a new version of the couchsurfing script is up. i'm very happy with this one, and hope to refine the tag cloud.

it seems that what the tag cloud shows is the multi dimensional coordinates of each person's reality bubble as seen by others, and the reality bubble they see in others. haven't worked out a way to get the tag cloud out of the profile page, but it's in the works. here are some screenshots and their description, using my own profile as an example.

reference behavior



i found that there is a clear pattern on how people behave towards references. some people leave more than they get, some get more than they give, and some are more "spread out" meaning they have a lot of non mutual references. though it might be hard to understand what's going on, and i'm still working on it, i see it as a basic reciprocity factor. someone with a negative reciprocity factor probably feels "owed to", meaning, they consider that some people are not worth the reference, whereas some others feel they "owe", considering that everyone deserves a reference, be it out of gratitude or whatever. as i said, i have no good explanation for this, all i know is that this happens.

my test case: 15.98% of the people i hosted considered i did not deserve a reference back, as i leave a reference for pretty much everyone i meet on the system. i'm still working out why this happens. on some cases people just forget/have no time/whatever, on some other cases, people are expecting more from CS than just a place to stay. i've had some clear examples of this, where people got the impression i didn't like them, and despite the fact that they got the free place to stay, free food, free water, free internet, etc, they still did not consider they should leave a reference because, i postulate, they were looking for more than that and did not get it. this just means that their basic requirements aren't just the facilities, they require some "hanging out" too, some catering, some pampering, some friendliness. some hosts provide this. i do not. i could argue that it's on my profile, but only 25% of the people read it anyway, so it's pointless.


surf/host balance



this one is easy. if you host more, you're a hoster, if you travel or surf more, you're a traveler. people come in all shapes on this one, and the "nothing" isn't working yet. some people host a lot (like myself), some others surf a lot, and everything in between. how that will help you decide, i have no idea.


tag clouds



this is probably the most useful one. i have a set of keywords/phrases and draw a tag cloud out of them. an optimal tag list is still in the works. basically, what each of these tag clouds estimates is what words are most often used in the references, both inbound and outbound. so inbound are what people say about this person (in my case, "interesting" comes up top), and outbound are what the person says about others (in my case, "interesting" again). some linguistics work is necessary and also the removal of some keywords that are too common in english (like "nice" or "good"). anyway, each word is a dimension of a person's personality, and the total amount of times it shows up is the amplitude of that particular dimension on that human being's reality. in a way, we are peer-reviewing someone's character by statistically analyzing what people say about them. i've found this very accurate, and despite the keywords not being fully optimized, it clearly shows, for instance in my case, that there is an intellectual focus, both inbound and outbound. some other profiles have tags such as funny or fun as top tags, showing a different kind of focus. this one is probably the most useful one to get a general picture of what the person is like.


warmness analysis



though incomplete, this one is a special one i'm working on. i postulate that despite almost everyone being selective about who they meet, their friendships through the system will tend to be a bell curve. hiding in this is the fact that there is no way of predicting, even if you read the profile, how you'll get along with someone. this is in line with my previous studies, and will prove, if i get lucky on these numbers, that no matter how well you draw up a profile, people that you meet will inevitably follow a uniform distribution from love to hate and everything in between, i.e., we are all fucking clueless! some normalization of the curves might be necessary, as the average relationship differs from person to person, and i'd argue, from culture to culture, as the meaning of "friend" is different. i argue that "warmer" cultures will rate their friendships "higher" than "colder" cultures. quotes because all these categories are unfair generalizations. but this might be a good place to demonstrate the "warmness" that everyone talks about when they travel.
also in this, i will add awkwardness metrics. the awkwardness is measured by calculating the relationship differential. get this: i note you down as good friend, you note me down as couchsurfing friend. fucking awkward. in this case, it would be a positive awkwardness, where i'm rating people higher than they rate me. i think it indicates extroversion. mirroring it, you will also get people with negative awkwardness, rating people consistently lower than they rate them. i'd say these are the introverted people. also, the total amplitude of awkwardness: how bad are we at being reciprocal in terms of friendship ratings. i'd say, on this one, that this varies tremendously between people. i'd rate pretty high on this one as i do not rate anyone below friend.


i'd recommend posting some metric requests on the script page or comment and request metrics. as i dive into these things, i see some really interesting patterns showing up. it might be that this is a good start for an actual numerical analysis on how people connect with each other. or it might be just another exercise to demonstrate how hard it is to demonstrate anything regarding people.


by the way, license is MIT now due to the inclusion of JQuery.

godspeed starstuff ☆★☆☆★★★☆☆☆☆☆