today Prof David Krackhardt (CMU) gave a very very nice talk titled “Simmelian Ties in Organizations “. david krackhardt greatly contributed to the discipline of cognitive social networks and has extensively studied the power of simmelian ties in organizations (his bio). here is the result of my live blogging during his talk:
what’s a simmelian tie? it’s a tie embedded in closed triples. think about a couple: him-her is a dyadic relationship. then think about the same couple but with a baby now: the dyadic relationship him-her becomes a simmelian tie in 1908, simmel argued that the fundamental building block for understanding social relations is a triad not a dyad! more specifically: a solely dyadic relationship is quantitatively different from a relationship embedded in a group. by different, he didn’t refer only to the strength of the tie (to see why, think about the example of the relationship him-her before and after the baby. the relationship change is NOT just function of strength, of course)
Why simmelian ties matter. david showed that simmelian ties matter for different reasons:
1) cooperation is affected by whether a dyadic relationship is embedded in a triad or not. based on iterations of a prisoner dilemma game among people whose social network was known, he concluded that if my relationship with you is embedded in a triad, then I’m more likely to cooperate with you than if the relationship wasn’t in a triad. big conclusion: cooperation may not be governed by reciprocity (by the famous “norm of reciprocity”) but may be governed by the fact that a 3rd person is “watching us”
2) ties embedded in cliques last longer [stability of tie types, krackhardt'98]. david analyzed the data coming from the newcomp study in which 17 new students were given free lodging in exchange for filling out a network questionnaire each and every week for 15 weeks! by plotting p(me-you friend at week w+1 | me-you friend at week w) versus weeks, he showed that the decay rate of simmelian ties is very slow compared to sole symmetric ties and to asymmetric ties (which were the faster to decay).
3) role analysis based on simmelian ties is preferable . the purpose of role analysis in social network studies
is to find a relatively small number of roles that explain a substantive amount of interaction patterns. one says that A and B are in the same role set, if A’s ties and B’s ties are highly correlated. in a study of a company called, say, “silicon systems”, david found that, by removing all the ties that are not simmelian, one is able to find the best set of roles. in that study, people linked by simmelian ties tended to form an “affective/emotional group”
4) generating synthetic network s based on the concept of simmelian ties creates more realistic networks than generating networks based on heiderian assumptions. heider assumed that people dislike imbalance (so they strive for balance) and that balanced relationships are simmelian ties, that is ties that are reciprocal and transitive. by contrast, simmel assumed that triadic structures are stable once they become both reciprocal and transitive. on the newcomb data, david showed that simmelian assumptions work far better than heiderian assumptions.
4) simmelian ties explain generation of innovations. based on a study of a social network across variety of research labs, david showed that the more strong/simmelian ties researchers had, the more likely those researchers were not highly productive. more importantly, productivity (in terms of number of patents) was best explained by whether one bridged simmelian ties or not. the “best” researchers were bridging simmelian ties. for more, see [the role of bridge, strong ties, and simmelian ties in generation of innovations]. the definition of the embedness index E-I of a node is interesting…