Social role, capital and trust

Choice of Online Communities

In this session's study, we are going to compare the social capital and trust mechanisms of two online communities (OCs). Based on this goal, I decided to choose two OCs where social capital and trust play an important role in their daily services. Inspired by the buying camera example presented in the A Survey of Trust Use and Modeling in Real Online Systems paper, I chose Amazon.com and mitbbs.com/exchange. Most us are familiar with Amazon.com. It is a US-based multinational electronic commerce company. It started as an online bookstore, but soon diversified, selling DVDs, CDs, MP3 downloads, computer software, video games, electronics, apparel, furniture, food, and toys. [wiki]. Mitbbs.com, is a Chinese bulletin board system site. The majority of its users are overseas Chinese. The focus of this particular forum, mitbbs.com/exchange, is to provide a platform where its users buy, sell, and exchange a broad variety of goods. Since money is involved, it is obvious that the social capital and trust are an important factor that affect people's activities on these two OCs.

Social Capital and Trust Mechanisms

Amazon.com

Amazon.com applies different trust models to different categories of sellers. For any of these trust models to work, an overall assumption is made. It is assumed that the users trust the host, Amazon.com. Under this assumption, Amazon.com authorizes sellers. For companies, such as Macy's and Nike, Amazon hosts stores under their company or brand names. This is similar to a public certification authorized by Amazon.com. Users would trust the identities of these stores. For example, we would trust the Macy's store on Amazon.com is really Macy's rather than some other organization that claimed this name ahead of Macy's. We trust Amazon.com to not letting any other organization to own the name of Macy's other than Macy's itself. In this sense, again, it works like a public certificate signed by a trusted third party, where the third party here is Amazon.com. In this sense, Amazon.com embrace the spirit of social capital, as mentioned in On and Off the Net: Scales for Social Capital in an Online Era, what is Social Capital? For some researchers it means the social groups and networks that create positive outcomes, while for other researchers it means the outcomes themselves. We can think of both in this case.

On this other hand, however, this trust mechanism would not work for individual sellers for obvious reasons. For instance, A person, P, has goods to sell and P opens a store under the name “IamP”. P would easily pass the Amazon.com authentication step, assuming no body has claimed this name already and this name does not corresponds to any well-known merchandise brand. Basically, Amazon.com tells its users that P is the real “IamP”. Well, this tells very little regarding whether or not P should be trusted. Meanwhile Amazon.com did it job to sign the certificate. Apparently, to facilitate transactions and thrive the OC, other trust mechanisms are needed. Amazon.com uses user rating system for individual sellers.

The sreenshot above demonstrates the rating for a book seller named “oldbookcase”.

The screenshot above demonstrates some of the comments from buyers of this book seller. As we can see this book seller has a large number of transactions and a overall high score. Amazon.com supplies this data as a trust credit for “oldbookcase”. Again, since the assumption is that buyers trust Amazon.com, this information is going to be taken into account when the buyers purchase goods.

The screenshot above demonstrates how the trust credit of sellers are used in the search results. They are placed a clear locations with emphasized fonts.

Mitbbs.com/exchange

The user activities in mitbbs.com/exchange differs from Amazon.com in many ways. First, people not only buy and sell, they also exchange goods. Judging from the name of this forum, the exchange action was probably the initial intention. Of course, trading with money counts as exchanging as well. Second, from my observation, very few users are there to make a profit. In many cases people are there to get ride of goods they don't want. Most of the goods are new but not always. With the increasing trend of online shopping, people, especially girls, often find out they've purchased goods that are not actually suitable for their needs. An obvious example is clothing. Without trying them in person, it's hard to tell whether or not they fit. So, after receiving their goods from other online stores, they decide to sell of exchange for something else from mitbbs.com/exchange. Third, generally, things getting exchanged are pretty cheap. This is determined by the first two characteristics. Since the sellers are not interested in making a profit, they tend to offer very reasonable prices in order to get ride of the goods quickly. Since the buyers are aware of fact that the goods they are going to receive has been marked as “unwanted” by its first-hand buyer already, they won't be willing to spend a good deal of money in purchasing them.

The nature of the bulletin board system and the aforementioned characteristics of mitbbs.com/exchange forum determined its social capital and trust mechanisms. The host, mitbbs.com, is not assumed to be a trusted third party. Anybody can participate in the discussion and hence can participate in the exchange activities. Mitbbs.com is not involved in the trust model. This is a fundamental difference from Amazon.com. For instance, assume a person, Q, purchased something from a seller at Amazon.com, and the product did not match with the its description. This transaction, hence, involves cheating or even commercial fraud. Q can not only rate negatively of this seller, Q can also report this seller to Amazon.com, which has the responsibility to maintain it business environment. Q could demand his money back and Amazon.com will assist Q to do so. Mitbbs.com, however, is not involved in any user activities, and it is not responsible for any fraud transactions. It does not provide a uniform rating mechanism as Amazon.com.

