Kansei engineering for assessing emotional responses to document design
A second approach to exposing people’s feelings about stimuli has its origins in product engineering. Kansei Engineering (Nagamachi 1995) has been applied to products as diverse as footwear (Solves et al. 2006) and cars (Nagamachi 1999). The core of the method is a semantic differential experiment where participants are asked to rate candidate designs against a series of bipolar adjectives (e.g., attractive/ not attractive, traditional/not traditional) chosen for the purpose. Figure 27.2 shows an example of such a set of adjectives.
Figure 27.2: Example score card for the Kansei method (from Delin et al. 2007:16)
Statistical analysis then tells the researchers which design features correlate with positive reactions from the participants. Because corporate language and design, as we have seen, combine to create “designed artefacts” from websites to utility bills, this method of design evaluation could be applied to a wide range of communication types, or to samples of language alone.
Delin et al. (2007) discuss how to improve the Kansei method by making the choice of adjectives to be used in testing more principled and more useful. Here it is important to note, first, that the “adjectives” could be adjectival phrases or even clauses, so that the concepts it is possible to test are not limited by what is lexicalised as a single word in the language. The choice of adjectives or phrases is constrained in other ways, however. It is important to avoid ambiguous adjectives, for example, as different participants may choose different meanings to react to, causing a split score for that adjective. Adjectives must be well spread across a spectrum of meanings to avoid giving too much weight to one “semantic cluster” of document attributes. And there are adjectives that can confuse participants in a particular context; for example, in one study, the adjective “oppressive”, translated into English from Japanese, was presented to participants asked to judge the design of a wristwatch.
Delin et al.’s revisiting of the Kansei method applied linguistic principles to the selection of adjectives for evaluating laundry-product packaging. They isolated the importance of testing three areas, all of which are relevant not just for packaging, but for the design of communications:
- - Functional qualities of the item being examined (in the case of a communication, these might be usability, ease of understanding, completeness, relevance);
- - Qualities of presentation (e.g., design, use of space, attractiveness, visual clarity, size, print quality;
- - Qualities relating to the brand and its “personality” (e.g., genuine, friendly, kind, authoritative, moral).
Initially, examination of brand materials and discussion, including talks with brand owners about their ambitions for the communication in terms of brand qualities, yield a first set of words. These can then be used as “seed” words to generate other candidate words and phrases from a corpus. As Delin et al. describe, the selection of further words for testing is based on the hypothesis that words with similar meanings will be distributionally similar, that is, they will share a significant number of other words occurring in their context. Each seed word is input into the British National Corpus (or, if preferred, another relevant corpus) to identify other adjectives that have similar lexical behaviour in naturally occurring language. The method then produces a list of the most significant collocations for each sufficiently frequent word from the corpus chosen. The method can be used with any relevant corpus, such as a Web-derived representative corpus (Sharoff 2006). It can be also employed to conduct a study across different locales by using test adjectives in different languages that are generated from scratch, rather than by translating existing Kansei test materials, which may not be suitable for the purpose.
Each seed word results in 10-20 significantly related words, a list that can then be reduced by removing duplicates, and by editing down with the help of the brand owner. Stimuli can be tested against the final set of adjectives (for example, clear/not clear, friendly/not friendly, bright/not bright, warm/not warm, etc.) by asking participants to rank each communication on the appropriate number of scales. This yields a statistically analysable set of results (assuming a sufficient number of participants), thus turning qualitative evaluation into quantitative data. Ranking the criteria for importance can further help the researcher to order the “winning” documents or samples.