Blackboard Expert Systems As Tools For Knowledge-Engineering in Cross-Cultural Advertising Communications

ABSTRACT - Cross-Cultural advertising evaluation with ESWA (Expert System for advertising evaluation) is compared to copy research results from AD*VANTAGE/ACT. Results from two TV-spots for a leading chocolate bar brand in six European countries are available.

Citation:

Thomas Andresen and Bruno Neibecker (1993) ,"Blackboard Expert Systems As Tools For Knowledge-Engineering in Cross-Cultural Advertising Communications", 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: 118-124.

European Advances in Consumer Research Volume 1, 1993      Pages 118-124

BLACKBOARD EXPERT SYSTEMS AS TOOLS FOR KNOWLEDGE-ENGINEERING IN CROSS-CULTURAL ADVERTISING COMMUNICATIONS

Thomas Andresen, GfK Market Research, Nnrnberg, Germany

Bruno Neibecker, University of Karlsruhe, Germany

ABSTRACT -

Cross-Cultural advertising evaluation with ESWA (Expert System for advertising evaluation) is compared to copy research results from AD*VANTAGE/ACT. Results from two TV-spots for a leading chocolate bar brand in six European countries are available.

INTRODUCTION TO CROSS-CULTURAL RESEARCH IN MARKETING

From a formal point of view, cross-cultural research can be definined as research that has culture as its main independent or dependent variable but not as an extraneous and/or residual variable (see Nasif et al. 1991). Taking global marketing theory (see Buzzell 1968; Clark 1990; Jain 1989; Levitt 1983; Meffert 1989; Porter 1989), support can be found for standardization of advertising worldwide based on the reasoning that human nature springs from similar motivations and lifestyles. Empirical findings show, that a number of different advertising appeals like status appeal, soft-sell, modernity/youth etc. can be found in the advertisements of different countries but to varying degrees (e.g. Mueller 1987). At the same time numerous differencies were found e.g. between Japanese and American advertisements and a trend back from standardized advertising could be found recently by Hite and Fraser (1988).

In spite of the lack of "good" theories in cross-cultural research, meanwhile the problems are more and more identified. Nasif et al. (1991) addresses seven methodological problems in cross-cultural-research, Netemeyer et al. (1991) provides evidence for the validity and reliability of cross-national measurement using the CETSCALE from Shimp and Sharma (1987). Actual theoretical improvement stems from social-psychological theories of stereotyping (see Linville 1982; Qualls and Moore 1990) and first improvements to develop a commonly acceptable and empirically-based terminology to describe and categoriese the cultures in different countries (Hofstede 1984; Kale and McIntyre 1991), based on four largely independent dimensions called individualism, power distance, uncertainty avoidance and masculinity.

METHOD AND PROCEDURE

Two TV-spots from a leading chocolate bar brand, Kit Kat are available. One film made for the United Kingdom, the other for continental European markets (Davison 1991) were tested in six countries using AD*VANTAGE/ACT, an International Pretest System for TV-spots.

Background information about the market (Davison 1991, p.59): "Kit Kat was launched in the UK over 50 years ago, and is the leading brand in the very large UK chocolate bar market. The product itself and the slogan "Have a break, have a Kit Kat" have been part of the English culture for many years. The Quackers film was made recently, and continues a long line of successful UK Kit Kat films. (...) The brand has also been present in continental Europe for several years, but has not yet achieved the success which it has in the UK. The Hitchhiker film was made to support the relaunch of Kit Kat across continental Europe. The key objectives of the film were to communicate the name and to dramatise the break situation."

The Hitchhiker spot shows a young man as a hitchhiker on an american highway. The American/international executional style appealed particularly to the young. The product has a strong presence throughout the film but communication of the nature of the product were weak. The Quackers spot shows two hunters. The "English humor" in the film was well accepted (especially in Germany) and the spot was seen as original and different from other commercials.

The films were testet using samples of 120 people in each country. Half of the sample were mothers with children aged 5-15; the remaining respondents were consumers aged 15-35 which are consumers of chocolate bars. Fieldwork took place in summer 1991.

The AD*VANTAGE/ACT procedure is the same in every country. Evaluative measures like attitude shifts are gathered together with the diagnostic ones, e.g. likes and dislikes. The approach is based on natural exposure, many measures, multiple exposures and multiple locations. The test session opens with basic questions, e.g. demographic questions. Afterwards the first TV programme is shown with a first break after 12 minutes. A break of seven spots with three constant "control" commercials and four test commercials is exposed. A second programm follows with questions, twenty minutes recall etc. and a third film with a second break finishes the test procedure.

