Evaluating Cross-Cultural Advertising By Expert Systems: the Caas-Diagnostic System



Citation:

Franz-Rudolf Esch (1993) ,"Evaluating Cross-Cultural Advertising By Expert Systems: the Caas-Diagnostic System", in E - European Advances in Consumer Research Volume 1, eds. W. Fred Van Raaij and Gary J. Bamossy, Provo, UT : Association for Consumer Research, Pages: 87-98.

European Advances in Consumer Research Volume 1, 1993      Pages 87-98

EVALUATING CROSS-CULTURAL ADVERTISING BY EXPERT SYSTEMS: THE CAAS-DIAGNOSTIC SYSTEM

Franz-Rudolf Esch, University of the Saarland, Saarbrncken, Germany

INTRODUCTION

Expert systems in advertising are used for different purposes. The first one, MORE, developed in the USA in 1985, served as decision support system for the selection of recipients for direct advertising. This system developed for use in practice, however, is insufficiently documented for scientific purposes (Cook, Schlede, 1988 p.51,52). Besides this system, there meanwhile exist expert systems for advertising design, advertising evaluation and media planning, which were primarily developed in German- and English spoken countries (Rangaswamy, Burke et al. 1986; Burke, Rangaswamy et al. 1988; Burke 1991; Rossiter, Winter 1989, 1992; Mitchell 1987, 1988; Esch, Kroeber-Riel 1992; Esch 1990, Neibecker 1990). These systems are in different stages of development. Some - like ADEXPERT by Rossiter/Winter - are still in development, others - like the CAAS-diagnostic systems - are already used in practice. First validation studies comparing the results of expert systems - as for example recall - with pretest-results were made with encouraging outcomes (Neibecker 1990).

The question whether these expert systems can also be used in other countries, whether there are modifications necessary for such a cross-cultural use and to what extent they can contribute to a possible standardization of advertizing campaigns in different countries, has not been discussed so far. These aspects will be discussed for the CAAS-diagnostic system, an expert system for advertising evaluation. It has been developed within the research project Computer Aided Advertising System at the Institute for Consumer and Behavioral Research at the University of the Saarland under direction of Prof. Kroeber-Riel. It is already in use by 11 companies (among others P&G, Jacobs-Suchard, Ciba-Geigy, BASF etc.) which supported the project. The further updating and commercialization have been taken over by Allcomm Advertising AG, Basel, a subsidiary of Ciba-Geigy.

THE CAAS-DIAGNOSTIC SYSTEM

Structure and Proceeding

The CAAS-diagnostic system was developed with the hybrid expert system tool GoldWorks. It runs on a 386 IBM compatible PC with 16 megabyte working storage expansion. The CAAS-diagnostic system is a rule-based system with a data driven, but goal-directed inference mechanism. Goal-directed means that the advertising evaluation depends on the specific advertising objective. For this purpose, the advertising objectives are specified in the antecedents of the corresponding relevant rules. Data-driven means that the rules are linked in a forward chaining process, e.g. the consequence of a rule is only applied if the antecedents of the rule are satisfied. According to Winston (1987, S.198), forward-chaining is appropriate if "the objective is to uncover everything that can be derived from the data." This reasoning complies with our goal to present to the user after the evaluation procedure all positive and negative aspects of an ad.

The consultation between the expert system and the user takes place in an interactive way. Depending on the advertising objective, on the general advertising conditions (for example involvement) and on the advertising medium to be considered, the user is asked several questions to which he must answer. The expert system only considers strategic advertising goals, e.g. positioning goals like an informative or an emotional positioning or a mixed positioning which follows the means-end-scheme. At the end of a consultation the user receives an expertise in form of a verbal report with an overall judgement of the ad, the strengths and weaknesses of each construct of advertising effect and recommandations for optimization.

To acknowledge media specific characteristics, separate expert systems for magazine advertising, radio advertising, television commercials and newspaper- and supplements advertising were developed. For example, the CAAS-diagnostic system for TV commercials evaluates storyboards and animatics. Furthermore, the TV spot is divided in two intervals, the first 10 seconds and the remaining part because according to results of audience research, the first 10 seconds are of primary importance for the effectiveness of the TV spot (Capocasa, Denon 1985; Krugman 1986; Meyers 1986). Since TV is regarded as a low involvement medium (Krugman 1965), the expert system for TV commercials automatically assumes a low involvement situation (Lorson 1992). However, the different involvement of consumers is explicitely taken into account in the case of the CAAS-diagnostic system for advertising in magazines, journals and supplements.

The knowledge-base of the CAAS-diagnostic system

Theory-supported development of the knowledge base

Rossiter and Percy explicitely warn to use the knowledge of advertising professionals like David Ogilvy or Leo Burnett since this kind of knowledge is mostly characterized by one-sided subjective experiences. Therefore, it cannot provide the necessary theoretical bachground (Rossiter, Percy 1987). For that reason, most developers of expert systems in the field of marketing refer to knowledge obtained from publications (Wierenga 1990, p. 9). Should one refer nevertheless to a human expert, his knowledge is then transformed in a theoretically based model of the subject rather than developing a model of human expert knowledge or rather than simulating his way of problem solving (Horowitz, Russo 1989).

The theory-supported development of the CAAS-diagnostic system was also founded upon these considerations. The procedure of formalization of knowledge for a knowledge base whereby theories and practices of a discipline are transformed in an expert system, will be described in the following sections.

