I wanted to measure trust between users, mostly neighbours; if there are measures based on social network graphs metrics then could you provide me with some please, thanks in advance.
I am not aware of these metrics. I will just tell you what occurs to me as I think about this. On a general social network like Facebook, I guess trust might be related to the cardinality of your network of friends. if you have thousands of friends to me that means that your level of trust with them should be a lot smaller than if you have just a dozen. There are some things that you simply won't share or say because you don't even know who will see them. So maybe you can take a function over the node degree as a proxy of the level of trust between the user on that node and all users of the nodes directly connected to it.
On a social network directed towards some specific goal, like Research Gate, things might be diferent. Trust might be more content dependent. Trust in Research Gate might mean something like not being afraid of sharing your research problems and ideas because you know that nobody will steal them. So you open yourself towards collaboration. From this point of view, the bigger, more open, and more widely recognized (and available in search engines) you network is, maybe the less likely will be that somebody violates on your trust, because everything is left out on the open for everybody to see. So in this case, the number of nodes, as a proxy for the network importance and degree of stablishment, might give an idea of its trustworthiness.
Combining both ideas, maybe you can think of a metric of the sort Number of nodes of the network / average user node degree as a first idea towards a way to measure trust in a social network. But you would have to leverage/normalize the expression. The idea is that network trust comes from it being bigger, but with very small parts or clusters (the groups of each user closest links)
One thing that comes to mind to me to validate this sort of model is that you could try to develop a sort of a short dictionary of trust words or sentences. It seems a dificult but maybe challenging task. For example I believe the word "dear" could be one of those, because it is a kind word, and trust and kindness are often connected. With this dictionary you could search the social network content against those words or sentences. Maybe you would find a predominance of small degree nodes on those results. Or maybe not, but some other interesting patterns.
But these are just my (probably naive) thoughts. If you do get to know how people measure this please let me know. I am curious.
I am not aware of these metrics. I will just tell you what occurs to me as I think about this. On a general social network like Facebook, I guess trust might be related to the cardinality of your network of friends. if you have thousands of friends to me that means that your level of trust with them should be a lot smaller than if you have just a dozen. There are some things that you simply won't share or say because you don't even know who will see them. So maybe you can take a function over the node degree as a proxy of the level of trust between the user on that node and all users of the nodes directly connected to it.
On a social network directed towards some specific goal, like Research Gate, things might be diferent. Trust might be more content dependent. Trust in Research Gate might mean something like not being afraid of sharing your research problems and ideas because you know that nobody will steal them. So you open yourself towards collaboration. From this point of view, the bigger, more open, and more widely recognized (and available in search engines) you network is, maybe the less likely will be that somebody violates on your trust, because everything is left out on the open for everybody to see. So in this case, the number of nodes, as a proxy for the network importance and degree of stablishment, might give an idea of its trustworthiness.
Combining both ideas, maybe you can think of a metric of the sort Number of nodes of the network / average user node degree as a first idea towards a way to measure trust in a social network. But you would have to leverage/normalize the expression. The idea is that network trust comes from it being bigger, but with very small parts or clusters (the groups of each user closest links)
One thing that comes to mind to me to validate this sort of model is that you could try to develop a sort of a short dictionary of trust words or sentences. It seems a dificult but maybe challenging task. For example I believe the word "dear" could be one of those, because it is a kind word, and trust and kindness are often connected. With this dictionary you could search the social network content against those words or sentences. Maybe you would find a predominance of small degree nodes on those results. Or maybe not, but some other interesting patterns.
But these are just my (probably naive) thoughts. If you do get to know how people measure this please let me know. I am curious.
[1] Cai-Nicolas Ziegler and Georg Lausen, "Propagation Models for Trust and Distrust in Social Networks", Information Systems Frontiers 7:4/5, 337–358, 2005.
[2] Nepal, S, Sherchan, W. , Paris, C., "STrust: A Trust Model for Social Networks", 10th IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom), pp. 841 - 846, Nov. 2011.
I would gather that trust, as our colleague above was hinting at, is variable depending upon the specific social environment. For example, again, as above, Facebook or Twitter might have a different trust metric than LinkedIn. Yes, I would agree to develop that dictionary of trust terms, but I would also begin to develop a spectrum of what trust means across social environments....what trust would one be seeking/feeling in the social media environment? Professional networking environment? Trust in what exactly? That's what I would be trying to determine.
But, you said "trust, mostly among neighbors"....physical neighbors? But, then you mention users of a social network. In my thought process, those are, or could be, two very different things with two very different meanings of trust.
thanks for mentioning, let me clarify myself, in "neighbours" i meant a graph theoretic sense of the word, imagine you show a social network as a graph and each node will have a set of neighbouring nodes; and i agree it seems the definition of "trust" is depending on for what kind of social network it's used;
Surely, as I suspected. The reason why I brought up the physical neighbors topic was b/c there is a vein of research in the sociological literature that look at literal neighbors in a community and the levels of trust among them, predictors, etc....
You cannot measure "trust" per se other than in a personal interview. But you can measure indicators of trust by counting interactions and referrals over time. Specifically, frequency of contact, both directed and undireected, and per content and content-neutral. In SNA terms, you are looking for centrality based on in-degree and out-degree measures and temporal measurements of brokerage (referrals) from a neighbor with high frequency of contact to the first actor of other neighbors or non-neighbors to that first actor.
May I suggest the paper "Merging trust in collaborative filtering to alleviate data sparsity and cold start by Guibing Guo, Jie Zhang, Daniel Thalmann".
Trust of a neighbor can be measured by how much one rates that neighbor.
thank you everyone for the answers; may i ask what do you think about inferring trust mainly and crudely based on structural properties of a user/node on a social network graph? i mean in social networks there are parts which users/nodes are more densely interconnected to each other; can such a measure(density of links in a particular part) be indicating a relation about trust(i'm assuming people before trusting each other must have known one another quite well), is anyone aware of a work that've been done around this?