RESULTS

Both spots achieved high scores of brand awareness (recall) in all six countries. AD*VANTAGE/ACT results are given as index scores based on the national database norms from the different countries. This norm values are different between the countries, reflecting the differencies in the national market structure and the cross-cultural differences (Table 1).

Based on the national norm-scores, which are calibrated to 100, the results for the two Kit Kat-spots are as shown in Figure 1 and 2. Starting with the Hitchhiker-spot, we can see that for brand awareness the results are very positive, ranging from 119 in Germany to 152 in England, all above the norm score of 100. Overall ratings start around the average in Belgium with an extreme of 171 in England. Finally, the persuasion scores, operationalized as attitude shift, are strong below the norm.

Looking on the response pattern across the countries, Belgium (awareness = 143; overall rating = 94, attitude shift = 32), Holland and France show the same ranking - with the brand awareness score as the highest score and with the lowest value for attitude shift. England (152; 171; 57) and Germany (119; 139; 42) as a second group, show the same response pattern with the overall rating as comparative highest score. In Italy the highest value was reached for attitude shift, a result completely differing from the other countries.

For the Quackers-spot again, brand awareness lies above the norm value but below the Hitchhiker-spot (mean brand awareness Quackers = 123; mean brand awareness Hitchhiker = 135), attitude shift is somewhat better but below the norm (Quackers mean = 70; Hitchhikers mean = 61). Again some response patterns can be found: Belgium and Holland vs. England Germany and France vs. Italy.

TABLE 1

AD*VANTAGE/ACT NORMS FROM DIFFERENT COUNTRIES

FIGURE 1

AD*VANTAGE/ACT: HITCHHIKER

FIGURE 2

AD*VANTAGE/ACT: QUACKERS

FIGURE 3

CRITERION RELATED VALIDITY FOR RECALL

There are considerable differences in the overall ratings of the films. Quackers achieved an exceptionally high score in Germany and a good score in England, whereas Hitchhiker was very well accepted in England and good accepted in Germany. A theoretical explanation for these somewhat surprising results can be found in social-psychological theories of stereotyping, where some authors (see Linville 1982; with different results Qualls and More) found empirical support for the polarized appraisal theory, which overcomes the unidirectional evaluative bias for in-group members. Rather the direction of the bias depends on the favorability of the information about a specific group member, consequently positive appealing out-group members will be evaluated more extreme based on the premise that people have less complex cognitive schemas of out-group members.

EXPERT SYSTEM EVALUATION

Expert Systems (XPS) are sophisticated computer programs that manipulate knowledge to solve problems efficiently and effectively in a narrow, specific problem area (Decker and Gaul 1990; Wierenga 1990). ESWA is a theory-based XPS to evaluate advertising and to make diagnostic and evaluative predictions for advertisements. To our knowledge, ESWA is the first fuzzy-based XPS in a qualitativ dominated knowledge domain which has undergone an objective validity test (Neibecker 1992a). Up to now, the findings for the criterion related validity of ESWA-results with copy-test results from AD*VANTAGE PRINT (GfK-Market Research, Germany) are available. Figure 3 shows the correlation between GfK-results for unaided recall and the predicted values from ESWA in T-scores (a standard scale ranging from 0-100, with a mean of 50; Neibecker 1990). Actually 45 ads are available to validate the system. ESWA results are derived from an evaluation session of about one hour for each ad and with blindness to the copy-test results. Including the three outliers, the correlation diminishes to r=0.35, still significant at the 1% level. To analyze the outliers, one recall score for an outlier-ad belongs to a campaign, as a consequence the available AD*VANTAGE PRINT score represents the mean for the campaign ads only. Another stands for the top brand in the market, giving the recall score an additional shift. This should be reflected in a later version of the ESWA knowledge base. Above all, a correlation of r=0.45 (p<0.001) is a convincing proof of the appropriateness of the ESWA knowledge base including the inference algorithm for uncertain knowledge.

Background information to ESWA: ESWA is a rule-based system with a fuzzy inference engine to handle multiple evidence and to allow for intercorrelated subdimension in calculating input certainty. It is completely programmed in COMMON LISP and will run on DOS-systems with 1 MB extended memory.

Fuzzy rules of this kind can easily be transformed into an influence network. A part of this influence network is shown in Figure 4. The symbols are used according to the conventions of causal diagrams and reflect in a formal sense a nomological network (Bagozzi 1980). This influence diagram shows the main influence path of the theory based knowledge base from ESWA including the available AD*VANTAGE PRINT items, which can be used for construct validation later on, if more cases are available (omitted influence paths are indicated by dotted arrows). Including the rules for the verbal expertise, ESWA has about 500 rules. It is worth to note, that such influence networks have significant theoretical and practical advantage over combinatorially explosive decision trees which grow exponentially. Influence networks grow linearly with the size of the problem they represent (Neibecker 1992b).