The formalization of knowledge is problematic because knowledge as represented in colloquial speech and professional terminology is superficial knowledge that is often expressed in unprecise and inconsistent terms. Additionally, it is not sufficiently operationalized. This superficial knowledge thus represents the raw material still to be adapted for the knowledge base. Example: in publications about ad effects, the following statements coinciding with many experts' opinion and being frequently considered as so-called rules of thumb for ad evaluation, are often used (vgl. Kroeber-Riel, Esch, 1992 p. 290-292):

S1: Pictures with persons enhance the viewing time of ads.

S2: Pictures with persons increase exchangeability.

S3: Exchangeable ads are looked at more shortly.

These empirical statements cannot be directly transformed into rules. Yet, one cannot simply ignore them as contradictions or only partially integrate them into the system. The hidden hypotheses beneath the surface of these empirical statements have to be laid down in theoretical constructs. It then becomes evident that

- Pictures with persons in S1 is an empirical indicator for the construct "emotional stimulus".

- Pictures with persons in S2 is an indicator for the construct "schema consistent illustration".

A schema consistent illustration can be understood as a picture that exactly matches an inner schema and that is therefore stereotypical and boring (Fiske, Linville 1980; Alba, Hasher 1983; Venkatraman, Villareal 1984; Kroeber-Riel 1991). These illustrations are quite common in advertising and lead almost necessarily to exchangeability.

S1 represents therefore a special case of the hypothesis "Emotional stimuli increase the activation potential of an ad". Together with the hypothesis "Illustrations with higher activation potential are looked at longer", it follows the logical conclusion "Ads with pictures with persons increase the viewing time". We can therefore formulate the rule

S1*: If persons are depicted in an ad, then the ad will be looked at longer.

S2 also can be transformed into a rule if it is connected with additional conditions. A reformulation of this sentence leads to the rule:

S2*: If persons are depicted in ads in a schema inconsistent way, the viewing time increases (Meyers-Levy, Tybout 1989; Caudle 1989; Kroeber-Riel 1990).

In addition, another confining rule can be found within that context: Schema consistent pictures with persons increase the viewing time as compared to ads without activation stimuli.

This example shows that one cannot use superficial knowledge and transform it 1:1 into the knowledge base of an expert system. Therefore, the request for a theory-supported information search is crucial to the development of a knowledge base.

But even a knowledge base developed from theory is fuzzy. The reason is that different sources of knowledge have to be harmonized whereby no theory can cover the entire advertising knowledge required. Furthermore, even in empirical research, the relations between independent and dependent variables are never completely conceived; in addition, the research is carried out under different conditions and test designs and analyzed with different statistical methods (Burke et al. 1992; Esch 1990). The comparability and harmonization of those results is therefore difficult and requires a certain degree of abstraction which necessarily results in a certain fuzziness. We accomodate ourselves to these circumstances by two factors. First, the relations between constructs of ad effect are considered in relatively simple degrees, for example strong-medium-weak. Second, only those factors of influence are aggregated to the results of certain constructs of ad effect which can be regarded as proven, based either on theoretical or on empirical evidence. Other factors of influence that were found only in a single study or those where a theoretical underpinning is missing, are not part of the aggregation. They will later only be placed at the user's disposal in the verbal print-out of the expertise as an indicator for a possible causal relationship.

We believe that such a proceeding is more useful than a mechanism as implemented in the expert system ADDUCE. In that system, the user's input data are compared with the knowledge stored in the system. The data of the experiment that comes closest to the input data is presented to the user. In case the facts do not exactly match the empirical studies contained in the knowledge base, one or two similar studies are offered to the user as an evaluation result. To what extent the experimental results, which match the respective situation more or less closely, can be used for evaluation, has to be considered by the user who will probably lack the theoretical knowledge of ad effect relations (Rossiter, Winter 1992). Furthermore, not all possible ad situations can be backed up by corresponding research results. For that reason, the system cannot provide assistance in some cases.

Structure of the knowledge base: the hierarchical model of advertising effectiveness

The expert system's knowledge base is constructed in a modular way. It consists of two parts: a part for advertising strategy and one for social techniques. Social techniques are devices to design social systems by means of principles found by social sciences (Kroeber-Riel 1991). In the strategic part, it is checked whether the advertising goal is appropriate under the given advertising conditions, whether the ad is appropriate for the target group and the company's corporate identity, whether the positioning is unique and aspects concerning integrated communication are checked. The evaluation of the social technical part comprehends the classical constructs of consumer and advertising research. Yet, these advertising constructs are arranged in a modified form (Petty, Cacioppo 1983; Petty, Cacioppo, Schumann 1983; Batra, Ray 1986; MacKenzie, Lutz, Belch 1986; Mitchell, Olson 1981; Kroeber-Riel 1990: Ray 1973; Rossiter, Percy 1987). Basically, there is a division in two aspects which are both crucial to advertising (table 1):

1. Is the ad powerful enough, has it sufficient impact, e.g. can it exist in a competitive environment and in comparison with competitive ads so that it is noticed and can be remembered?

2. Is the "goal achievement" thru advertising fulfilled, e.g. can the brand and its key message be learnt and is the acceptance by the target group guaranteed?

The impact of an ad is therefore a necessary condition, the "goal achievement" a sufficient one for a potential ad effectiveness.

