Diffusion of Social Science (continued):
This repeated theme is Rogers’ paradigmatic Law of Diffusion/Marketing: In deciding whether or not to adopt an innovation, people depend on the opinions of people like them who have already adopted.
For instance, Rogers and Kincaid (1981) studied the diffusion of family planning innovations in 25 Korean villages. Choices were homogenous within villages, creating, say, “pill” or “IUD” villages. Closer study revealed that after an opinion leader adopted, say, an IUD, they shared their personal experience with other villagers; opinion-sharing diffused through interpersonal networks until, after several years of diffusion, the villagers came to use the same method.
Or, consider Dr. William Whyte, who in the mid 1950’s noticed from aerial photographs that air conditioners tended to cluster by neighborhood in Philadelphia. Personally interviewing adopters, Whyte found that neighbors had recommended the experience of air conditioning itself, without recommending a brand, because any cooler air felt good. Whereas tetracyline had specific medical advantages, family planning and air conditioning were commodities that simply satisfied; but everyone still relied on a friend’s opinion.
Bridging Links.
As Rogers noted, Dr. Mark Granovetter (1973) produced the classic study of “The Strength of Weak Ties.” Granovetter gathered data from 282 people in Newton Massachusetts who had found new jobs in the last year. Granovetter discovered that successful leads came, not from close friends, but from acquaintenances “only marginally included in the current network of contacts, such as an old college friend or a former workmate or employer, with whom sporadic contact had been maintained.” Only 17% of Newton respondents found their new jobs through close friends and relatives.
Why? Perhaps, because your close friends are also friends of each other, already sharing among your clique both a shared lifestyle and held-in-common information. Such networks can be described as interlocking personal networks in contrast to radial personal networks that consist of a set of individuals all linked to a focal person, but who do not interact with one another. Rogers (p. 308) used the term, communication proximity, defined as “the degree to which two linked individuals in a network have personal communication networks that overlap,” to specify this generalization:
The information exchange potential of communication network links is negatively related to their degree of (1) communication proximity, and (2) homophily.
Rogers speculated that the strength of weak ties lies in their ability to bridge networks by dispersing information. Whereas, the strength of strong ties is its ability to convey interpersonal influence. Aren’t strong ties what relationship marketing is all about? (The other part is getting to know what someone wants, needs now, and will pay for.)
Rogers discussed the link that was significant in recruiting college students to the 1964 Mississippi Freedom Summer. Twenty years after that epic summer in which three volunteers to register Black voters were murdered by racists, Dr. Doug McAdams (1986) obtained access to the applications of 720 volunteers and 239 students who originally volunteered but withdrew prior to departing for Mississippi. McAdams conducted in-depth interviews with eighty participants and withdrawers. Their application forms included the names of ten people who the students wanted to be kept informed of their whereabouts and summer activities. These ten listed people were defined to represent “strong ties.” Indirect network links were identified as anyone not on the list of ten strong ties, but who was one degree of separation distant, that is, named by someone whom the student had had named. Any indirect link was classified as a “weak tie.”
Neither weak ties, nor college, major, personal characteristics, prior activism, race, or distance from Mississippi had much impact in explaining whether or not a student did go to Mississippi. By far, the best predictor of going, which entailed considerable personal risk, was a strong relationship with another participant who was going or to a Freedom Summer activist. Having a friend dropout or having upset parents who exerted counter pressure were typical reasons for withdrawal. It is not what a friend of a friend says or wants, but only what your friend says or wants that counts for you. Only friends, intimates, and opinion leaders with strong ties to us exert much interpersonal influence.
Critical Mass.
Rogers recognized that a concept used by scholars of social movements, who had borrowed it from physicists, was essential to explain an important aspect of the diffusion of interactive adoptions, where interactivity occurs among parties using an innovative communication channel. This crucial concept was critical mass—a point in the social process when diffusion becomes self-sustaining.
