Explaining online bahavioral differences: An Internet dependency perspective

INTRODUCTION

Understanding users has always been a major challenge in designing software. With the advances of Internet-based electronic commerce, many practitioners discovered that the concept of "user focus" is not solely limited to traditional software design. Early web design philosophy assumes "If you build it, they will come." However, this concept is no longer a plausible assumption (25).

A web site should give the users a reason to visit but many designers overloaded the web visitors with too much information. Bettman et al. (2) cautioned web site designers that delivering superfluous information might impede consumers from making good decisions. Pereira (21) further indicated that certain features or information are only relevant to a certain group of users.

Even though the Internet has become a popular marketing channel, knowledge gained about user behavioral differences is still scarce. Thus, research (6, 19) still stresses the importance of Internet behavioral studies. Along with Pereira's suggestions on the possibility of Internet group behaviors, we decided to focus on the level of Internet-dependency and its effects on Internet usage.

Internet dependent users and non-dependent users have shown behavioral differences in using the Internet (9). While many Internet users are not addicted to the Internet by psychological standards, they do show different degrees of Internet dependency (23). Therefore, a study into the levels of Internet dependency can potentially help delineate behavioral differences of Internet users.

Although dramatic online differences were discovered between dependent and nondependent users, very limited empirical research has been conducted to examine the behavioral differences in various Internet features. The primary goal of this study is to examine the behavioral differences among Internet dependent and non-dependent users and their responses to several Internet features, such as game playing online chatting, online shopping, searching for information, and email.

BACKGROUND

The concept of Internet dependency (or Internet addiction), coined by Goldberg (11), was intended to describe the effects of excessive Internet use on personal lives. However, an agreement on the precise definition of this terminology among clinicians, researchers, and social policy makers has not yet been reached (24). Thus, several terms describing similar concepts have been used in the literature. We adopted a generic term - Internet dependency to describe the adverse effects resulting from excessive Internet use.

Internet dependency has gained significant attention in recent years. The concept of Internet dependency no longer stays in academic definitions. Instead, many real life cases were reported (5, 8, 17). Research on Internet dependency, however, has been predominantly psychologically oriented where the goals are identifying symptoms and finding cures. To date, there exists very little research on how dependent users (and nondependent ones) react to several internet features, such as chat rooms, search engines, etc.

RESEARCH HYPOTHESES

Current research on technological dependency focuses on the time factor in two different ways: the history of Internet usage and the duration of the time spent online. The former includes the time span of a user's involvement in using the Internet. The familiarity in using the Internet is gained through the usage process and grows with time spent online. Early research (26, 27) indicates that Internet or computer dependent users tend to be technologically savvy. Recent research (31), however, shows that Internet dependent users do not necessarily like programming and exploration into software and hardware. They could be newcomers to the Internet, too.

Another aspect of time regarding Internet dependency centers around the duration of time spent online for logon sessions. Internet dependents normally spend more time online than non-dependent users. However, most Internet dependency research does not treat longer logon sessions as the sole factor that distinguishes dependent users from non-dependent ones. Researchers, nonetheless, have discovered that excessive long time sessions are associated with Internet dependency. Thus, we hypothesize the effect of time on two aspects:

Hypothesis 1: Internet dependent users are not any different in terms of history of Internet use.

Hypothesis 2: Internet dependents spend more time online than non-dependent users.

Entertaining features of the Internet seem to attract most users and indulge them into the cyber world. Current literature suggests that Internet dependent users tend to prefer entertaining features more than non-dependent users. These users are pleasure seeking users (16) and possibly game players (9, 11). This discussion leads to the following hypothesis:

Hypothesis 3: Internet dependent users play online games more than non-dependent users.

Smyth (28) shows that addictions increase when technology becomes more lifelike. While playing games is not necessarily a feature of "live" function on the Internet, chatting with other users on the other end of the computer delivers real human interactions. Internet dependent users are likely to fulfill the need of communication through chat rooms (20). While online chatting is instantaneous where a message is sent or received without very much delay, sending or receiving email messages does not have the same level of interaction.

Even though emailing is less interactive compared with online chatting, it is becoming a necessity of Internet users no matter if they are addicted to the Internet or not. As Kingsley and Anderson (13) indicated, email is the prime example of an interactive communication technology. Its popularity leads us to suspect whether dependent users and non-dependent ones show discernable differences in using email. Therefore, we separate the communication needs into chatting and emailing, which implies the following hypotheses:

Hypothesis 4: Users who engage in extensive chat activities on the Internet are more likely to be Internet dependent users.

Hypothesis 5: Users who engage in extensive emailing on the Internet are more likely to be Internet dependent users.

