본 연구는 사전지식과 관여도가 신념편향의 발생에 미치는 영향을 공분산기반모형을 사용하여 살펴보았다. 신념편향이란 어떤 주장의 결론을 강하게 믿는 사람이 그렇지 않은 사람보다 그 주장의 타당성 내지 논리성을 높이 평가하는 경향을 말한다. 연구 1에서는 사전지식 수준이 신념편향의 발생에 어떤 영향을 미치는지 조사하였다. 연구결과, 사전지식 수준은 신념편향의 발생에 영향을 미치며, 구체적으로 사전지식 수준이 높은 경우에는 신념편향의 발생이 억제되고 사전지식 수준이 낮은 경우에는 신념편향의 발생이 촉진되는 것으로 나타났다. 이것은 사전지식 수준이 높은 경우에 휴리스틱 시스템이 작동하고 그로 인해 신념편향이 발생할 것이라는 기존 견해와는 상반되는 결과이며, 사전지식 수준이 낮은 경우에 휴리스틱 시스템이 작동하여 신념편향이 발생하고 사전지식 수준이 높은 경우에는 분석적 시스템이 작동하여 신념편향이 발생하지 않을 것이라는 본 연구의 견해를 지지해 주는 것이라고 할 수 있다. 연구 2에서는 연구 1의 결과가 다른 제품 카테고리에서도 나타나는가와 관여도가 신념편향의 발생에 어떠한 영향을 미치는지를 알아보고자 하였다. 연구 결과, 다른 제품 카테고리에서도 연구 1과 동일한 결과가 나타 났고, 신념편향의 발생은 관여도에 따라 달라지지 않는 것으로 확인되었다. 이로써 사전지식 수준이 신념편향의 발생에 미치는 영향은 제품 카테고리에 따라 달라지지 않는다는 것이 입증되었고, 연구 1의 결과를 일반적인 소비재 카테고리에까지 확장할 수 있는 가능성이 발견되었다. 본 연구는 다음과 같은 시사점을 갖는다. 이론적인 측면에서 연구는 이원적 프로세스 이론의 작동과 신념편향의 발생에 미치는 사전지식의 영향을 새롭게 조명함으로써, 이원적 프로세스 이론과 신념편향 이론의 정교화에 기여할 것으로 보인다. 또한 신념편향의 측정에 공분산기반모형을 적용함으로써, 신념편향의 측정방법과 범위는 물론 공분산기반모형의 외연을 확장하는 효과도 기대할 수 있다. 마케팅적인 측면에서 본 연구는 사전지식 수준이 낮은 브랜드에 대한 마케팅 메시지의 결론을 소비자들이 믿을 수 있도록 구성하는 것이 왜 중요하고 어떤 효과를 기대할 수 있는지 분석함으로써, 설득적 마케팅 커뮤니케이션의 전략적 근거를 마련하는 데 기여할 것으로 보인다. 마지막으로 소비자 정책 측면에서 본 연구는 소비자들이 신념편향으로 인해 잘못된 결정을 내리지 않도록 소비자 교육을 강화하고 소비자 보호를 위한 법적ㆍ제도적 장치의 마련과 같은 정책적 뒷받침이 필요하다는 점을 부각시킬 수 있을 것으로 보인다.

Using the covariation-based model, this research addresses consumers’ belief bias, which is the tendency to assess the validity of message arguments on the basis of the believability of argument conclusion. In investigating the effect of beliefs on consumer information processing, this research examines whether belief bias is affected by prior knowledge or involvement through two experiments. The covariation-based model posits that beliefs in causal power are represented in terms of the degree of covariation between the cause and its effect. A person could be said to be biased by beliefs, when he or she with a high subjective probability about the conclusion of an argument(P(B)) shows a relatively higher evaluation on the validity or logicality of that argument(ΔB(=P(B∣A)-P(B∣￢A))) than a person with a low subjective probability. Therefore, the occurrence of belief bias can be detected first by observing if the ΔB value of the high P(B) group is larger than that of the low P(B) group, and then by analyzing whether the difference in the ΔB value between the two groups is statistically significant. Experiment 1 is performed to examine whether the occurrence of belief bias is affected by consumers’ prior knowledge. The results showed that for both ‘sneakers’ and ‘toothpaste’ (the two product categories used as experiment stimuli), when prior knowledge levels about the brand were low, belief bias occurred, whereas when prior knowledge levels about the brand were high, belief bias did not occur. Let us look at details in the case of sneakers. With ‘Nike,’ a high prior knowledge brand, the high P(B) group had a smaller ΔB value than the low P(B) group, and the difference was not significant(Mlow_P(B)=.1833 vs. Mhigh_P(B)=-.0281; t(36)=1.627, p=.112). On the other hand, with ‘Ethonic’ (a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant(Mlow_P(B)=.0214 vs. Mhigh_P(B)=.1292; t(36)=-1.790, p<.1). Thus, we can infer that the belief bias occurs not for brands with high prior knowledge, but only for brands with low prior knowledge. Similar results were obtained in the case of toothpaste. With ‘Perioe,’ a high prior knowledge brand, the high P(B) group had a smaller ΔB value than the low P(B) group, and the difference was not significant(Mlow_P(B)=.0556 vs. Mhigh_P(B)=.0100; t(36)=.575, p=.569). On the other hand, with ‘Tripledenta’ (a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant (Mlow_P(B)=.0077 vs. Mhigh_P(B)=.1280; t(36)=-1.724, p<.1). Thus, it can be inferred that the belief bias phenomenon occurs only for brands with low prior knowledge, not for brands with high prior knowledge. These results support Hypothesis 1 predicting that belief bias is affected by the level of consumers’ prior knowledge. Specifically, when the level of prior knowledge is high, the occurrence