As non-interactive innovations become more popular, their value for subsequent adopters is perceived as increasing. This property is called sequential interdependence. For instance, in 1981, a forerunner of the Internet called BITNET (Because It’s Time NETwork) was begun at CUNY as a network for sharing information among Universities about grants and research projects. Each Eastern University who created a link to BITNET increased the sequential benefit for subsequent adopters by reducing the initial cost of adoption because each new member only had to pay the cost of a leased telephone line to the nearest participating University.
In interactive adoptions, however, there is a process of reciprocal interdependence: not only do earlier adopters influence later adopters, but also later adopters influence earlier adopters, increasing value for all as the potential for interactivity increases.
For instance, in 1982, UCal Berkeley leased a transcontinental phone line to join BITNET. Other West Coast Universities joined, each paying a portion of the cost of that leased line to the East Coast. Adoptions took off; by 1983, nineteen members had adopted. Between 1984 and 1985, adoptions doubled every six months as lines were leased in Canada, Europe, and Japan.
Compared to the usual S-shaped curve of adoption, the curve for an interactive innovation is depicted as more sinusoidal, with a steeper inflection and rate of adoption following the achievement of critical mass.
Following Dr. Thomas Schelling (1978), Rogers (p. 318) viewed critical mass as a concept that “bears on the relationship between the behavior of individuals and the larger systems of which they are a part.” Rogers called this “a crucial cross-level relationship;” a complexity theorist might choose “an emergent at the system level based on local individual decisions.” Schelling noted that an individual’s actions often depended on how many other individuals around him were behaving in a particular way so that a self-sustaining critical mass occurred across phenomena as diverse as jaywalking, panic behavior, riots, fashion, political movements, epidemiology, and species extinction.
The interest of social scientists in such phenomena probably stemmed from Dr. Mancur Olson’s (1965), The Logic of Collective Action. Olson pointed out that, “Even if all the individuals in a large group are rational and self-interested, and would gain, if, as a group, they acted to achieve their common interest or objective, they still would not act to achieve that common or group interest.” As in Dr. Garrett Hardin’s Tragedy of the Commons, each individual pursues his own rational course of action in a way that drives the system to disaster. The individual’s decision, say, to graze just one more cow on the commons, when combined with similar rational, but shortsighted, independent decisions from many individuals creates a collective system-disaster for the commons. Or, to strike closer to home, when many corporations and many investors try to keep pace with the unusual excess rewards in an accelerating market, they jointly create a bubble—a self-sustaining critical mass of irrational exuberance—that will eventually burst, leaving them all surprised and filled with regrets.
In describing network externalities and critical mass in new telecommunications services, Dr. David Allen (1983) pointed out that reaching critical mass depends on “everybody watching while being watched.” If first adopters thought only of their own immediate benefits then no one would adopt a reciprocally interdependent service. “I won’t buy a fax machine unless you buy one too. Although it is in all of our best interests, I won’t be first; but also I won’t be last and lose out on the collective benefit.” Hence, everyone anticipates and watches to see who adopts an interactive innovation while everybody else also watches while being watched.
Adoption-decisions are based on what people expect others to do. A critical mass is a tipping point in diffusion occurs when everybody suddenly perceives everybody else as deciding to adopt the innovation. Markus (1987) noted that just as a critical mass triggers self-sustaining adoption, it also speeds up discontinuance of the competing innovation. Just as people can jump on a popular bandwagon, they can jump off a faded star because everybody is watching everybody while also being watched.
Just as a critical mass defines at a macro level a collective tipping point for a system, a threshold defines at a micro level an individual’s tipping point.
Granovetter (1978) described this classic threshold model:
“Imagine 100 people milling around in a square—a potential riot situation. Suppose their riot thresholds are distributed as follows: There is one individual with threshold 0, one with threshold 1, one with threshold 2, and so on up to the last individual with threshold 99. This is a distribution of thresholds. The outcome and is clear and could be described as a ‘bandwagon’ or ‘domino” effect: The person with threshold 0, the ‘instigator,’ engages in riot behavior—breaks a window, say. This activates the person with threshold 1. The activity of these two people activates the person with threshold 2, and so on, until all 100 people have joined.” Hence, if the person with threshold 1 were removed, there would be no riot, just a lone vandal.