Very little research has focused on the difference of online shopping behaviors between dependent and non-dependent users. While this holds true, Pike (22) shows that online shopping is actually a possibility of Internet addicts. However, he did not indicate whether dependent users are distinguishable from non-dependent ones on this factor. We suspect that Internet dependency will contribute a difference in users' preference to use online shopping.

Hypothesis 6: Internet dependent users do not shop online more than non-dependent users.

Research has shown contradicting results in interpreting Internet dependent users' preference in searching information online. Orzack (20) indicated that Internet dependent users could be information seekers. Young (32), however, argues that these users do not regard the Internet as an informational tool. Instead, they treat the Internet as a form of escape that allows them to get away from their problems. Thus, we hypothesize:

Hypothesis 7: Internet dependent users do not prefer online information searching more than non-dependent users.

Research on Internet dependency does not seem to agree on whether gender difference does exist in dependent users. Some research (3, 9, 12, 26, 27) indicates that males are more likely to be addicted to the Internet. Other research (33) shows that females are prone to Internet dependency as well. In addition, Kingsley and Anderson (13) indicate that men would experience the loss of not having Internet access more deeply than women. Thus, we tentatively advance the following hypothesis.

Hypothesis 8: Gender difference does not exist in Internet dependency.

METHODOLOGY

Sample and Procedures

Since there is no central directory of Internet users, it is very difficult to select users from the population at random. A survey that is self-selection in nature was used for this research. The questionnaire was placed on a WWW site. Both Javascript and Web server scripts were used to reduce incomplete answers. To encourage participation, respondents were promised an executive summary of the results. The announcement of the survey was posted on news groups, web pages, search engines and posters. Data collected was stored in a backend database server.

Table 1 summarizes the demographic profile of respondents. 401 users responded to the survey; 245 male respondents (63%) and 144 female respondents (37%). The ages ranged from 14 to 71 with an average of 29. Most respondents fall in the 20-29 year old group. These figures match GVU's 8th survey results. The 20-29 age group is also the largest group in GVU's survey.

The items used to measure Internet dependency were adapted from Young (32) and Egger (9). This results in 20 items in five point Likert-type scale. Game playing, online chatting, online shopping, information searching and emailing were measured based on frequency of use.

Analysis and Results

Reliability of instrument was measured by Cronbach alpha. Typically, reliability coefficients of 0.70 or higher are considered adequate (18). Cronbach alpha of 0.88 was obtained for these 20 items indicating a high internal consistency or reliability.

A K means cluster analysis was then conducted to classify Internet users into dependent and non-dependent groups. Table 2 shows that there is a mean difference (p<0.01) between dependent and non-dependent users on classification items in the cluster analysis. The first cluster consists of 232 respondents with a high level of Internet dependency. The second cluster, consisting of 159 respondents, shows that these users are relatively low in Internet dependency.

A discriminant analysis was used to examine the classification power of a set of discriminating variables: history of Internet usage, duration of Internet usage, gender, online game playing, online chatting, online shopping, information searching, and emailing. Table 3 shows the standardized discriminant coefficients and discriminant loadings for these variables.

Lambert and Durand (15) suggest that discriminant loadings larger than 0.30 are acceptable for their significance. Based on this, the discriminant loadings of five variables including frequency of Internet use, online chatting, game playing, searching for information, and gender, were above the cut-off point. These variables also have moderate to high discriminant coefficients. Both criteria suggest that these variables are important discriminators. The results of t-tests, testing the mean differences on the five discriminating variables, are shown in Table 4.

The first hypothesis suggests that Internet dependent users were new to the Internet compared with non-dependent users. The result of discriminant analysis shows that the history of Internet usage does not have a high discriminating power. T-test statistics also suggest no statistical difference among these two user groups. However, the duration of online usage was a strong discriminant whose discriminant coefficient is 0.459. Dependent users were found to spend much more time online than non-- dependent users (p<0.001).

Hypothesis 3 purports that dependent users are more likely to be online game players. Although relatively weak in its discriminating power (0.194 in discriminant coefficient), game playing is nonetheless an influential factor that makes dependent users different.

Online chatting is fairly strong in its discriminating power, while emailing does not seem to be a determinant at all. T-test statistics also show a significant mean difference in online chatting between dependent and non-dependent users (p<0.001 ).

Discriminant analysis shows that online shopping was not strong enough to be a determining factor. Therefore, Hypothesis 6 is supported in that dependent users and non-dependent users do not show a significant difference in shopping online.

Hypothesis 7 conjectures that dependent users do not care for searching information online. Information searching was found, in this study, a discriminating factor that contributes to the group separation. Dependent users were more comfortable searching online, as there is mean difference in terms of online searching for information (p<0.001).

