Erratum: Motivation for using search engines: A two factor model (Journal of the American Society for Information Science and Technology (2008) 59:11 (1829-1840))
Journal
Journal of the American Society for Information Science and Technology
Journal Volume
61
Journal Issue
1
Pages
214-216
Date Issued
2010
Author(s)
Abstract
We have recalculated the results according to the proper formula and present the correct results in the revised version of Table 6. In addition to the confidence intervals, we calculated the t-values and p-values to compare the differences between the “attract” and “retain” responses for each factor. As in hypothesis testing, where the traditional method for measuring the confidence interval and the p-value method are two approaches that can be used alternately, the Bonferroni intervals and the p-value method used in our work should produce exactly the same results. However, Bonferroni's inequality method, which evolved from the probability inequality bearing the same name, is used to approximate the lower bound of the coverage probability of simultaneous confidence statements when the random variables involved in the statements are correlated (Johnson & Wichern, 2001). Therefore, it is relatively conservative in terms of Type I errors, albeit at the expense of Type II errors. On the other hand, the t-test treats each factor as independent; hence, it is a more lenient significance test. As we know, it is always difficult to achieve a balance between Type I and Type II errors. Since we do not know the correlation between the components, i.e., the 23 factors, a priori, we provide the results of both analyses here. The results of the analyses are basically the same, suggesting that the probability of a strong correlation between the 23 factors is low. After applying the correct calculation formula, most of the results remain the same, except for the “readability of search results” and “search tip” factors, which are not significant after correction. However, for these two factors there is a difference between the results of Bonferroni's confidence interval and those of the t-tests. The “readability of search results” factor is not significant in Bonferroni's procedure, but it remains significant in the t-tests at the 0.05 significance level. All the hygiene factors generate significant differences in the “attract” and “retain” responses. Most important, 12 of the factors induce more “attract” than “retain” responses, whereas only two of the factors in the table (#12: multimedia search, and #14: search instruction) induce more “retain” than “attract” responses. Hence, the first hypothesis in our paper is strongly supported. The results of the hygiene factors were unchanged after correction of the calculation error. With regard to the motivation factors, three factors (#16, #18, and #19 in the table) induce more “attract” than “retain” responses, but only one of them, “information highlighting,” is significant. On the other hand, five of the remaining six motivation factors that induce more “retain” than “attract” responses are significant. The results still suggest that motivation factors are more likely to “retain” than to “attract” search engine users, even though the tendency is not as strong as the opposite effect seen in the hygiene factors. Since our hypotheses deal with the overall tendencies of hygiene and motivation factors in “attracting” and “retaining” users, we also assess their effects by applying Hotelling's T2 test. The results show that, at the 0.01 significance level, hygiene factors induce significantly more “attract” responses than “retain” responses (T2=380.607). On the other hand, motivation factors induce more “retain” responses than “attract” responses (T2=44.958). In summary, the results of the individual factors and the overall trends support the proposed hypotheses and the conclusion is the same: Hygiene factors are more likely to attract than retain search engine users, whereas motivation factors are more likely to retain than attract such users. We apologize for the mistake made in our paper, and for any inconvenience caused. Moreover, we would like to emphasize that the slight change in the results after the correction process does not affect the validity of the two-factor model in explaining users' motivation to use search engines. Empirically, the results still support the hypotheses developed from the two-factor model. More important, the hypotheses were developed inductively through logically sound argumentations based on the two-factor model. The validity of this theory is based not only on empirical results, but also on logical reasoning. As mentioned in the paper, it is not our intention to emphasize the importance of one factor over the other in search engines. Rather, we want to use the proposed systematic model to explore the possibly different impacts of these two kinds of factors on users' motivations. Traditionally, hygiene factors are strongly emphasized and fully developed in search engines, but motivation factors are typically ignored, and thus not well developed. This explains why far fewer motivation factors are adopted by search engines. It may also be the reason that the effects of motivation factors are not as strong as those of hygiene factors in our findings. Given that we highlighted the importance of motivation factors in retaining users, we expect that further development of those factors will induce stronger effects than we have found in our research. More important, they will provide stronger competitive advantages for search engines. Google is an example of a search engine that has done an excellent job in applying both hygiene and motivation factors. For instance, its document ranking method is based on the relevance of the information to users, instead of its commercial value to the information provider. This is exactly the point that we want to stress in Wu, Chuang, and Chen (2008): user satisfaction should be the central concern when developing a search engine. Since people use search engines to find information, which involves querying (retrieving relevant information, supported by hygiene factors) and browsing (digesting the retrieved information, supported by motivation factors), a well-designed search engine should provide functions that support both types of factors.
Type
other