Main differences of the CAAS-diagnostic system compared to other expert systems

The expert systems for ad evaluation in the English spoken countries, ADDUCE and ADEXPERT (Burke 1991; Winter, Rossiter 1992; Rossiter, Winter 1989) essentially try to answer the question how effective an ad will be. This is an evaluative kind of ad evaluation. The "how" refers to the question whether the respective ad changes the attitude towards the brand, e.g. roughly spoken, whether the attitude is improving, stays the same or is deteriorating. Yet, the development of such evaluative systems has some disadvantages for the use in practice; we want to point out two important ones:

1. It is not clear, why an ad leads for example to an improvement of attitude or not.

2. In case of a negative result, it is difficult to derive from such an evaluative system, whether the ad in question should be used at all or whether improvements of single elements of design could produce the desired result.

Because of these disadvantages, the CAAS-diagnostic system does not only deal with the evaluative question "How effective will a specific advertising medium be?". It also explicitely deals with the diagnostic question "Why will the advertising medium be effective in one way or the other?". Only by answering this question, the user will have the possibility to optimize the respective advertising: He obtains an analysis of strengths and weaknesses to each construct of ad effect being part of the knowledge base. With the help of these constructs of ad effect, deficiencies of knowledge representation can also be found more easily. By comparing the results of each construct of ad effect with corresponding results from ad pretests, divergences can be faster localized and analyzed.

TABLE 1

KNOWLEDGE BASE OF THE CAAS-DIAGNOSTIC SYSTEM: THE HIERARCHICAL MODEL OF ADVERTISING EFFECT

General problems when evaluating advertising with the CAAS-diagnostic system

Problems of evaluation

From our experiences with the CAAS-diagnostic system the system's user is a crucial bottleneck for the evaluation of ads. In general, following statements can be made:

- The deviation of answers evaluating the same ad is the greater, the more different the knowledge of the users is (for example experts vs. novices).

- The deviation of answers for evaluation of the ad execution (= social technical part) is greater than for evaluation of ad concepts (= strategical part).

- The deviation of answers evaluating questions to the same ad is the greater, the greater the answers depend on the subjective values and judgements of the recipient.

In the last case, the use of results is the more problematic, the stronger the users' subjective values differ from those of the target group. Yet, especially emotional aspects of an ad like pleasantness or irritation aroused by an ad depend to a large extent on the subjective values of the recipient.

Approaches to problem solving in the CAAS-diagnostic system

We accomodate ourselves to the different user knowledge by providing helps like definitions and explanations. In addition, there are suggestions for measuring specific ad factors such as an ad's contrast. However, these optionally available verbal hints are not always sufficient for demonstrating certain problems. Therefore, the CAAS-diagnostic system offers pictorial examples representing cultural norms for specific ad factors which the user can automatically call upon from a U-MaticVCR connected to the expert system. For example, when dealing with questions concerning the contrast of a picture, he obtains norm examples with strong, medium and weak contrast. Depending on the user's state of knowledge, both the explanations and the pictorial examples can differ in scope and intensity. The pictorial examples can also be adapted to product categories that are of interest to the company. Thus, the users obtain valuable assistance for evaluation. Yet, the explanations cannot - as would be possible in the case of a human expert (Dreyfus, Dreyfus, 1986) - gradually be adapted to the specific needs and skills of the respective user.

A further strategy to reduce subjective judgements is an operationalization of advertising constructs that can be grasped in single objective criteria as refined as possible. Example: According to findings of research in eye-mark-recording, ads in magazines are regarded on the average for about 2 seconds. This is the typical low involvement situation. To find out whether the key message in the headline can be picked up within that time, an over-all question could be used. In such a case, it would be more useful, however, to choose another procedure and to operationalize the information processing of the key message in the headline. Among others, this depends on

- how many words the headline has,

- how large the headline is as compared to the entire ad,

- how the headline's contrast is as compared to the background,

- whether the headline is designed in an activating way (f.ex. thru the use of colors),

- whether rare types are used rendering reading more difficult,

- whether other ad elements distract from the headline,

- whether a clear and precise language is used etc.

Through such an operationalization in various factors of influence, the information processing of the key message in the headline can be analyzed in a much more objective way than through an over-all question. Through the following aggregation of the input data, a result can be derived from the information processing of the key message in the headline.

Such a proceeding can only be realized, however, if an additional operationalization based on the actual knowledge is possible and if this also leads to a more objective assessment. For some constructs - such as acceptance - an operationalization is not yet possible. Also pictorial examples - which basically can be interpreted as norm values for specific ad factors - are not always helpful in these circumstances. While norm examples about a picture's contrast are valid across target groups and cultures, the same is not necessarily true for the acceptance of an ad when considering for example irritation or pleasantness. In these cases, differences in target groups often differ on the judgement of ad effect: therefore, the norm examples have to be adapted to the several target groups in an intracultural way.

GENERAL POSSIBILITIES FOR EVALUATING CROSS CULTURAL ADVERTISING

Two basic questions arise concerning the use of the CAAS-diagnostic system for a cross-cultural evaluation of ads:

1) Is it possible with the CAAS-diagnostic system to decide whether an advertising campaign can be used in different countries?

2) Can the CAAS-diagnostic system be used in different countries in order to evaluate ads?

Apart from the different cross-cultural use of expert systems for advertising evaluation, the following general hypothesis can be formulated:

The larger the cultural distance between countries, the smaller is the possibility of using a uniform expert system for ad evaluation.

The cultural distance between countries could be for example operationalized by different attitudes and values, different market and communication conditions like different media structures, different levels of technological development etc. To what extent cultural distance determines possibilities and limits of cross-cultural evaluations by expert systems, cannot be examined in general. Instead, the following aspects have to be discussed separately:

- Is it possible to evaluate questions on a conceptual level as well as on an executional level on a cross-cultural basis?