of belief bias is suppressed. In contrast, when the level of prior knowledge is low, the occurrence of belief bias is promoted. The results also confirm the viewpoint of our study. When prior knowledge levels are low, a heuristic system operates and belief bias occurs. On the other hand, when prior knowledge levels are high, an analytic system operates and belief bias does not occur. This viewpoint should be contrasted with the existing viewpoint that when prior knowledge levels are high, a heuristic system operates and as a result belief bias occurs. Experiment 2 was conducted to replicate this result in other product categories and to examine whether involvement moderates this relationship. ‘College preparation books’ and ‘instant noodles’ were used as experiment stimuli, and they showed lower involvement levels than the stimuli in Experiment 1. When prior knowledge levels about the brand were low, belief bias occurred. But when prior knowledge levels about the brand were high, belief bias did not occur. Let us take a look at the case of college preparation books. With ‘Concepts and Principles: Mathematics,’ a high prior knowledge brand, the high P(B) group had a smaller ΔB value than the low P(B) group, and the difference was not significant(Mlow_P(B)=.5400 vs. Mhigh_P(B)=.3765; t(37)=1.342, p=.188). On the other hand, with ‘Giltoraebi Mathematics’(a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant(Mlow_P(B)=.0333 vs. Mhigh_P(B)=.2238; t(37)=-2.264, p<.05). We can infer that in the case of college preparation books the belief bias phenomenon occurs not for brands with a high prior knowledge level, but only for brands with a low prior knowledge level. The case of instant noodles is also examined. With ‘Shin Ramen,’ a high prior knowledge brand, the high P(B) group had a larger ΔB value than the low P(B) group, but the difference was not significant(Mlow_P(B)=.0962 vs. Mhigh_P(B)=.1462; t(37)=-.726, p=.473). On the other hand, with ‘Mapsi Noodle’ (a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant(Mlow_P(B)=.0083 vs. Mhigh_P(B)=.2267; t(37)=-2.334, p<.05). Thus it can be inferred that in the case of instant noodles the belief bias phenomenon also did not occur for brands with a high prior knowledge level, but it occurs only for brands with a low prior knowledge level. The results of Experiment 2 show that even under low involvement, belief bias occurs only when prior knowledge levels are low, but not when prior knowledge levels are high. Hypothesis 2 predicting that the influence of prior knowledge levels on belief bias will differ according to involvement levels was not supported. Thus, involvement has no moderating effect on the influence of prior knowledge levels on the occurrence of belief bias. Nevertheless, the results of Experiment 1 are replicated in different product categories and the findings can be extended to general consumer products. Theoretical and managerial implications of these findings are discussed as follows. From a theoretical perspective, this study sheds new light on the effect of prior knowledge on the operation of the dual process and belief bias occurrence. This study then makes it possible to sophisticate the dual process theory and theories on belief bias. This study has also extended methods and ranges of belief bias measurement. Furthermore, the study makes it possible to extend the covariation-based model. From a managerial perspective, this study suggests that marketers should make a conclusion of marketing message believable for brands with a low prior knowledge level. In such a case, consumers who have positive responses to and strong beliefs in the conclusion of a message are likely to give a positive evaluation and believe the basis of the conclusion as well as the conclusion itself. From a consumer policy perspective, this study implies that it is necessary to prevent consumers from making wrong choices due to belief bias. It could be achieved by bolstering consumer education and political support such as legislation and institutional arrangements.