FIGURE 4

PART OF THE ESWA INFLUENCE NETWORK FOR AD-EVALUATION

FIGURE 5

INFERENCE TREE FOR ESWA RECALL

We now used ESWA to evaluate the Kit Kat spots. ESWA has four options to present results: a complete Table of all facts, a verbal expertise, an interface to HARVARD GRAPHICS and an inference tree with an integrated reporting of the results. Figure 5 shows the predicted T-score for the recall of the Hitchhiker-spot (62), which means, that a recall strong above average (50) is predicted. In addition the first two hierarchy-levels of the inference tree are shown. Numbers behind the facts represent a linear attenuation function (see Neibecker 1990) according to the PROSPECTOR-algorithm (e.g. for picture recall y = .23 + .53 x) which causes a weighting of the rule (and the rules facts). Predicted T score for attitude toward the ad (comparable to the GfK item "overall rating") is 77 and for persuasion (GfK item "attitude shift") 39.

The predicted scores from ESWA are comparable with the market reseach data. Recall ("brand awareness") and attitude toward the ad (Aad) are above or equal the average or norm, persuasion (attitude shift) is below the average (see Figure 7).

Somewhat different are the results for the Quackers spot. ESWA predictions are 44 for recall, 78 for attitude toward the ad and 49 for persuasion. For both films attitude toward the ad is predicted strong above the average, persuasion is predicted below average, all this scores are in accordance with the market research results. But to improve the ESWA expertise for cross-cultural advertising, the available and future theories like the stereotyping theory should be implemented in a later knowledge base.

Only the recall score with 44, a value somewhat below the T-scale average of 50, deviates from the AD*VANTAGE/ACT-results which is for all six countries above the AD*VANTAGE/ACT norm value of 100, but looking at the mean for all countries, brand awareness for Hitchhiker is above the mean for Quackers, a ranking which ESWA predicts correct and market research results also should be interpreted in light of the situational exposure bias.

In Figure 6 a gain and loss analysis for Quackers recall, as predicted by ESWA, is given. The first hierarchy-level of the knowledge base for recall inference by the XPS is shown, in other words: these influence components for recall can be analysed to explain the low recall score. Only the score for picture recall is above average, indicated by the zero-line which match with a T-score of 50. Especially the score for brand recall remains with only 47 after attenuation. Brand recall is considered as the most important single influence component for the overall recall score. Thus a recall of only 44, six points below the average is given by ESWA.

BLACKBOARD MODEL OF PROBLEM SOLVING

Looking on the really heterogenious results even among the European countries, considered in the near future as part of the "European Common Market" the question arises, how the different national, culture dependent aspects for advertising evaluation can be handled within XPSs. Given the lack of theories in multicultural research, these XPSs must also integrate a lot of the heuristics, the experts are working with.

A flexible alternative to conventional XPSs is discussed as blackboard systems. With blackboard systems it is now possible to develop an architecture to coordinate different independent XPSs. Thus cross-cultural specific compenents could be integrated into country or cultural specific XPSs. A strategic or a coordinating XPS can summarize and evaluate the results from the country specific systems.

FIGURE 6

ESWA GAIN AND LOSS ANALYSIS (INFLUENCE COMPONENTS FOR RECALL)

FIGURE 7

ESWA RESULTS (T-SCORES)

Such a blackboard model of problem solving is a very simple yet powerful idea for coping with problems characterized by the need to deal with qualitative and uncertain knowledge and data with the need to apply non-deterministic solution strategy. Blackboard systems can be seen as a natural evolution of XPSs (Craig 1989; Engelmore and Morgan 1989; Kroeber-Riel, Lorson and Neibecker 1992). Knowledge is segmented into modules, which are implemented in separate and independent XPSs with, if needed, seperate inference engines. All this independent XPSs communicate via a common memory, the blackboard. They must be able to put new information and incrementally solutions to the blackboard and to read and understand the informations which are found on the blackboard (the global database). Communication and interaction between the XPSs take place solely through the blackboard. In this pure model no control component is specified. If one watches the task being performed, the solution is built incrementally and opportunistically - i.e. in an autonomous reactive mode based on the systems own intelligence. XPS with these capabilities are well suited for a coarse-grained parallelism on a conceptual level.

We have integrated a blackboard option into ESWA and are willing to realize this concept step by step.