- Is the knowledge base e.g. the hierarchical model of ad effect, appropriate for cross-cultural evaluation?

- Is the surface design of the expert system appropriate for cross-cultural evaluation of ads?

- Which influence has the different user behavior in the case of cross-cultural evaluation of ads?

Especially the consideration of user behavior is important for evaluating from one country whether or not an ad campaign can be transferred to other countries. Factors like acceptance, as already discussed above, that are mainly characterized by subjective values, do not allow a judgement from one country whether an ad can be used in other countries too. In such a case, even with a data base containing pictorial norm examples from other countries as an assistance for judgement, the user would not be able to make a valid judgement. Problems already arising in one country when evaluating an ad's acceptance for a specific target group, become even more complex when cross-cultural differences have to be taken into account. The expert system user in one country is unable to consider cross-cultural differences especially concerning advertising constructs highly dependable on subjective values. Because of this, in many cases the user in one country will be unable to evaluate whether an advertising campaign can be transferred to other countries. Therefore, such an evaluation might only be possible if the expert system is directly used in the respective countries. This is the reason why the focus of the following discussion will be the problem of ad evaluation with the CAAS-diagnostic system in different countries.

POSSIBILITIES AND LIMITS OF AD EVALUATION WITH THE CAAS-DIAGNOSTIC SYSTEM IN DIFFERENT COUNTRIES

Transferability of evaluation levels

Ads can be judged on different levels, each possibly having shortcomings (Kroeber-Riel, Session A 7):

- on the conceptual level where the main strategic aspects of an ad are stated in a verbal form, for example the underlying positioning concept, the consideration of the target group, the competition and the corporate identity of the company itself,

- on the executional level where the strategic concepts are specified through the design of the respective advertising.

The relation between conceptional and executional level as well as between strategical and socialtechnical evaluation, as being realized in the CAAS-diagnostic system, is shown in a comprehensive way in table 2.

TABLE 2

RELATION BETWEEN AD EVALUATION LEVELS AND STRATEGIC AND SOCIALTECHNICAL PART OF EVALUATION OF THE CAAS-DIAGNOSTIC SYSTEM'S KNOWLEDGE BASE

TABLE 3

TRANSFERABILITY OF THE CONCEPTUAL AND THE EXECUTIONAL EVALUATION LEVEL

On the conceptual level, an harmonization between basic positioning and the company's corporate identity as well as the ability to be sufficiently distinct from the competitors can be verified. Just so, the adequacy of the general advertising goal can be evaluated since the criteria for such a decision are valid across countries. For example, the appropiateness of an ad objective as an emotional positioning depends, among others, on following points:

- whether the market for the respective product is saturated

- whether the consumer is barely interested in objective functional product qualities because these aspects are about the same across products

- whether the consumers only have a low product- and brand involvement

- whether the consumers only have a low cognitive but a high emotional involvement etc (Kroeber-Riel 1991; Mittal 1989; Park, McClung 1986; Bloch 1983; Mittal, Lee 1988).

These evaluation criteria are valid across cultures. Therefore, a transfer on the conceptual level is possible despite cultural distances between countries.

Yet, this is not true for the executional level. The elements of ad design - especially emotional ones - largely depend on the recipient's subjective values. Since these values differ the more, the greater the cultural distance, one and the same realization cannot automatically come into use in different countries. Instead, modifications with respect to the knowledge base as well as to the surface design of the CAAS-diagnostic system are necessary. (Table 3).

Transferability of the Knowledge Base

The following questions have to be checked when transferring the knowledge base:

- Are the single advertising constructs specified in the knowledge base also appropriate for other countries?

- Are the advertising constructs specified in the knowledge base sufficient for an evaluation in other countries or is it necessary to add new constructs?

- Are the relations between independent and dependent variables as represented in the knowledge base also appropriate for other countries or do they have to be modified?

- Is the strength of the relations between independent and dependent variables, as represented in the knowledge base, appropriate for other countries?

The extent to which modifications are necessary is expressed in the above order, being very large at first and decreasing as one goes further down. In general, we can assume that the model of ad effectivenes which is the foundation of the CAAS-diagnostic system's knowledge base can also be used in other countries. The constructs of ad effect as specified in the knowledge base are valid in many countries. American research results, for example, can rather directly be applied to the German situation. The purpose of the theoretically based formalization of knowledge as contained in the knowledge base was to accumulate, as completely as possible, the relevant advertising constructs. Within the context of advertising evaluation, constructs of ad effect such as an ad medium's acceptance have to be evaluated for any country. Yet, it is possible that the signification of single constructs differs very much from country to country, respectively that the relations between independent and dependent variables can have different attributes. This is due to the cultural environment as well as the cultural values and norms in single countries. F.ex., one can assume that countries like Italy or the USA where electronic media are very dominant, the use of fast cuts in TV commercials has to be judged different than in countries like Germany or Sweden.

TABLE 4

TRANSFERABILITY OF THE CAAS-DIAGNOSTIC SYSTEM'S KNOWLEDGE BASE

Modifications that are necessary primarily refer to the strength of relations between independent and dependent variables as well as to constructs of advertising effect depending on subjective values (table 4, table 5).

Transferability of the surface design

In this context, the transfer of surface language from a semantic point of view as well as the transfer of pictorial examples which represent cultural norm values have to be examined.

