User participation patterns over time in Yahoo! Answers

Authors


Introduction

On many accounts, Yahoo! Answers is a success case of an online community in general and of a social Q&A service in particular (Hitwise, 2008). In a social Q&A system, people post a variety of questions on an open space seeking to benefit from ‘the wisdom of the crowd’. On the other hand, people voluntarily share information and offer advices to other people who are unknown to them. Research on online communities has been trying to explain dynamics of such voluntary participation. In our view, understanding the participation patterns of the community as a whole constitutes a prerequisite step towards understanding the social dynamics involved therein.

In a previous study, we compared two different Q&A sites, Google Answers and Yahoo! Answers, and argued that the active user participation in Yahoo! Answers accounts for its success to a large extent (Shah, Oh, & Oh 2008). We characterized the participation structure of each site with the statistics pertaining to two major activities that take place in a Q&A site, questioning and answering, and observed that Yahoo! Answers has developed a responsive community with the majority of the users actively engaged in providing answers as well as asking questions. On average, each question posted on Yahoo! Answers received 7.86 answers from peer users, demonstrating the level responsiveness of the communities. In addition, the detailed analysis of 55,000 user profiles revealed that those who earned higher “levels” within the Yahoo! Answer's point system are even more focused on answering questions than asking.

In the current study, we will continue our analysis of user participation in Yahoo! Answers in terms of questioning and answering, while adding the dimension of time to the analysis. In particular, we are interested in seeing whether the participation patterns of users change over time and when people develop certain participatory behaviors in Yahoo! Answers. In theories of social exchange, it is suggested that positive outcomes in the early stage of interactions result in positive affect (Kollock 1994; Lawler 2001), which, in the context of a community, leads to commitment to the community and social behaviors strengthening the social capital of the community (Cummings, Sproull & Kiesler 2002). In other words, for individual users the early stage of their involvement in a community is critical in determining further participation. If that is the case in Yahoo! Answers, we may suppose people's participation patterns, in aggregate, change over time in the direction that their contributions increase. The question we are addressing in this study, therefore, is whether and how aggregate patterns of user participation are different depending on the duration of their membership in Yahoo! Answers.

Methods

In order to get information about individual users such as their membership duration, we looked at the user profiles. Each registered user of Yahoo! Answers has their own profile page. Although Yahoo! Answers provides APIs allowing developers/researchers to collect data from their database, there is no API support to access user profiles. Therefore we adopted a two-step approach:

First, we drew a sample of user IDs from the large database11 of questioning/answering data that we had collected from Yahoo! Answers using its APIs. This step was necessary to construct URLs of corresponding user profiles.

Second, we crawled user profile pages and parsed them to get information we need for the present study. Yahoo! Answers' user profile provides a number of statistics regarding the membership and activities of each user. Among others, we were interested in looking at the following about a user 222:

  • The date the he/she joined (registered as a member)

  • The number of questions he/she has asked

  • The number of answers he/she has provided

  • The current level of the user

In order to incorporate the time dimension into analysis, we calculated the duration of membership for each user. The duration is determined by the difference, in days, between the date the user joined Yahoo! Answers and the date the profile was collected. This study reports the result of the analysis of 100,000 user profiles.

Result

The level of participation of 100,000 users, in terms of the number of questions and the number of answers, is largely in line with what we found in the previous study. On average, users asked 19.74 questions while providing 232.7 answers during 309.5 days of their membership with Yahoo! Answers333. Figure 1 shows the changes in the number of questions and the number of answers over time. The numbers were obtained by averaging over all users whose duration of membership (in days) is the same. Note that it is a semi-log graph for the number of answers. The distribution of users by their membership duration (days) is shown in Figure 2.

Figure 1.

Questions and answers over time

Figure 2.

Distribution of days

As shown in Figure 1, both questioning and answering lines show a clear trend of increase in activities in early days (up until about 100 days), with relatively large fluctuations. The increase in the number of answers is drastically higher. In fact, considering that the answer graph is in log scale, the increase in the number of answers in these days is almost exponential to the increase in the number of questions. After the early stage of vigorous changes, users who are in the range of approximately 100 to 300 days of membership, which is the vast majority of the sample, show stable patterns of steady increase in their activities.

Although both questioning and answering are vital participatory activities, providing answers makes more of contribution a to the formation of a strong community. In order to see the change in the proportion of these activities in detail, we examined the number of answers divided by the number of questions, over time. Figure 3 shows the result.

Figure 3.

Changes in the answer / question ratio

In Figure 3, we observe a noticeable increase in the proportion of answering activities in relatively ‘young’ users. After a certain period of time, the ratio of the number of questions to the number of answers more or less converges to 1:10, and remains stable until around 500 to 600 days. This finding suggests that Yahoo! Answers successfully engages new members while sustaining existing members to a large degree.

Conclusion

With a large sample of user profiles, we have investigated participation behaviors of users of Yahoo! Answers, with a particular emphasis on the changes in participation patterns over time. The results show that users of Yahoo! answers contribute substantially to the community, particularly with their provision of answers to other people's questions. More importantly, their contributions consistently increase over time for the majority of users, and the proportion of answering activities, on average, grows up to 10 times in the early stages of their involvement in Yahoo! Answers. Such strong participatory behaviors of Yahoo! Answers' users defy the classic economic theory that assumes people are not inclined to contribute to public goods (Bowles & Gintis 2002). We suppose the theoretical account of the commitment to a community, suggesting the critical impact of the outcome of the early stage, has potential explaining this phenomenon. Much remains to study, though, before we can fully describe and explain the participatory behavior of users in a social Q&A site. This study analyzed user participations only in aggregation. In our future study, we plan to trace behaviors of specific individual users over time. A study of aggregated patterns show the overall participation at a macro level, such as the present study, and a micro level analysis of individual behaviors will complement each other to bring a better understanding of user participations.

Footnotes

  1. 1

    At the time of writing, our database had more than 1.5 million questions and 12 million answers.

  2. 2

    We did not obtain any identification information regarding each user. As for the Yahoo! Answers' user IDs, we used encoded version of those IDs to access their profiles. Once we obtained profiles, user IDs were not used in any kind of analysis that can lead to their true identity.

  3. 3

    The membership duration in our dataset ranges from 1 day to 1156 days. Yahoo! Answers launched its service in December 2005, and one user in our dataset has been a member since December 8, 2005.