Using the covariation-based model, this research addresses consumers’ belief bias, which is the tendency to assess the validity of message arguments on the basis of the believability of argument conclusion. In investigating the effect of beliefs on consumer information processing, this research examines whether belief bias is affected by prior knowledge or involvement through two experiments. The covariation-based model posits that beliefs in causal power are represented in terms of the degree of covariation between the cause and its effect. A person could be said to be biased by beliefs, when he or she with a high subjective probability about the conclusion of an argument(P(B)) shows a relatively higher evaluation on the validity or logicality of that argument(ΔB(=P(B∣A)-P(B∣￢A))) than a person with a low subjective probability. Therefore, the occurrence of belief bias can be detected first by observing if the ΔB value of the high P(B) group is larger than that of the low P(B) group, and then by analyzing whether the difference in the ΔB value between the two groups is statistically significant. Experiment 1 is performed to examine whether the occurrence of belief bias is affected by consumers’ prior knowledge. The results showed that for both ‘sneakers’ and ‘toothpaste’ (the two product categories used as experiment stimuli), when prior knowledge levels about the brand were low, belief bias occurred, whereas when prior knowledge levels about the brand were high, belief bias did not occur. Let us look at details in the case of sneakers. With ‘Nike,’ a high prior knowledge brand, the high P(B) group had a smaller ΔB value than the low P(B) group, and the difference was not significant(Mlow_P(B)=.1833 vs. Mhigh_P(B)=-.0281; t(36)=1.627, p=.112). On the other hand, with ‘Ethonic’ (a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant(Mlow_P(B)=.0214 vs. Mhigh_P(B)=.1292; t(36)=-1.790, p<.1). Thus, we can infer that the belief bias occurs not for brands with high prior knowledge, but only for brands with low prior knowledge. Similar results were obtained in the case of toothpaste. With ‘Perioe,’ a high prior knowledge brand, the high P(B) group had a smaller ΔB value than the low P(B) group, and the difference was not significant(Mlow_P(B)=.0556 vs. Mhigh_P(B)=.0100; t(36)=.575, p=.569). On the other hand, with ‘Tripledenta’ (a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant (Mlow_P(B)=.0077 vs. Mhigh_P(B)=.1280; t(36)=-1.724, p<.1). Thus, it can be inferred that the belief bias phenomenon occurs only for brands with low prior knowledge, not for brands with high prior knowledge. These results support Hypothesis 1 predicting that belief bias is affected by the level of consumers’ prior knowledge. Specifically, when the level of prior knowledge is high, the occurrence of belief bias is suppressed. In contrast, when the level of prior knowledge is low, the occurrence of belief bias is promoted. The results also confirm the viewpoint of our study. When prior knowledge levels are low, a heuristic system operates and belief bias occurs. On the other hand, when prior knowledge levels are high, an analytic system operates and belief bias does not occur. This viewpoint should be contrasted with the existing viewpoint that when prior knowledge levels are high, a heuristic system operates and as a result belief bias occurs. Experiment 2 was conducted to replicate this result in other product categories and to examine whether involvement moderates this relationship. ‘College preparation books’ and ‘instant noodles’ were used as experiment stimuli, and they showed lower involvement levels than the stimuli in Experiment 1. When prior knowledge levels about the brand were low, belief bias occurred. But when prior knowledge levels about the brand were high, belief bias did not occur. Let us take a look at the case of college preparation books. With ‘Concepts and Principles: Mathematics,’ a high prior knowledge brand, the high P(B) group had a smaller ΔB value than the low P(B) group, and the difference was not significant(Mlow_P(B)=.5400 vs. Mhigh_P(B)=.3765; t(37)=1.342, p=.188). On the other hand, with ‘Giltoraebi Mathematics’(a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant(Mlow_P(B)=.0333 vs. Mhigh_P(B)=.2238; t(37)=-2.264, p<.05). We can infer that in the case of college preparation books the belief bias phenomenon occurs not for brands with a high prior knowledge level, but only for brands with a low prior knowledge level. The case of instant noodles is also examined. With ‘Shin Ramen,’ a high prior knowledge brand, the high P(B) group had a larger ΔB value than the low P(B) group, but the difference was not significant(Mlow_P(B)=.0962 vs. Mhigh_P(B)=.1462; t(37)=-.726, p=.473). On the other hand, with ‘Mapsi Noodle’ (a low prior knowledge brand), the high P(B) group had a larger ΔB value than the low P(B) group, and the difference was significant(Mlow_P(B)=.0083 vs. Mhigh_P(B)=.2267; t(37)=-2.334, p<.05). Thus it can be inferred that in the case of instant noodles the belief bias phenomenon also did not occur for brands with a high prior knowledge level, but it occurs only for brands with a low prior knowledge level. The results of Experiment 2 show that even under low involvement, belief bias occurs only when prior knowledge levels are low, but not when prior knowledge levels are high. Hypothesis 2 predicting that the influence of prior knowledge levels on belief bias will differ according to involvement levels was not supported. Thus, involvement has no moderating effect on the influence of prior knowledge levels on the occurrence of belief bias. Nevertheless, the results of Experiment 1 are replicated in different product categories and the findings can be extended to general consumer products. Theoretical and managerial implications of these findings are discussed as follows. From a theoretical perspective, this study sheds new light on the effect of prior knowledge on the operation of the dual process and belief bias occurrence. This study then makes it possible to sophisticate the dual process theory and theories on belief bias. This study has also extended methods and ranges of belief bias measurement. Furthermore, the study makes it possible to extend the covariation-based model. From a managerial perspective, this study suggests that marketers should make a conclusion of marketing message believable for brands with a low prior knowledge level. In such a case, consumers who have positive responses to and strong beliefs in the conclusion of a message are likely to give a positive evaluation and believe the basis of the conclusion as well as the conclusion itself. From a consumer policy perspective, this study implies that it is necessary to prevent consumers from making wrong choices due to belief bias. It could be achieved by bolstering consumer education and political support such as legislation and institutional arrangements.

Belief Bias, Covariation-Based Model, Subjective Probability Model, Prior Knowledge, Involvement