Gabriel Rossman
Sociological Science, March 3, 2014
DOI 10.15195/v1.a5
Abstract
This article models the implications of innovations being nested within categories. In effect, social actors assess the legitimacy of innovations vis-à-vis conformity to categories such that a sufficiently legitimate innovation may be adopted without direct reference to the behavior of peers. However, when innovations lack categorical legitimacy, actors default to proximately peer-oriented heuristics such as information cascades. Eventually, if enough similarly novel innovations achieve widespread popularity, their conventions will become accepted as a legitimate category. Thus density creates legitimacy, but this density can be at the level of the particular innovation or of the category within which it is embedded.
This article models the implications of innovations being nested within categories. In effect, social actors assess the legitimacy of innovations vis-à-vis conformity to categories such that a sufficiently legitimate innovation may be adopted without direct reference to the behavior of peers. However, when innovations lack categorical legitimacy, actors default to proximately peer-oriented heuristics such as information cascades. Eventually, if enough similarly novel innovations achieve widespread popularity, their conventions will become accepted as a legitimate category. Thus density creates legitimacy, but this density can be at the level of the particular innovation or of the category within which it is embedded.
- Citation: Rossman, Gabriel. 2014. “The Diffusion of the Legitimate and the Diffusion of Legitimacy.” Sociological Science 1: 49–69.
- Received: September 17, 2013
- Accepted: September 20, 2013
- Editors: Jesper Sørensen, Ezra Zuckerman
- DOI: 10.15195/v1.a5
This is a great, thought provoking article piece which argues that the strongest “social” forms of diffusion may be invisible. Gabriel Rossman’s key idea, as I understand it, is that highly legitimate, cognitively available actions are often taken without reference to what others are doing — they don’t require the support provided by witnessing similar behavior elsewhere. Less legitimate actions are by contrast more likely to be contagious — faciliated by adoption elsewhere, via safety in numbers, spread of the perception that a given action is possible, or won’t be sanctioned, etc).
I have tended to argue the opposite, that more legitimate “innovations” spread more easily via contagion (and non-contagion) than do less legitimate innovations — because news of the practice is more public, explanations of why the innovation is a good idea are more forthcoming and more compelling, etc. An extreme case is the diffusion of an illegal behavior, where secrecy is paramount and its very hard for news about what others do, and how they do it, to spread. The main study I know that tries to distinguish diffusion patterns based on legitimacy is Davis & Greve, which argues that the more legitimate practice (poison pills) diffuses faster over thinner social ties than the less legitimate one (golden parachutes). In Gabriel’s example of audience clapping rather than booing, I would think clapping is more contagious, not less — at a talk yesterday noticed how the clapping at the end of the talk and of the session both involved an acceleration as people primed to clap responded to the first ones who put their hands together, easily reaching a crescendo. This seems a good example of a timed, contagious behavior that feeds on its acceptability and legitimacy (standing ovations would be another good example). I see that an individual who wants to boo might be particularly encouraged by another booer (boor?), but in contexts where its illegitimate booing is often an isolated, planned activity of a few opponents of the speaker that has trouble converting even their fellow travellers.
Its very hard to work out these issues empirically, because there are so many inter-dependent parts: overall propensity to adopt, degree of overall contagiousness, network-based variation in contagiousness, shifting overall legitimacy, changing benefits to adoption as others adopt, changes to the innovation as it spreads…Simulations can help here, both Gabriel’s idea of influence across types of actions/categories, and I suspect other types of models. But I would think that the key work will be empirical in more carefully comparing adoption patterns across innovations, especially if one also looks below the surface pattern of adoption dates to the motives of actors.
In any case, Gabriel is raising a fundamental issue for approaches to diffusion that seek to be comparative across “innovations”. He also expands our vocabulary and imagination concerning what we mean by social, and pointing to forms of social influence that are more profound and stronger than social networks — which is a terribly weak form!
Sincere thanks to David Strang for his careful reading, general appreciation, and specific thoughts about my paper. I think his thoughts are very interesting and considering them improves our understanding of diffusion. However first a simple and straightforward point of empirical disagreement. My understanding both from reading Heritage and Greatbatch (1986) and Clayman (1993) and discussing it with Heritage and Clayman (both of whom our colleagues) is that it is indeed the case that booing generally starts slow then builds whereas applause generally starts almost simultaneously. This may have to do with specific circumstances and in particular how obvious it is that applause is merited, but in general my understanding is that the “slow clap” is a film trope that seldom occurs in reality. On the other hand issues like the onset of standing ovations or the cessation of applause probably do follow s-curves.
However let’s focus on the more general point, which is that illegitimate behavior will tend to be furtive/inconspicuous and therefore less contagious. That is whereas I was focusing on the role of legitimacy to reduce the awareness-adoption gap among actors who are at risk of adoption, David rightly draws our attention to whether adoption will be conspicuous or furtive by of past adopters. I think this is a brilliant point and I’m especially sorry not to have put it together before as I have been thinking about a series of related issues. (In particular furtive shameful behavior is the main theme of my forthcoming paper in Sociological Theory, although I did not connect that argument to diffusion. Likewise, although I did not draw the connection to legitimacy motives for conspicuousness, just two days ago I was having a conversation about how among the most brilliant things Apple did with the iPod was giving it distinctive white earbuds which made adoption of the device extremely conspicuous and so magnified its contagious effect.)