Speculation. Assume three samples of distributions of thresholds where there is, say, a 3-unit gap in Set 1 between a person with threshold 2 and a person with threshold 5, a 3-unit gap in Set 2 between a person with a threshold 22 and a person with threshold 25, and a 3-unit gap in Set 3 between a person with threshold 92 and a person with threshold 95. Models could be formed to specify various parameters and their relative strengths as a function of their distributions through time.
For instance, a model might assume that an early, say, 3-unit gap between people with thresholds 2 and 5 may or may not be sufficient to disrupt riotous behavior given their comparatively low threshold and the growing strength of contagion, which might be assumed to increase with the number of participating rioters. As contagion grows in strength, it might reach critical mass and become able to bridge increasingly large gaps from the inflection point on, say, above a threshold of 20. Thus, bridging 22 and 25, but not bridging the 3-unit gap between 12 and 15, which came before critical mass. Next, assume that contagion might reach an asymptote with larger numbers. Here, it might no longer bridge the gap between thresholds 92 and 95. Or, a modeler might assume that persons with very high thresholds, say from the 90th percentile up might have unusual resistance to the violent contagion induced by other rioters. That is, models could vary the size of the gap between thresholds, the contagiousness in riots of increasing size, virulence of riots, and the like in samples of individuals varying in high- or low-riot-thresholds and the like.
Such models of the network diffusion of innovations are applicable in marketing, which can use existing data to form its initial assumption, which could be subject to research using computer simulations patterned after those introduced by Dr. John Holland and extended to networks by others, such as Dr. Albert-Laszlo Barabasi. Such computer simulations have considerable investigative and demonstrative power. (To demonstrate this kind of theory and research effectively would require me to review the recent work on networks stemming from complexity theory. It is a recent paradigm shift from this model of the diffusion of innovations that emanated 60 years ago, so brace yourself for a future presentation of the present state of the art.)
In 1995, Rogers s knitted diffusion networks into the research paradigm of his roots, which are “socially” rather than “network-” oriented. He (p. 322) pointed out that Dr. Jack Valente (1994) introduced the idea “that network thresholds can be used to classify individuals according to (1) their innovativeness with respect to their system and (2) their innovativeness with respect to their personal network partners. Valente’s analysis showed that some individual were late adopters in spite of exposure to the innovation by their personal network partners; whereas other late adopters because their personal cluster had not been exposed to the innovation by network links.
So, Rogers’ final and eighteenth generalization in his chapter on diffusion networks remains steadfastly locked into the individual, rather than the network, level of analysis:
An individual is more likely to adopt an innovation if more of the other individuals in his or her personal network have adopted previously.
It remains to be seen what Everett Rogers will do in his forthcoming Fifth Edition. And, whether he includes and discusses the recent work on networks by complexity theorists is important because his classic work is widely read as the introduction to diffusion research.
Nonetheless, dependent on the distribution of thresholds and the set of existing links between sets of local clusters defining a distributed diffusion network, many models of diffusion networks are possible. The best-known model in social science is the small- world web model, which was rendered in mathematical graph-form through the use of computer simulation by Drs. Duncan Watts and Steven Strogatz and published in Science in 1998.
The magic of their model was its demonstration from a conceptual point of view of precisely how it was possible that a social world, given a few random weak ties, could form a seamless web having only six degrees of separation between diverse individuals, each predominantly interacting within their own tightly clustered local group, but still connected to everybody in the world even when the individuals-to-be linked were located anywhere in the world.
Their simple model captured the real-life complexities of sets of tightly clustered communities and the intertwined social structures of groups and communities in an increasingly global world. The non-interactive ties of TV culturally unite the world. But, interactive ties forged through using the worldwide Web promises to bridge distances beyond the reach of local-clusters. Affinity groups create new virtual-clusters with strong ties, yet they remain widely distributed. Diffusion accelerates; ideas grow exponentially.