Transferability of the surface language

As problematic as the transfer of technical in practical language is the transfer of the expert system's surface language. The latter refers to the questions the user is being asked, the answer categories of closed questions, the explanations and hints to specific problems and the verbal print-out of the expertise. The Creation of a dictionnary of synonyms is not sufficient as is not a simple retranslation for control purposes (Ger, Belk 1990, S.186). Example: The German adjective "sch├Ěn" ("beautiful", "nice") has two equivalents in the French language, a female ("belle") and a male form ("beau"). French associations to the word "beautiful" are different from German associations to this word. But even the French associations are different depending on whether associations to a female or a male word are inquired (table 6). In order to translate such a word for a question of the expert system, an explicit examination would be necessary which French word comes closest to the respective German one. This problem often arises when doing intercultural research where certain scales are much more appropriate for some countries than for others (Wallendorf, Arnould 1988; Brislin et al 1973; Lonner, Berry 1983). In this respect, one can agree to Ger and Belk who demand that "In cross-cultural research it appears crucial that questions are decided upon by joint efforts of researchers from the cultures involved. Subtle culture factors play in and a rigorous translation procedure is not enough to guarantee cross-cultural standardization." (Ger, Belk 1990, p. 188). Only through differentiated control measures by researchers from different countries can be guaranteed that the connotations connected with a word are interpreted in a uniform way across countries.

Transferability of pictorial examples representing cultural norm values

Pictorial examples for a specific ad factor represent cultural norm values. These norm values were partly obtained from empirical studies, partly from the judgement of experts. Yet, cultural norm values can tremendously differ from country to country: pictorial scenes that are perceived as being irritating in one country, are not perceived in the same way in another country, a landscape that is perceived as being very emotional in one country, can only evoke weak emotions in another. Hence, those norm values have to be checked with respect on congruency across countries. If required, they have to be modified. Also in this case, the need for modification is greatest with emotional norms that largely depend on subjective attitudes. On the contrary, despite of cultural distance, a transfer of norm examples for ad factors that can be judged in objective terms, for example a picture's contrast, does not cause any problems since one can hereby assume cross-culturally equal reactions to the stimuli (table 7).

Cross-cultural evaluation of emotional ad elements

When using emotional elements in advertising, there are two basic possibilities (Kroeber-Riel 1986):

1. the peripheral use of emotional stimuli in advertising; they catch the consumer's attention, lead to a better information processing and create an emotional atmosphere.

2. the central use of emotional stimuli in order to create specific emotional brand experiences. We call this an emotional positioning. This is especially important for saturated markets with functionally exchangeable products.

TABLE 5

TRANSFERABILITY OF SINGLE CONSTRUCTS OF AD EFFECT CONTAINED IN THE KNOWLEDGE BASE

TABLE 6

ASSOCIATIONS FROM FRENCH AND GERMAN STUDENTS TO THE ADJECTIVE "BEAUTIFUL" (GERMAN: "SCHON"; FRENCH: "BEAU","BELLE")

Notwithstanding the peripheral use of emotional stimuli, in the case of the central use of emotional stimuli strategic aspects of advertising evaluation have to be taken into account as well as the question whether the emotional key message can be learnt. Since the strategic part of evaluation causes the least problems when using the CAAS-diagnostic system in other countries, we confine ourselves to the social technical part of evaluation which is relevant for the peripheral as well as for the central use of emotional stimuli.

For the social technical part of evaluation, there basically are two possiblities for measuring emotions. They can be measured either by a differential or by a dimensional approach. The first one aims at different emotions like pleasure, exotism etc. Although there is broad agreement with respect to fundamental emotions (Plutchik 1980; Izard 1977), the emotions derived from those primary emotions and used in advertising can hardly be overseen (Zeitlin, Westwood 1986). For that reason, this measurement is not recommandable for the use in expert systems. Hence, the dimensional approach was chosen for the CAAS-diagnostic system. According to this dimensional approach, emotions can be classified into the following dimensions (Kroeber-Riel 1986):

1) the intensity of emotions

2) the positive or negative direction of emotions (pleasant-unpleasant)

3) the degree of consciousness (conscious-unconscious)

4) the quality of emotions (subjective feeling/experience)

All these attributes of emotions are examined by the CAAS-diagnostic system except the degree of consciousness. The intensity of emotions, e.g. of an emotional picture, is besides the intensity of physical and collative stimuli an influence factor for an ad's activation (Berlyne, 1974; Kroeber-Riel, 1990; Finn, 1988). The direction of emotion is evaluated within the advertising construct "acceptance". The quality of emotion is recorded for the learning of the key message. It is only relevant if the advertising goal of an emotional positioning is pursued. Otherwise, with respect to emotional stimuli it is important to measure their contribution to an ad's acceptance.

TABLE 7

TRANSFERABILITY OF THE CAAS-DIAGNOSTIC SYSTEM'S SURFACE DESIGN

The degree of difficulty when measuring emotional dimensions with the CAAS-diagnostic system is varying. Problems arising in that context are amplified in the case of cross cultural use.

Example: Concerning the intensity of emotions, especially in the case of biologically founded stimuli like a baby or a naked woman, we can assume that reactions to these stimuli are similar on a cross-cultural basis. With respect to direction and quality, however, we have to differentiate: in the case of the baby, direction and quality of the emotions of different target groups are perceived in a similar way, yet, this is not true for the naked woman. In the latter case and in many others, the emotion's direction and quality depend on the subjective values and attitudes of the respective consumers. A picture with a naked woman will be judged entirely different in countries like Turkey as opposed to countries like Germany, for example.