Anyway, David is exactly right to draw our attention to the fact that when adopters attempt to behave furtively this will reduce the extent to which their effect can be contagious. I would add to this the point that this may be a point where we may see the effects vary depending on whether the endogenous effect occurs through contagion, information cascades, or network externalities. For instance, if I were to install a hookup app on my phone but were ashamed to tell my peers about my intentions to commit digitally abbetted adultery then my adoption of the app would not be contagious, but it would still increase the app’s download count and so make it more visible to other users of iTunes or GooglePlay, and would increase the thickness of the hookup market and so increase the app’s network externalities (or depending on your opinion of me, impose congestion costs, though I leave that distinction to the reader). Why this matters is that for an actor to furtively adopt an illegitimate innovation may reduce one form of endogenous diffusion but not other forms, though it depends on the information regime. If adoption takes the form of buying a lowbrow novel or installing a sleazy app then I can refrain from telling my friends but it will still show up in the bestseller list and so the consequentiality of furtive adoption depends on how important you think word-of-mouth is relative to information cascades. On the other hand, not every behavior is observable to market information regimes. And indeed there is no SoundScan for golden parachutes.
That said, while David has identified a key intervening mechanism for the endogenous effect, I do not believe that this implies an exogenous effect. Obviously if we identify the overall shape of the CDF curve as the ratio of a to b then it is a mathematical identity that reducing b increases the ratio of a to b. However there is no reason to suspect that a itself goes up and for reasons suggested in my article I think it is likely that a goes down when the innovation is illegitimate. To return to the hookup app example, if I download the app but refrain from telling my friends about it, this means there is no contagious effect for my friends but there is also no reason that my secret adoption should make my friends (or anyone else) any more likely to respond to advertising for the app.
I fully sympathize with David’s closing remark, that diffusion is an issue with very high causal density and there are many scope conditions to consider and in particular that both simulations and empirical work are necessary to unpack and identify them. In the paper I tried to identify many scope conditions for so-called exogenous diffusion (p 54, pp. 58-61) but in my focus on this issue I neglected to acknowledge that there are also important scope conditions or exacerbating conditions for endogenous diffusion, such as how conspicuous the past adopters have allowed their adoption to be.
I very much appreciated David’s reaction and also where Gabriel took it– i.e., some interesting speculation about the implications of furtive adoption of an innovation.
That said, I am worried that there is a miscommunication between you on the core issue in Gabriel’s paper. I read David as saying that it is odd that more legitimate practices are ‘less contagious’ than are less legitimate practices. This does seem odd, and David seems skeptical that when we put this matter to data, that this odd prediction will hold up.
I think though that there is a way to clarify the matter such that David’s way of thinking meshes with Gabriel’s (and then the good news I guess is that we still don’t have to do any empirics!)
In particular, I think Gabriel’s key point is that the most legitimated practices are such that they don’t *need* contagion to spread. Rather, they spread “instantaneously” on the basis of common exposure to the same cues (e.g., we all see that it’s raining so we open our umbrellas; we all see a great performance and so we all clap– without waiting)
Moreover, I think Gabriel would agree that the more legitimate the practice, the easier does the practice spread via diffusion. For example, I think that there is an easy bridging proposal for your views on booing versus clapping. Basically, I think you would both agree that the average adoption threshold for booing would be higher than it would be for clapping, and that as a result, attempts to get a crowd to boo often fail. And you would both agree that instances of applause differ in that sometimes people are indeed sensitive to adoption thresholds (e.g., at events where it is not clear when exactly we’re supposed to clap) and others where we know exactly when to clap (at the end of the star spangled banner) and it is effectively instantaneous (because it is so “legitimate,” in Gabriel’s terms.
In short, while the issue of furtiveness is a very interesting one, I think it is a bit of a tangent relative to the issue that David raised. But actually, when we look at the issue David raised for a moment, there is nothing incompatible about his idea that legitimate practices are more contagious than illegitimate ones (at least through public channels, per Gabriel’s point). It’s just that when they are *really* legitimate, they can spread even without contagion– the key point of Gabriel’s article.
Since the code provided in the supplemental materials for the simulation and graphs is for Stata, I tried to reproduce it in R. You can find it here: https://www.mirosa.org/blog/post/2015/01/Rossmans-Simulation-in-R/
I was reading a paper: Delmestri & Wezel (2011)
http://www.palgrave-journals.com/jibs/journal/v42/n6/pdf/jibs201122a.pdf
and was reminded of this one. The linked paper suggests different diffusion of multiplex cinemas depending on whether the logic underlying multiplex cinemas (cinema-as-commerce) is new to a country or not (cinema-as-art). Their data show that cinema density’s effect is smaller if the logic behind multiplex cinemas is already accepted in a country. Analogously, this paper argues that endogenous diffusion is smaller if an innovation is situated within an already legitimate category.
As the linked paper was not cited, mentioning it may bring some new inspiration, and confidence in this idea’s empirical validity.