In the diffusion of innovations there are multiple potential thresholds that depend on how finely diffusion networks are segmented, from local to worldwide; from segmented by demographics to segmented by lifestyles; from proximal and overlapping in tight local clusters to distal and virtual clusters; from regional homophilous clusters to worldwide clusters where physical distances and cultural heterophily give way to newly developed strong ties; from distant clusters linked by weak ties to distant clusters linked by strong ties; from clusters linked directly to clusters linked indirectly by up to six degrees of separation.
However, you need a small-world model to account for a social phenomenon like the diffusion of music through a Napsteresque world wide web. Today shared affinity interests, a web culture, and contemporary digital communication-computing technology permit a virtual world to blossom on top of the material world. We have yet to explore all the similarities and differences, all the delights and conundrums, all of the value enhancements and the dangers that this may entail.
Marketing Joins In Diffusion Research
When Rogers published his First Edition in 1962, there were only a handful of marketing studies of diffusion. However, by 1994, the marketing literature had exploded, producing a cumulative total of 585 studies, ranked second among ten disciplines with 15% of all the then available 3, 859 diffusion studies. Much of this research came from social marketing because corporate studies of marketing are often proprietary and confidential. Fortunately, Professors always seem to be willing to share their ideas as freeware.
The biggest impetus to the increase in marketing research came from studies that sought to predict the rate of adoption of new products. The most influential investigator was Dr. Frank Bass (1969), then a Marketing Professor at Perdue University, who introduced the Bass Forecasting Model.
Having reviewed the literature, Bass assumed that potential adopters of an innovation were influenced by two communication channels: (a) mass media, and (b) interpersonal word-of-mouth, with the former concentrated early in the adoption-process and the latter concentrated within the first-half of the S-shaped adoption curve. Bass realized that an S-shaped curve of cumulative adoption, when plotted as frequencies across times, approximated a normal (bell-shaped) curve. This meant that the frequencies of adoption in the first and second halves were symmetrical around the mean. Bass formulated a predictive model of adoption-curve to forecast mathematically how many adoptions occur in a specific time period. The parameters in the Bass model are: (1) a coefficient of mass media influence; (2) a coefficient of interpersonal influence; and, (3) an index of market potential. Market potential was estimated on the basis of frequencies of adoption in early time periods, pilot studies with the new product, or by analogy previous introductions of similar product in similar markets.
Collating data across fifteen research studies, Sultan, Farley, and Lehmann (1990) found a trivial mass media coefficient of only .03, but also a significant interpersonal influence coefficient of .38. This outcome was consistent with the body of data suggesting that adoption-decisions are a predominantly a social process occurring in interpersonal networks, not a mass media inoculation.
It is appropriate to end with a somewhat rephrased version of Rogers’ four familiar-sounding strategies to get to critical mass, which go beyond the most direct approach of giving free service for a limited time period: (1) Target top officials for initial adoption of interactive innovations; (2) Shape individual’s perceptions that the innovation is inevitable, desirable, approaching critical mass; (3) Introduce the innovation to key intact groups that have a compelling need to adopt; and (4) Provide incentives for early adoption until critical mass is reached.
The key to understanding high-tech marketing is captured by universal principles that specify the processes by which an innovation is diffused by communication through social channels in a homophilous diffusion networks that are linked by worldwide weak ties to various heterophilous global, but regional, clusters of local homophilous diffusion networks that enable all of Earth’s people to be separated by no more than six degrees of freedom. More specifically, the master key for marketing requires scientific and practical understanding of networks and their links.
Moore’s Use of The Technology Adoption Life Cycle.
Although the single point to be made here is far from covering the contributions of Geoffrey Moore, who obviously adapted the S-shaped curve discovered in the diffusion of innovations literature into a systematic marketing strategy, I am whetting your appetite for a closer examination of his various models in my next contribution to the RTW Report. I contend that Moore fertilized the ovum generated by the TALC with the sperm of applied social science/marketing ideas. His recipes for how to combine ideas into a variety of new models were innovative and generative. But, cross-fertilizing the TALC with the idea of segmentation produced his grandest offspring.
By segmenting the TALC at each of its temporal stages, Moore created new target markets that required new marketing strategies because Moore could now perceive the segmented homophilous groups within the TALC that served as reference groups while they watched while being watched. |