Since direction and quality of emotions are accounted for in the CAAS diagnostic system primarily by the construct of ad effect "acceptance", we consider that uncertainty of evaluation as follows: Can a user make no definite judgements about direction and quality of emotions in an ad, he can answer accordingly. Then, the respective construct of ad effect is not part of the aggregation to constructs of a higher hierarchical level. Instead, the expertise of the verbal print-out obtains a remark about the importance of the respective construct of ad effect and a demand to solve this question with the help of market research tests. For these purposes, he also receives scales of measurement that have been proven appropriate as proposals.

The procedure is completely different concerning the intensity of emotions. Hereby, we can generally distinguish between three categories (Kroeber-Riel 1991, p.153):

- biologically founded (for example baby, naked woman),

- culturally founded (for example mediterranean landscape),

- target group specific emotional schemata (for example a football team).

The judgement of the intensity of an emotional stimulus becomes more difficult as we go from top to bottom of the above categories. What is activating consumers in one cultural environment, does not arouse feelings in consumers in another cultural environment. The same is true for target specific stimuli. A consumer who is not interested in tennis, will not be activated by looking at a tennis match. According to our experiences, the user hereby has only problems when deciding for scale degrees that are close to each other, e.g. he cannot differentiate between weak and medium stimuli or medium and strong ones (Lorson 1992). So far, it has not been problematic, however, to evaluate scale realizations that are distinct enough from each other (for example weak vs. strong emotional stimuli).

To account for this user-uncertainty (uncertainty of decision between several answer possibilities), a further answer category "Cannot make a decision" has been inserted to the respective questions in the CAAS diagnostic system. If the user chooses this answer category, he is then asked between which answer categories he is indecisive (based on the fact that this inability to make a decision mostly happens between neighboring answer categories).

The system then internally continues with one medium value. This value is the arithmetical mean of the answer categories between which the user was indecisive. In the case of emotional stimulus strength, this procedure seems to be acceptable since emotional stimuli determine, next to physical and collateral ones, an ad's potential of activation. The factors of influence are related to each other in a compensatory manner so that a certain inexactitude can be accepted when evaluating emotional stimuli in order to obtain an overall judgement about the ad's potential of activation. We believe that this proceeding is more appropriate than asking the user to assign confidence factors as it happens in the expert systems MYCIN or ADCAD (Burke et al 1992). First, the user is hardly able to exactly quantify his indecisiveness when answering to questions. Second, the factor of confidence can only be analyzed in the case of dichotomous scales. When dealing with interval scales, a confidence factor of .8 for the value "medum" leaves the question unanswered whether the remaining .2 refer to the value "weak" or "strong" (Esch 1990).

CONCLUSIONS

The use of an expert system for advertising evaluation in different countries is problematic but not insoluble. It is possible to transfer the basic model of ad effectiveness and the advertising constructs of the knowledge base as well as the aspects of surface design. If the cultural distance is low, it can be assumed that the adaptation of the CAAS-diagnostic system's surface to cultural particularities are sufficient. Pictorial examples which serve as cultural norm values for evaluation purposes, could be stored in country specific data bases to which the expert system user has access. With increasing cultural distance, modifications of the knowledge base become necessary, especially on the executional level. In this context, greater problems arise for those components of ad effect whose evalution largely depends on the recipients' subjective attitudes and values. In these cases, the expert system is able to differentiate between extremes but it is difficult to measure subtle gradual realizations. For these purposes, the possibilities of operationalization complex advertising constructs are not yet sufficient. Whereas for consumer research there already exist a multitude of scales for measuring an ad's acceptance (Lutz 1985; Wells et al. 1971; Schlinger 1979; Aaker, Bruzzone 1981), in problematic cases it is difficult for the user to measure the acceptance of an ad for specific target groups. This problem therefore also occurs when using the expert system in other countries. In these cases, primary market research is still necessary. The CAAS-diagnostic system takes this into account, for example by integrating results from a program analysor in order to judge a TV spot's acceptance.

REFERENCES

Aaker, D. A., D. E. Bruzzone (1981), "Viewer Perception of Prime-Time Television Advertising", in Journal of Advertising Research, Vol. 21, No. 5, October, 15-23.

Alba, J. W., L. Hasher (1983), "Is Memory Schematic?", in: Psychological Bulletin, Vol. 93, Nr. 2, 203-231.

Batra, R., M. L. Ray (1986), "Affective Responses Mediating Acceptance of Advertising", in Journal of Consumer Research, Vol. 13, September, 234-249.

Berlyne, D. E. (1974), Konflikt, Erregung, Neugier. Zur Psychologie der kognitiven Motivation (Conflict, Arousal, and Curiosity), Stuttgart: Klett Verlag.

Bloch, P. H., M. L. Richins (1983), "A Theoretical Model for the Study of Product Importance Perceptions", in Journal of Marketing, Vol. 47, Summer, 69-81.

Buchanan, B. G., D. Barstow, R. Bechtel, J. Bennett, W. Clancy, C. Kulikowski, T. Mitchell, D. A. Waterman (1983), "Constructing an Expert System", in Building Expert Systems, F. Hayes-Roth, D. A. Waterman, D. B. Lenat (eds.), Reading/Mass.: Addison-Wesley, 127-167.

