Summer Milestone 10 Reputation Systems research and exploration Rating Friends Without Making Enemies surabhi and kajal
Introduction to the paper
Ratings among platform users is a simple task but involves complex subjective dimensions which cannot be overlooked. The paper discusses some of the factors that influence the rating system of an Online Social Networking Site(OSNS): 'Couchsurfing.org'
Couchsurfing is an OSNS established in 2003 (before facebook). It lets people build their profiles, connect with people and view their connection's profiles. A user's profiles on Couchsurfing shows the user's friends and how they are rated, and the vouches that others make but not the trust ratings. It is a network of travelers and hosts where hosts are the people who are ready to make a room in their own homes for the travelers to stay. Such a process needs high levels of trust.
Points discussed in the paper
- Challenges faced by the rating system
- Relation and or pattern observed between ‘friendship’ and ‘trust’.
- Reciprocal ratings
- Public versus private rating system and its effect on the reciprocal ratings.
- Quantification of relationships on a predefined scale
- The effect of language and culture on ratings
- Effect of ratings on others reputation
- Timing of ratings relative to the evolution of a relationship
Challenges faced by the rating system
- It was observed that close friendship involves trust factor but high levels of trust need not need good friendship: For example, having 1000 friends on facebook does not mean that the person is trustworthy.
- Concern over the reaction of friends on rating: Only 41.7% of the survey respondents noticed a difference in selecting friendship levels between them and their friends. Off these 201 users, 79 responses expressed neutral sentiments, 26 expressed curiosity as to why the difference, but not enough to ask, some cases also felt guilty or ashamed for assigning a lower rating or level.
- Concern over the effects of rating on a person's profile and reputation: Users generally refrained from giving a negative reference since it would affect the person's profile. A bad remark was avoided unless the experience was extremely bad. Hence, very low negative feedbacks.
- Very little or no incentive for a negative feedback.
- Influence of similarity of user profiles and preferences on trust ratings: It was observed that people with similar profiles and/or people who joined the platform almost at the same time had high trust rates among them.
Results of the paper
Understanding trust vs friendship
When one user tends to assign a high friendship level, the trust levels also seem to be high. However, high trust levels can occur with a range of friendship levels. Following are some of the responses from the survey respondents:
-I think close friends you trust, but I don’t think everyone you trust is a close friend. [P12]
-Friendship includes trust. You can trust someone, but still without that person being a friend. I guess friendship is a more elusive concept and therefore more difficult to judge. [S312]
These responses support the findings.
Following are the responses which do not support the findings:
-In general I only *really* trust my closest, real life friends. That takes years to earn. [S256]
-Trust takes, you know, in a lot of cases years to build, whereas friendship and that sort of thing can sort of happen instantly [P8]
The conclusion of such mixed responses was that on average both friendship and trust increases with time, but even more with familiarity.
Quantifying relationships on a predefined scale
On an average, respondents evaluated trust and friendship as being equally difficult to rate, several of them considered that trust can be measured in 'levels', while friendship cannot because of its multidimensional nature.
The effect of language and culture on ratings
Fixed labels may not translate well into all languages, for example, consider the following response:
Best friend, I understand what that means but between good friend and close friend, I don’t know exactly what is the higher level [P10]
Reciprocity was observed more in the case of publicly displayed friendship ratings than privately displayed trust ratings. Also, mutual friendship ratings were more aligned than mutual trust ratings. In this case, there can be two possible explanations:
1. Ratings reflect a true difference in mutuality between trust and friendship
2. People rate friendship more carefully since it is made public
Reciprocity in friendship
General expectation is that friendship is mutual but one must choose the level before they are shown how the friend rated them. Following are the responses that attributed the difficulty in selecting friendship level to this matching problem:
-It can be difficult to select a friendship level if I am unsure of how the other person may react or if I think they may see our friendship as being at a different level.[S114]
-Cause sometimes you don’t want to be unpolite by just adding the person as ”Couchsurfing friend” ... you want to have the person the same friendship level. [S175]
-The trust level is anonymous, and I tend to trust people more easily. that’s why friendship level is more difficult: everyone can see it [S276]
41.7% of our survey respondents ever noticed selecting a friendship level different from what their friend had designated.When asked the resopnses were “OK”, “neutral”,‘ ‘didn’t care”,“it’s normal”, “it doesn’t matter”. 26% had felt “weird”,“strange”,“confused” or “awkward”.
Reciprocity in references
Reciprocity starts with the CS norm, mentioned though not held by all, that both the host and surfer will leave a reference. In paper we saw that people where very specific about writing references. They didn't write one for the person they hosted and even if they wrote about it they would meter first how other person wrote about them.