is not implausible that the same applies to other instances of fun, whether it is sharing a joke, using a product, or interacting with a computer.
Knowledge of the process of emotion, i.e. how emotions are evoked, can enhance our understanding of what makes us enjoy interacting with a computer or, for that matter, with any other kind of product. So far, however, little is known about how people respond emotionally to products and what aspects of design or interaction trigger emotional reactions. In my view, an instrument that enables us to measure emotional responses can support the study and exploration of relationships between subjective affective responses, and objective interaction and design characteristics. Given this application purpose, the instrument should be able to measure subtle (i.e. low intensity) emotions, and mixed emotions (i.e. more than one emotion experienced simultaneously). In addition, our research requires an instrument that is applicable cross-culturally and therefore language independent.
The quest for instruments to measure emotions has a long history. Traditionally, attempts to measure emotions have been done in the field of psychology and sociology. More recently (i.e. the last twenty years), acknowledging the important role of emotions in their field of research, consumer and marketing researchers have developed instruments which measure the emotional responses to advertisement and consumer experiences. Even more recently (i.e. the last ten years), and as a result of the rapid invasion of computers into our daily lives, computer science has also become a player in the field of measurement of emotions. Unfortunately, none of these developed instruments appears to be applicable for the measurement of emotional responses to products because none meets all the above stated requirements. Given the limitations of existing instruments for the current measurement aims, a new instrument was developed: the Product Emotion Measurement instrument (PrEmo). This development was an iterative design process. Over a period of five years, six versions of the instrument have been created. The creation of each version was followed by an evaluation of its strengths and weaknesses, and on the basis of this evaluation renewed starting points were defined. Each subsequent version was designed to correct the flaws of the previous version. This chapter presents the development, validation, and application possibilities of the sixth PrEmo version. Note that the development of all PrEmo versions is reported in detail in Desmet (2002). Earlier versions have also been discussed in previous publications (see e.g. Desmet & Hekkert 1998; Desmet, Overbeeke, & Tax 2001; Desmet, Hekkert, & Jacobs 2000).
2. APPROACHES TO MEASURE EMOTION
Before one can measure emotions, one must be able to characterise emotions and distinguish them from other states. Unfortunately, this is a problem that currently belongs to those yet unsolved. Although the concept of emotion appears to be generally understood, it is surprisingly difficult to come up with a solid definition. When surveying emotion research in psychology, one finds various traditions that hold different views on
P.M.A. Desmet (in press) Measuring Emotions
Delft University of Technology; Department of Industrial Design
how to go about defining, studying, and explaining emotions. The last 100 years, psychologists have offered a variety of definitions, each focussing on different manifestations or components of the emotion. As there seems to be no empirical solution to the debate on which component is sufficient or necessary to define emotions, at present the most favoured solutions is to say that emotions are best treated as a multifaceted phenomenon consisting of the following components: behavioural reactions (e.g. approaching), expressive reactions (e.g. smiling), physiological reactions (e.g. heart pounding), and subjective feelings (e.g. feeling amused). Each instrument that is claimed to measure emotions in fact measures one of these components. As a consequence, both the number of reported instruments and the diversity in approaches to measure emotions is abundant. Today’s instruments range from simple pen-and-paper rating scales to dazzling high-tech equipment that measures brain waves or eye movements. In this chapter, the distinction is made between non-verbal (objective) instruments and verbal (subjective) instruments.
Non-verbal instruments to measure emotions.