Brislin, R. W., W. J. Lonner, R. M. Thorndike (1973), Cross-Cultural Research Methods, New York: John Wiley & Sons.

Burke, R. (1991), "Reasoning with empirical marketing knowledge", in: International Journal of Research in Marketing.

Burke, R. R., Rangaswamy, A., J. Wind, J. Eliashberg (1988), "ADCAD: A Knowledge-Based System for Advertising Design", Working Paper No. 88-027, The Wharton School, University of Pennsylvania 1988.

Burke, R., A. Rangaswamy, J. Wind, J. Eliashberg (1992), "ADCAD - ein Expertensystem zur Werbegestaltung (ADCAD: Advertising Communication Approach Designer)", in Expertensysteme - Perspektiven fnr die Werbung (Expert Systems-Perspectives for Advertising), Kroeber-Riel, W., F.-R. Esch (eds.) (1992), in preparation.

Clancy, K. J. (1990), "The Coming Revolution in Advertising", in Journal of Advertising Research, Vol. 30, No. 1, 47-52.

Cook, R. L., J. M. Schleede (1988), "Application of Expert Systems to Advertising", in Journal of Advertising Research, Vol. 28, No. 3, June, 47-56.

Caudle, F. M. (1989), Advertising Art: Cognitive Mechanisms and Research Issues", in Cognitive and Affective Responses to Advertising, Cafferata, Tybout (eds.), Lexington, Mass., Toronto: Lexington Books, 161-218.

Esch, F.-R. (1990), Expertensystem zur Beurteilung von Anzeigenwerbung (Expert System for the Evaluation of Printed Ads), Heidelberg: Physica.

Esch, F.-R., W. Kroeber-Riel (1992), "Expertensysteme in der Werbung (Expert Systems in Advertising)", in Electronic Marketing - Handbuch der Informations- und Kommunikationstechnik im Marketing, Hermanns, A., V. Flegel (eds.), Mnnchen: C.H. Beck Verlag.

Finn, A. (1988), "Print Ad Recognition Readership Scores: An Information Processing Perspective", in Journal of Marketing Research, Vol. 25, May, 168-177.

Fiske, S. T., P. W. Linville (1980), " What Does the Schema Concept Buy Us?", in Personality and Social Psychology Bulletin, Vol. 6, No. 4, 543-557.

Ger, G. and R. W. Belk (1990), "Measuring and Comparing Materialism Cross-Culturally", in Advances in Consumer Research, Vol. 17, M. E. Goldberg, G. Gorn, R. W. Pollay (eds.), Provo, UT: Association for Consumer Research, 186-192.

Harmon, P., D. King (1985), Expert Systems - Artificial Intelligence in Business, New York: John Wiley & Sons.

Horowitz, A. D., J. E. Russo (1989), "Modeling New Car Customer-Salesperson Interaction for a Knowledge-Based System", in Advances in Consumer Research, Vol. 16, T. K. Srull (ed.) (1989), Association for Consumer Research, Provo, UT., 392-398.

Izard, C. E. (1977), Human Emotions, New York: Plenum Press.

Kroeber-Riel, W. (1986), "Nonverbal Measurement of Emotional Advertising Elements", in Advertising and Consumer Psychology, Vol. 3, J. Olson, K. Sentis (eds.), New York, Westport, London: Praeger, 35-52.

Kroeber-Riel, W. (1990), Konsumentenverhalten (Consumer Behavior), 4th edition, Mnnchen: Vahlen.

Kroeber-Riel, W. (1991), Strategie und Technik der Werbung. Verhaltenswissenschaftliche AnsStze (Strategies and Techniques of Advertising), 3rd edition, Stuttgart: Kohlhammer Verlag.

Kroeber-Riel, W., F.-R. Esch (1992), "Expertensysteme im Marketing Expert Systems in Marketing)", in Electronic Marketing - Handbuch der Informations- und Kommunikationstechnik im Marketing, Hermanns, A., V. Flegel (eds.), Mnnchen: C.H. Beck Verlag.

Levermann, T. (1992), Entwicklung eines Expertensystems zur strategischen Durchsetzung von Werbung (Development of an Expert System for the Strategical Evaluation of Advertising), Dissertation an der UniversitSt des Saarlandes,in preparation.

Levitt, T. (1983), "The Globalization of Markets", Harvard Business Review, Vol. 61, May-June, 92-102.

Lonner, W. J., J. W. Berry (1986), Field Methods in Cross-Cultural Research, Beverly Hills: Sage Publications.

Lorson, T. (1992), Expertensystem zur Beurteilung von Fernsehwerbung (Development of an Expert System for the Evaluation of Commercials), Heidelberg: Physica.

Lutz, R. J. (1985), "Affective and Cognitive Antecedents of Attitude Toward the Ad: A Conceptual Framework", in Psychological Processes and Advertising Effects, Alwitt, L. F., A. A. Mitchell (Hg.), Lawrence Erlbaum Associates, Hillsdale/New Jersey, 45-63.

MacKenzie, S. B., R. J. Lutz, G. E. Belch (1986), "The Role of Attitude Toward the Ad as a Mediator of Advertising Effectiveness: A Test of Competing Explanations", in Journal of Marketing Research, Vol. 23, May, 130-143.

Meyers-Levy, J., A. M. Tybout (1989), "Schema Congruity as a Basis for Product Evaluation", Journal of Consumer Research, 16, No. 1, 39-54.

Mitchell, A. A. (1987), "The Use of Alternative Knowledge-Acquisition Procedures in the Development of a Knowledge-Based Media Planning System", in International Journal of Man-Machine Studies, No. 26, 399-411.

