There are some new products and services that are very obviously good — a cure for a deadly disease, for example, or some other type of medical innovation. But other innovations have value that is more uncertain, such as an unproven technology. In her latest research paper, Wharton management professor Valentina Assenova examines the role of social networks, both online and offline, in the spread of these complex innovations. Her paper is titled, “Modeling the Diffusion of Complex Innovations as a Process of Opinion Formation Through Social Networks.” She joined Knowledge@Wharton to discuss her findings about which kinds of innovations spread more quickly than others in different networks, the role of influencers, and what that means for entrepreneurs.
An edited transcript of the conversation follows.
Knowledge@Wharton: What was the inspiration for this research?
Valentina Assenova: The inspiration for this research was looking at the spread of microfinance. Microfinance is one of those innovations that is not obviously good or bad, and there is a lot of mixed evidence around whether or not it is actually beneficial for women, whether it improves welfare and so forth. But it was something that really got me intrigued about the role of public opinions and of social networks — in the sense of people who you talk to for advice, for help in making a decision — and how some of these complex innovations spread.
Knowledge@Wharton: Tell us a little more about what you mean by complex innovations?
Assenova: Complex innovation is essentially an innovation that has a lot of uncertainty around its value for a potential adopter. When we think about a complex innovation, we might think about an unproven technology. This is a technology that nobody yet knows whether it will be more or less beneficial in relation to existing solutions.
Given the uncertainty around this technology, typically there is a need for some kind of social validation for other people to adopt it. This is contrasted to something like a simple innovation, like penicillin or a medical innovation that is very obviously good. With just a few adopters, there is not as much of a need for social validation for something like that to spread in a population.
“There are certain features of network structure that are more conducive to the diffusion of a complex innovation than others.”
Knowledge@Wharton: What is included in this kind of social validation or social network?
Assenova: When we think about social networks in social science and in management, we tend to think of patterns of interaction among people. A social network could be people in the office that you interact with daily. Those interactions can be represented as a graph, as a network, or it could be a network of people that you talk to on Twitter and that you follow. It is essentially who is influencing you in your opinions and in your beliefs about the value of certain ideas or certain technologies.
Knowledge@Wharton: How did you test this?
Assenova: The paper is a theoretical model that looks at learning on these networks using a DeGroot naive learning model. I begin by looking at random networks, then I move on to testing it. In the case of the spread of microfinance in India, the beauty of using these very simplified … random graphs is that we know exactly what is happening on the network. We can test very cleanly what the mechanisms are. In this case, the key mechanism is influence from the opinions of other people.
Knowledge@Wharton: What did you find when you looked at these models?
Assenova: I found a couple of interesting things. The first is that there are certain features of network structure that are more conducive to the diffusion of a complex innovation than others. Remember that we don’t know what the value of these innovations are, and in theory we might never know. What people are going off of is the opinions of other people about the value of these innovations. There are a couple of different features of this model. One of them is that we can manipulate the structure of the network itself in terms of the density — how connected people are to other people — and in terms of the asymmetry of the relationship — whether I am more likely to influence you than you to influence me.
What I found in using these models is that networks with both high density and high asymmetry are optimal for diffusing complex innovations when the barriers to adoption are low. That basically means we have a case of a fairly easy-to-adopt innovation. It’s very simple. It doesn’t have a lot of added steps to what it takes to adopt it.
Conversely, when we have an innovation that is very complex, where the barriers to adoption are very high, it’s actually the opposite. Here we might think of a pretty complex technology that requires a lot of additional input and knowledge about how to use it. So, low density and low asymmetry networks are the ones that are most likely to diffuse those innovations.
Knowledge@Wharton: What does a high density, high asymmetry group versus a low density, low asymmetry group look like in the real world?
Assenova: We can take an office as an example. If we look at an office where people are sitting and talking to each other, and we were to map out the network of their communications, a very high density and a very high asymmetry network would look like one where everybody in that office is talking to everybody else. They are interconnected. Moreover, there is one person or two people in that office who are dominating that conversation. Their opinions are far more influential than anyone else in that group.
By contrast, a low density and a low asymmetry group would look like one where very few people in the office are talking to few others. Communications are much more selective, and some of these differences in whose opinion matters are less pronounced. I am just as likely to listen to your opinion as I am to one of my other colleagues in the office, and there isn’t a single person who is dominating the conversation.
“There is this interplay between my own values and my own thresholds, and the opinions of other people that I am influenced by.”
Knowledge@Wharton: In my life, I am likely to be part of both kinds of groups. How does that come into play when we are looking at the spread of innovation across different parts of life?