This category comprises instruments that measure the either the expressive or the physiological component of emotion. An expressive reaction (e.g. smiling or frowning) is the facial, vocal, and postural expression that accompanies the emotion. Each emotion is associated with a particular pattern of expression (Ekman 1994). For example, anger comes with a fixed stare, contracted eyebrows, compressed lips, vigorous and brisk movements and, usually, a raised voice, almost shouting (Ekman & Friesen 1975). Instruments that measure this component of emotion fall into two major categories: those measuring facial and those measuring vocal expressions. Facial expression instruments are based on theories that link expression features to distinct emotions. Examples of such theories are the Facial Action Coding System (FACS; Ekman & Friesen 1978), and the Maximally Discriminative Facial Moving Coding System (MAX; Izard 1979). Generally, visible expressions captured on stills or short video sequences are analysed. An example is the Facial Expression Analysis Tool (FEAT; Kaiser & Wehrle 2001). Also subtle expressions, invisible to the naked eye, can be measured because the facial muscle activity of these expressions result in electrical potentials (facial electromyographic activity: EMG). This EMG activity can be assessed by determining the voltage from two electrodes placed on the skin’s surface over a particular muscle group (see Cacioppo & Petty 1989). Like the facial expression instruments, vocal instruments are based on theories that link patterns of vocal cues to emotions (e.g. Johnstone & Scherer 2001). These instruments measure the effects of emotion in multiple vocal cues such as average pitch, pitch changes, intensity colour, speaking rate, voice quality, and articulation.
A physiological reaction (activation or arousal, e.g. increases in heart rate) is the change in activity in the autonomic nervous system (ANS) that accompanies emotions. Emotions show a variety of physiological manifestations that can be measured with a diverse array of techniques. Examples are instruments that measure blood pressure responses, skin responses, pupillary responses, brain waves, and heart responses. Researchers in the field of affective computing are most active in developing ANS
P.M.A. Desmet (in press) Measuring Emotions
Delft University of Technology; Department of Industrial Design
instruments, such as IBM’s emotion mouse (Ark, Dryer, & Lu 1999) and a variety of wearable sensors designed by the Affective Computing Group at MIT (e.g. Picard 2000). With these instruments, computers can gather multiple physiological signals while a person is experiencing an emotion, and learn which pattern is most indicative of which emotion.
The major advantage of non-verbal instruments is that, as they are language-independent, they can be used in different cultures. A second advantage is that they are unobtrusive because they do not disturb participants during the measurement. In addition, these instruments are often claimed to be less subjective than self-report instruments because they do not rely on the participants’ own assessment of the emotional experience. For the current application however, this class of instruments has several limitations. First, these instruments can only reliably assess a limited set of ‘basic’ emotions (such as anger, fear, and surprise). Reported studies find a recognition accuracy of around 60-80% for six to eight basic emotions. Moreover, these instruments cannot assess mixed emotions. Given these limitations, it was decided not to use this approach for measuring emotions evoked by products.
Verbal instruments to measure emotions.
The above mentioned limitations of non-verbal instruments are overcome by verbal self-report instruments, which typically assess the subjective feeling component of emotions. Subjective feeling (e.g. feeling happy or feeling inspired) is the conscious awareness of the emotional state one is in, i.e. the subjective emotional experience. Each emotion involves a specific feeling which is a basic, irreducible kind of mental element (Titchener 1908). These subjective feelings can only be measured through self-report. The most often used self-report instruments require respondents to report their emotions with the use of a set of rating scales or verbal protocols.
The two major advantages of the verbal instruments is that rating scales can be assembled to represent any set of emotions, and can be used to measure mixed emotions. The main disadvantage is that they are difficult to apply between cultures. In emotion research, translating emotion words is known to be difficult because for many emotion words a one-to-one, ‘straight’ translation is not available. Between-culture comparisons are therefore notoriously problematic. To overcome this problem, a handful of non-verbal self-report instruments have recently been developed that use pictograms instead of words to represent emotional responses. An example is the Self-Assessment Manikin (SAM; Lang 1985). With SAM, respondents point out the puppets that in their opinion best portray their emotion. Although applicable in between-culture studies, these non-verbal scales also have an important limitation, which is that they do not measure distinct emotions but only generalised emotional states (in terms of underlying dimensions such as pleasantness and arousal). It was therefore decided to develop a new instrument for emotions evoked by products. This instrument was developed to combine the advantages of existing non-verbal and verbal self-report instruments: it measures distinct (and mixed) emotions but does not require the participants to verbalise their emotions.