Mitchell, A. A. (1988), "The Development of a Knowledge Based Media Planning System", in Data, Expert Knowledge and Decisions. An Interdisciplinary Approach with Emphasis on Marketing Applications, W. Gaul, M. Schader (eds.), Berlin, Heidelberg, New York, London, Paris, Tokyo 1988, 67-79.

Mitchell, A. A., J. C. Olson (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitudes?", in Journal of Marketing Research, Vol. 18, August, 318-332.

Mittal, B. (1989), "Must Consumer Involvement Always Imply More Information Search?", in Advances in Consumer Research, Vol. 16, T. K. Srull (ed.), Association for Consumer Research, Provo, UT., 167-172.

Mittal, B., M.-S. Lee (1988), "Separating Brand-Choice Involvement from Product Involvement Via Consumer Involvement Profiles", in Advances in Consumer Research, Vol. 15, M. J. Houston (ed.), Provo, UT, Association for Consumer Research, 43-49.

Neibecker, B. (1990), Werbewirkungsanalyse mit Expertensystemen (Analyzing ADvertising Effectiveness with Expert Systems), Heidelberg: Physica.

Park, C. W., G. W. McClung (1986), "The Effect of TV Program Involvement on Involvement with Commercials", in Advances in Consumer Research, Vol. 13, R. J. Lutz (ed.), Ann Arbor, Association for Consumer Research, 544-548.

Petty, R. E., J. T. Cacioppo (1983), "Central and Peripheral Routes to Persuasion: Application to Advertising", in Advertising and Consumer Psychology, L. Percy, A. G. Woodside (eds.) (1983), Lexington Press, Lexington/Mass., S. 3-24.

Petty, R. E., J. T. Cacioppo, D. Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement", in Journal of Consumer Research, Vol. 10, September, 135-146.

Plutchik, R. (1980), Emotion. A Psychoevolutionary Synthesis, New York, Hagerston: Harper & Row.

Ray, M. L. (1973), "Marketing Communication and the Hierarchy of Effects", in New Models for Mass Communication Research, P. Clarke (ed.) 2nd edition, Beverly Hills, CA: Sage Publications, 147-176.

Rangaswamy, A., R. R. Burke, J. Wind, J. Eliashberg (1986), Expert Systems for Marketing, Working Paper No. 86-036, The Wharton School, University of Pennsylvania.

Rossiter, J. R., L. Percy (1987), Advertising and Promotion Management, New York: McGraw-Hill.

Rossiter, J., F. Winter (1989), "An expert system for predicting advertisement performance", in Marketing Thought and Practice in the 1990's, G. Avlonitis, N. Papavassiliou, A. Kouremenos (eds.), Conference Proceedings, Vol. II, XVIII Annual Conference of the European Marketing Academy, Athen, Griechenland, 1471-1483.

Schlinger, M. J. (1979), "A Profile of Responses to Commercials", in Journal of Advertising Research, Vol. 19, No. 2, April, 37-46.

Venkatraman, M., A. Villarreal (1984), "Schematic Processing of Information: An exploratory Investigation", in Advances in Consumer Research, Vol. 11, Kinnear, Th. C. (ed.), Ann Arbor: Association for Consumer Research, 355-360.

Wallendorf, M., E. J. Arnoud (1988), "My favorite Things: A Cross-Cultural Inquiry into Object Attachment, Possessiveness and Social Linkage", Journal of Consumer Research, Vol. 14, No. 4, 531-547.

Wells, W. D., C. Leavitt, M. McConville (1971), "A Reaction Profile for TV Commercials", in Journal of Advertising Research, Vol. 11, No. 6, December, 11-17.

Wierenga, B. (1990), The First Generation of Marketing Expert Systems, Working Papier No. 90-009, The Wharton School, University of Pennsylvania.

Winston, P. H. (1987), Artificial Intelligence, Addison-Wesley, Reading/Mass.

Winter, F., J. Rossiter (1992), "ADEXPERT - ein Expertensystem zur Werbegestaltung und zur Werbebeurteilung (Expert Systems in Advertising: ADEXPERT, a new System for Advertisement Construction and Evaluation)", in Expertensysteme - Perspektiven fnr die Werbung (Expert Systems - Perspectives for Advertising), W. Kroeber-Riel, F.-R. Esch (eds.), Mnnchen: Vahlen (in preparation).

Zeitlin, D. M., R. A. Westwood (1986), "Measuring Emotional Responses", in Journal of Advertising Research, Vol. 26, No. 5, 34-44.

----------------------------------------

Authors

Franz-Rudolf Esch, University of the Saarland, Saarbrncken, Germany



Volume

E - European Advances in Consumer Research Volume 1 | 1993



Share Proceeding

Featured papers

See More

Featured

Morality Matters in the Marketplace: The Influence of Morally Based Attitudes on Consumer Purchase Intentions

Andrew Luttrell, Ball State University
Jacob Teeny, Ohio State University, USA
Richard Petty, Ohio State University, USA

Read More

Featured

Conjuring Creativity: The Impact of Fear

Ilgım Dara Benoit, Appalachian State University
Elizabeth Miller, University of Massachusetts, USA

Read More

Featured

J3. You Think I’m Yours but, Trust Me, I’m Not: How Consumers Value Dogs and Cats

Colleen Patricia Kirk, New York Institute of Technology
Samantha Renee Kirk, Boston College, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.