On the other hand, as a thriving forum that has been serving its purpose for a long time, it is inconceivable that no trust mechanisms are imposed. So, I investigated not only the tools provided by the host, I also observed closely of the contents of people's participation, where the trust mechanism lays implicitly. Of course the accuracy of information is still not guaranteed since the host does not work as trusted certification source as we just discussed. This is similar to the problem mentioned in Establishing trust management in an open source collaborative information repository paper, verifying the accuracy of the emergency information used by the proposed model is essential to establish trust management in this particular context.

In the first mechanism, people post threads to report particular sellers. In this case, only extremal good and bad transactions are mentioned since the buyers have to go out of his way to create these threads. Other users refer to these threads and adjust the implicit trust credits of particular sellers, which are other regular users.

The screenshot above demonstrated a thread created by the user “ppjy” reporting the user “suxin”. They are both regular users of this OC. I briefly translated and summarized this thread as below.

ppjy: I recently purchased 8 items costing $120 from “suxin”, who used a $5 bubble envelop for shipping. The envelop was broken when it arrived, and some goods were missing. What should I do about this?

The second mechanism is built on top of the first one. Once in a while, some OC participants would summarize reported sellers and post a overall list of sellers to avoid. These threads are usually promoted to the first page of the forum and stay there for a reasonably long time.

The screenshot above demonstrated a thread created by the user “mononoke” summarizing a list of sellers that were reported negatively. I briefly translated and summarized this thread as below.

mononoke: derong, Seni, sw103, demonbull, lisacute, dennisfeng, qingxi, xiaotaosu, jojojo, chuer2009 were reported with fraut transactions. Please avoid!!!

In the third mechanism, the forum provides a searching tool. When it comes time to purchase goods, the buyer could conveniently search for this seller and evaluate his implicit trust credit from his previous activities.

The screenshot above demonstrated the search result for a user with id “babies”. From this search result, people can see that in the last two months, this user has exchanged baby closing, baby shoes, small furniture, makeup, hair accessories, Chinese medicine, SD card, and HP ink. From this information, other users can tell “babies” is an experienced exchanger. This information helps other to adjust the implicit trust credit for “babies”.

Suggestions of Trust Mechanism Improvement

Amazon.com

I believe Amazon.com is continuously putting effort into researches and empirical studies of how to improve their social capital and trust systems. Meanwhile they indeed have accomplished a good deal in this aspect. The current social capital and trust model together with other facilities they provide for their sellers and buyers have made them one of the main electronic commerce companies in the world. There are, however, still improvements to make. If I were to give Amazon.com suggestions in improving their trust mechanism, I would recommend rewarding users for rating. Everybody likes free stuff, even if they're little.

1. Giving out a small amount of Amazon.com credit for a certain number of ratings.
2. Giving out small prizes. For instance, since Amazon.com has large storage and a variety of goods, it should be easy to reward people by giving small prizes from their “hard-to-sell” products. These small prizes can be shipped together with the next purchase from Amazon.com to save the shipping cost and encourage purchase.
3. Giving out as accumulative special credits. Users can choose to accumulate this special credit and get something bigger later. For example, Amazon.com can have a monthly menu of small prizes. When the user see what he wants, he can get it with this special credits.

Mitbbs.com/exchange

Since this public bulletin board system is not profiting from the users' exchange transactions, it would be difficult to persuade its founders to raise money for re-designing and re-implementing the system into something fundamentally different from what they have right now. Of course, if that's not a constraint, then many rating techniques can be added. With that constraint though, we have to think about improving strategies with small modifications of the system.

I think the strategy proposed in the paper Survey of trust use and modeling in current real systems is suitable here. It basically embrace the Sprite of Google's PageRank. A meaningful situation is the following: every single user predicts the trust scores he/she should place in other users from his/her personal point of view and on his/her behalf. In this way, every user is in charge of aggregating all the trust statements he/she deems relevant (and in this way, he/she can, for example, limit himself/herself to just fetch information expressed by friends of friends) and run the local Trust Metric on this data. In this situation, everyone maintains a dynamically calculated score that reflects his trust credit, and make it visible to others. This would require small addition to the bulletin board system rather than implementing a rating system.

Proposal for the Final Project

In the last paragraph of the previous section of this post, I didn't really elaborate on that idea. I think it could be an interesting topic to explore more. I recently read the original paper published in 1998 presenting Google and its PageRake system, The Anatomy of a Large-Scale Hypertextual Web Search Engine. It was a smart move by Google back then, and it was a smart proposal by the Survey of trust use and modeling in current real systems paper to relate PageRake with OC's trust model. The PageRake precisely represents the trust credit of a web page. It sounded like a promising social network analysis study case that I can investigate.