CONCLUSIONS

Despite global marketing theories, the standardisation potential for chocolate bar advertising seems to be limited. It should be stated, that food market are not considered as the primary global markets. The consequence should be a carefull consideration of the homogeneity of the countries within a specific market which could be the basis for a combination strategy between standardized and localized advertising.

In addition we could show, that expert systems can be very helpful in advertising evaluation when based on a theoretical relevant knowledge base. Further research is needed to develop cross-cultural theories in Marketing and to implement this knowledge into multiple, culture specific expert systems. The blackboard architecture is a fruitful framework to implement this concept.

REFERENCES

Bagozzi, R. P. (1980), Causal Models in Marketing, New York et al.:Wiley.

Buzzell, Robert D. (1968), "Can you standardize multinational marketing," Harvard Business Review, (November-December), 102-113.

Clark, Terry (1990), " International Marketing and National Character: A Review and Proposal for an Integrative Theory," Journal of Marketing (October), 66-79.

Craig, Iain D. (1989), The Cassandra Architecture: Distributed Control in a Blackboard System, Chichester: Ellis Horwood.

Davison, Andrew J. (1991), "Kit Kat - a Case Study on a TV-Advertising Test in Six European Countries," in The Contribution of Advertising Testing to the Development of Effective International Advertising, Nnrnberg, GfK-Marktforschung, 57-65.

Decker, R. and W. Gaul (1990), "Einige Bemerkungen nber Expertensysteme fnr Marketing und Marktforschung," Marketing ZFP, 12, 257-271.

Engelmore, Robert and Tony Morgan, ed. (1989), Blackboard Systems, Wokingham et al.: Addison-Wesley.

Hite, Robert E. and Fraser, Cynthia (1988), "International Advertising Strategies of multinational corporations," Journal of Advertising Research (August/September), 9-17.

Hofstede, G. (1984), Culture's Consequences: International Differences in Work-Related Values, Beverly Hills: Sage.

Jain, Subhash C. (1989), " Standardization of International Marketing Strategy: Some Research Hypotheses," Journal of Marketing, Vol. 53 (Januar), 70-79.

Kale, Sudhir H. and Roger P. McIntyre (1991), "Distribution Channel Relationships in Diverse Cultures," International Marketing Review, 8, No. 3, 31-45

Kroeber-Riel, Werner, Tania Lorson and Bruno Neibecker (1992), "Expertensysteme in der Werbung," DBW (Die Betriebswirtschaft), 52, 91-108.

Levitt, Theodore (1983), "The Globalization of Markets," Harvard Business Review, 92-102.

Linville, Patricia W. (1982), "The Complexity-Extremity Effect and Age-Based Stereotyping," Journal of Personality and Social Psychology, 42, 193-211.

Meffert, Heribert (1988), Strategische Unternehmensfnhrung und Marketing, Wiesbaden: Gabler.

Mueller, Barbara (1987), "Reflections of Culture: An Analysis of Japanese and American Advertising Appeals," Journal of Advertising Research, (June/July), 51-59.

Nasif, Ercan G., Hamad Al-Daeaj, Bahman Ebrahimi and Mary S. Thibodeaux (1991), "Methodological Problems in Cross-Cultural Research: An Updated Review," mir Management International Review , 31, 79-91.

Neibecker, Bruno (1990), Werbewirkungsanalyse mit Expertensystemen, Heidelberg: Physica.

Neibecker, Bruno (1992a forthcoming), "A Concept for Construct Validation of Expert Systems in Marketing", in Operations Research Proceedings, Berlin et al.: Springer.

Neibecker, Bruno (1992b forthcoming), "Expertensysteme zur Werbemittelgestaltung," in Handbuch Kommunikations-Management, ed. Ralph Berndt and Arnold Hermanns, Wiesbaden: Gabler.

Netemeyer, Richard G., Srinivas Durvasula and Donald R. Lichtenstein (1991), "A Cross-National Assessment of the Reliability and Validity of the CETSCALE," Journal of Marketing Research, 28, 320-327.

Porter, Michael E., ed., (1989), Globaler Wettbewerb, Wiesbaden: Gabler.

Qualls, William J. and David J. Moore (1990), " Stereotyping Effects on Consumers' Evaluation of Advertising: Impact of Racial Differences Between Actors and Viewers," Psychology & Marketing, 7, 135-151.

Shimp, Terence A. and Subhash Sharma (1987), "Consumer Ethnocentrism: Construction and Validation of the CETSCALE," Journal of Marketing Research, 24, 280-289.

Wierenga, Berend (1990), "The First Generation of Marketing Expert Systems," Working Paper No. 90-009, Marketing Department, Wharton School, University of Pennsylvania.

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