Assenova: That is a fantastic question, and it is a topic of a recent stream of research that I have been looking at, which is multiplex networks. In reality, people are embedded in more than one kind of network. You have your friends, your co-workers, your family, and those networks could look very different. We are just beginning to explore how diffusion dynamics might matter in those types of networks. The preliminary findings show that there are different types of multiplexity that are more or less conducive to diffusion, and one key factor that matters is how broadly you are spanning these different networks and how unconnected they are. That would be a pretty big predictor of whether you, as a key person who is a spanner across these networks, are able to influence other people within them.
Knowledge@Wharton: How does it help somebody who is trying to spread an innovation to understand the networks and how they influence people?
Assenova: That is a great question. When it comes to innovations with unknown value, like new technologies, what are the best communities to target and who are the best people to target for getting the word out and really promoting adoption? I think this research has a couple of key implications for an entrepreneur or practitioner. One of them is to really think about the barriers to adoption for this technology. Are those barriers relatively high or fairly low?
If somebody is developing a fairly simple app that they believe would be widely valuable and applicable, then targeting a high density and high asymmetry network where they are identifying the key influencer would be the way to spread it, and that would be the way to promote its very rapid diffusion.
Conversely, if somebody is developing a biomedical innovation that is difficult to understand and difficult to form an opinion about, then one might want to target a network that has lower density and lower asymmetry, such as a network among expert physicians who might be trained to use this technology and then talk to each other and rationally evaluate what the costs and the benefits of this technology are. In very practical terms, it just means it provides entrepreneurs a way of selecting the right audience and the right communities for diffusing their technologies.
Knowledge@Wharton: You have to understand who your customers are or who they could be, correct?
Assenova: Exactly. And who are the key people who will be forming an opinion about the value of what you are proposing, and what is the best way of creating social contagion within the networks of people that they touch and that they communicate with. I think that is a key takeaway in understanding what it means for a specific technology in a specific community that somebody is trying to target.
Knowledge@Wharton: I would think understanding how they are going to communicate about the innovation is also important. Are they going to do it via online social networks? Is it going to be word of mouth? Is it going to be something else?
Assenova: This model is agnostic as to the mode of implementation. Certainly in these types of networks, the assumption is that the influence that people are getting is primarily through the evaluations of other people, and that matters. It’s not the kind of technology where it’s easy to just go on a website, read about it and make sense of it.
This is really where the complexity comes from. It is this need for social validation through other people in a network to understand that technology. But certainly when it comes to these evaluations, trying to understand how the structure of the networks shapes the formation of these opinions and whether people are forming a consensus that something is really valuable or not, is an important element of making it successful.
“The fact that you can diffuse an innovation more quickly and more broadly in a sparser network … calls into question the conventional wisdom around choosing key influencers.”
Knowledge@Wharton: Were you surprised by any of the findings?
Assenova: I would say I was a bit surprised by the second finding of really looking at how the thresholds to adoption moderate the benefits of density and asymmetry because I think the popular conception in the literature is that density is always or relatively good. Density and asymmetry would be conducive to diffusion. When we talk about key influencers — that is exactly what this literature is referring to. These are the people who are very, very connected in a network and can immediately influence lots of other people by posting something on Twitter or voicing an opinion. What is surprising is that these kinds of networks and those types of people are not universally beneficial for diffusion. It depends on the barriers to adoption related to specific technology.
The fact that you can diffuse an innovation more quickly and more broadly in a sparser network is surprising and interesting, and it calls into question the conventional wisdom around choosing key influencers and choosing high-density networks.
Knowledge@Wharton: What is it about that high barrier to entry that created that result, where they did better in networks that did not have a lot of asymmetry and were not very dense?
Assenova: I think there are two elements to it, and there are two elements that I test in the model. One of them is the fact that people have a particular threshold for adoption. For example, I am willing to switch to the new iPhone if it is X many times better along the dimensions that I value in relation to the existing technology. People have this threshold about the value they would need to be able to get out of this for it to be worthwhile switching from the existing solution. The second element of that is related to threshold to adoption. There is obviously a consensus value that is forming within a group of people that I know, so if I value your opinion quite a bit and you’ve switched to the new technology and you are changing my opinion about it, I am more likely to switch over.
There is this interplay between my own values and my own thresholds, and the opinions of other people that I am influenced by. It’s that interplay that really does affect what people end up doing and how broadly an innovation spreads within a population.
Knowledge@Wharton: What is next for this research?
Assenova: There are a couple of different projects that I have been working on. One is looking at the role of multiplexity in networks and the role of community leaders or opinion leaders in these networks and how they span these networks over time. I have a second paper that is a follow-up to the study, really parsing how multiplexity, this overlap in different networks that people have, affects the diffusion of these innovations. I have another paper looking at how playing multiple roles within these communities — say, being both a developer and a user on a platform — might affect the diffusion of some of these innovations.
The post How social networks contribute to spread of unproven innovations appeared first on Businessamlive.