Gender proves to have a high power in discriminating dependent users from non-dependent ones. Male users were the major gender in dependent users (p<0.001). Therefore, Hypothesis 8 is rejected in favor of an alternative hypothesis.

DISCUSSION

The rejection of Hypothesis 1 is consistent with previous research in that Internet dependents are not different in their history of using the Internet. The history of Internet usage is about the same between the two groups under this study. This can perhaps be explained by the fact that Internet dependency is the concept of the after-use effects. These effects can occur to any users at different degrees (23).

Findings in the current study show that Internet dependent users tend to spend more time online than non-dependent users. Dependent users were online on an average of 15.5 hours per week, while non-dependent users spent an average of 8.5 hours per week online. This finding is consistent with previous research, which indicates that Internet dependent users spend more time online (33). A logical explanation can be that dependent users' ability to control online usage is minimal (20). These users also feel the need to spend more time online in order to gain satisfaction.

Even though games have the tendency of indulging players of many kinds, Internet dependent users seem more likely to play games than others do. An explanation can perhaps be derived from Eto et al.'s (30) work which suggested that perceived enjoyment is positively related to the frequency of Internet usage and daily Internet usage. Several previous studies (16, 28, 35) have also reported dependent users' game playing tendency. With Internet dependents! high frequency and duration in using the Internet, their perceived enjoyment in playing games could further shape their intention in using the Internet.

The support for Hypothesis 4 suggests that Internet dependents were involved in chat rooms more than others were. Young (32) also indicates that online games and chat rooms are most addictive. Her finding is supported in this study, as online chatting was the single most influential factor in the discriminant analysis. The sense of community is an effective feature to promote online offerings (10, 14) and improve user's acceptance of a new technology. Even so, we found that online Internet dependent users were more likely to use online chat rooms. Emailing did not play a significant role in Internet dependency. This might be because sending and receiving email is a fundamental part of the Internet learning experience. As the Internet gains popularity, using email software has become an essential tool for Internet user.

The support for Hypothesis 6 has a formal indication of the role online shopping plays in Internet dependency. As it turned out, online shopping was not a significant factor to discriminate user groups. The t-test statistics showed that the means of both groups with respect to online shopping were quite close (2.60 vs. 2.40, p. 0.1). while online shopping did not show significant discriminating power, online information searching was significant in this study. Korgaonkar and Wolin (14) indicated that the heavier users using the web for personal reasons sought higher enjoyment from the web for seeking information. This supports our finding on the role of Internet dependency in information searching.

The rejection of Hypothesis 8 seems reasonable since women might experience higher anxiety than men in using computers in general (10). Table I also suggests that the effect of male dominance is more likely to appear in the dependent user group than in others.

CONCLUSION

In this paper, we show the impact of Internet dependency on the usage patterns of various Internet features. Internet dependent users were not necessarily new to the Internet. They were online longer and more frequently than non-dependent users. However, these dependent users were not any more experienced in using the Internet. They tended to focus on a small set of Internet features and spend most of their time with these features. These features, when identified, would provide practitioners insights in determining what to deploy on the web to suit customer needs. As we discussed before, the limited viewable space in any given web browser leads to a design strategy that presents relevant information up front. Findings presented in this paper show web designers what to deploy up front so that visitors will experience less "information overload."

Dependent users felt more at ease in online information searching. They might have a higher chance to experience various web designs. Several researchers have shown the importance of information interchange function built into web sites (4, 34). The pre-purchase information is considered the primary value that web sites can provide to users (1, 4).

Within a homepage, the dependent users would feel more comfortable in using the search feature. As usability studies suggest that most Internet users will not scroll when visiting the home (main) page, putting the search function on the home page would benefit dependent users.

The chat function provides a sense of community that Internet dependent users would use more often than nondependent users (32). This, again, shows a practical value to practitioners. The chat function can have many variations including discussion groups, private chat rooms, online interactive seminars, etc. An up-front deployment of this feature is highly welcomed by these dependent users.

Even though we found no significant differences between two user groups in using email, web designers should interpret this finding with caution. In this study, both groups have shown higher frequency in using email. Thus, email support in the form of customer support can be presented to general users without concerns about their degree of Internet dependency.

Another interesting finding of this study shows that male dominance effect only existed in dependent users. Balanced gender distribution was found in non-dependent users. This study further suggests that the controversy of whether gender difference really exists on the Internet may better be explained by the effect of Internet dependency.

This study, while consistent with existing research in part, is still limited in scope. Future research can survey respondents from a broader area and adopt more objective measures where it is practical.