Table 4. Portion of Shared Information between a group and aggregators who are leading the information dissemination. There the each Member in Each Members collection was also another interesting observation about perfectly matched Absolute Group Member Jaccard users. Some users were just highly influenced by their group Fraction Fraction Similarity collection. For instance, the 'group #2 has 567 items in group collections and 28 members. Out of 10 members whose personal 48874216%1635%11.51% Metadata.05 45.73%18.98% 13 13% member A has 159 items and member B has 69 items in personal 51.10%1889% collection respectively and these items were all in the group's collection. However, we couldn't find any evidence showing that Micro 160.8638.06%1431%7.03% they had posted any item to the group. For the users who were active to disseminate information in there group collections, the tend to participate in the group in the early days and contribute to forming the collection. For another kind of users who were highly influenced by group are prone to join the group in the later time when the group collected abundant amount of information. Unfortunately, we failed to find constant pattems of behaviors in the members who had no common items with the groups However, the number of items that zero-overlapped users had (M 104.58)was significantly larger than the members who had perfect overlap with group(M= 30.70, [= 4.18, p< .001).We interpreted this result as the rich users have their own strategies to find useful information and inclined not to rely on somebody else or since they have amassed abundant information, they did not 2561024 look for another information source. In addition. these zero- verlapped users could have very specific information needs, No of shared In formation since they did share little information either with the group or Figure 6. Absolute numbers of Shared Items and Metadata between a group and the members 2 LOOLADAAEANDO ON No, of shared Information a macro tag Figure 8. Member Fractions of Shared Items and Metadata (from Members' Point of view) Figure 7. Absolute Numbers of shared Micro Tags and macro Tags between a group and the members 1400 micro Error! Reference source not found. and figure g show the 1200 distribution of the relative powers, especially member fractions from members' point of view. In these two graphs, we found interesting points. There are two distinctive peaks on the both extreme sides. A subset of members had the personal collection that was barely overlapped with their group's collection and HFTT another subset of members had the collection that was perfectly 0 overlapped with their group collections. We investigated what make this difference by tracing the differences in the posting 9p>23少p>393p>△9 times of the common information. Many members whose collections were perfect match signed up the groups on earlier time and did post items to the group collections. Using the Member Fractions Citeulike interface, users are able to add interesting items not only Figure 9. Member fractions of Shared Micro Tags and macro to their personal collection but also to the group collection Tags (from Members, Point of view)Table 4. Portion of Shared Information between a Group and the Each Member in Each Member’s Collection Absolute Group Fraction Member Fraction Jaccard Similarity Items 48.87 42.16% 16.35% 11.51% Metadata 167.05 45.73% 18.98% 13.13% Macro Tags 42.22 51.10% 18.89% 9.80% Micro Tags 160.86 38.06% 14.31% 7.03% Figure 6. Absolute numbers of Shared Items and Metadata between a group and the members Figure 7. Absolute Numbers of Shared Micro Tags and Macro Tags between a group and the members Error! Reference source not found. and Figure 9 show the distribution of the relative powers, especially member fractions from members’ point of view. In these two graphs, we found interesting points. There are two distinctive peaks on the both extreme sides. A subset of members had the personal collection that was barely overlapped with their group’s collection and another subset of members had the collection that was perfectly overlapped with their group collections. We investigated what make this difference by tracing the differences in the posting times of the common information. Many members whose collections were perfect match signed up the groups on earlier time and did post items to the group collections. Using the Citeulike interface, users are able to add interesting items not only to their personal collection but also to the group collection simultaneously. That is to say, they are active contributors or aggregators who are leading the information dissemination. There was also another interesting observation about perfectly matched users. Some users were just highly influenced by their group collection. For instance, the ‘group #2’ has 567 items in group collections and 28 members. Out of 10 members whose personal collections were 100% matched with the group collection, member A has 159 items and member B has 69 items in personal collection respectively and these items were all in the group’s collection. However, we couldn’t find any evidence showing that they had posted any item to the group. For the users who were active to disseminate information in there group collections, they tend to participate in the group in the early days and contribute to forming the collection. For another kind of users who were highly influenced by group are prone to join the group in the later time when the group collected abundant amount of information. Unfortunately, we failed to find constant patterns of behaviors in the members who had no common items with the groups. However, the number of items that zero-overlapped users had (M = 104.58) was significantly larger than the members who had perfect overlap with group (M = 30.70, t = 4.18, p < .001). We interpreted this result as the rich users have their own strategies to find useful information and inclined not to rely on somebody else or since they have amassed abundant information, they did not look for another information source. In addition, these zero￾overlapped users could have very specific information needs, since they did share little information either with the group or with the members. Figure 8. Member Fractions of Shared Items and Metadata (from Members’ Point of View) Figure 9. Member Fractions of Shared Micro Tags and Macro Tags (from Members’ Point of View) 1 2 4 8 16 32 64 128 256 512 1024 1 4 16 64 256 1024 No. of Pairs No. of Shared Information Items Metadata 1 2 4 8 16 32 64 128 256 512 1024 2048 1 4 16 64 256 No. of Pairs No. of Shared Information macro tags micro tags 0 200 400 600 800 1000 1200 No. of Members Member Fractions item meta 0 200 400 600 800 1000 1200 1400 No. of Members Member Fractions micro macro