Diffusion research is used in many fields and promotes interdisciplinary research by providing a common theoretical framework while still supporting various research goals (growth; institutional, behavioral, social, or policy change, information exchange; etc.). Innovations are particularly good subjects to study because of the relative ease of tracking their diffusion (they are often different, salient, and memorable). Diffusion research is "applied," studying real-world processes, but the theoretical framework allows for generalizations of diffusions. The framework is straightforward to apply.
Despite these benefits, there are a number of shortcomings in diffusion research as it is today. The first and most serious is the pro-innovation bias (recognized in the 1970's but not remedied): most diffusion research is innovations that diffuse successfully, and little has been done on ignorance, rejection, discontinuance, re-invention, or anti-diffusion campaigns (e.g. anti-smoking, anti-STDs). The reasons for the pro-innovation bias are numerous: many diffusions studied had very high relative advantage, much research is funded by change agencies with a vested interest, many diffusions are studied after-the-fact where failed diffusions are hard to trace, and researchers often see rapidly-adopted diffusions as more interesting. Different research methods would mitigate this, such as the collection of data several times throughout the diffusion process, the selection of a wider range of innovations, and the study of the social context of an innovation and individual motivations (especially non-economic) for adopting, rejecting, discontinuing, or re-inventing. (Often, adoption is automatically seen as "rational" and rejection as "irrational," and researchers assume potential adopters see the innovation the same way they do.)
A second shortcoming is the individual-blame bias. An individual-blame perspective would say, "If the shoe doesn't fit, there's something wrong with your foot." Researchers exhibit this by tending to side with the change agents rather than the adopters (hence even the passivity of the term "diffusion"), and thus blaming potential adopters for shortcomings (e.g. inability to understand the innovation) rather than the system (e.g. bad innovation design). Variables correlated with "innovativeness" are all individual variables, rather than system variables such as change agent assistance or appropriateness of the innovation, e.g. late adopters and laggards are seen as "irrational" for their reluctance (and sometimes people who fit the profile of a laggard aren't even contacted, reinforcing the bias against them). It's also easier to analyze individuals than systems (including social networks, the means of diffusion), so individuals are the most common unit of analysis, lending to individual-blame. Some researchers recognize system-blame but feel that the system is harder to change than individuals. It's important for researchers to use various units of analysis, include rejectors in innovation studies, and be open to the possibility that the innovation or the method of diffusion is inappropriate.
A third shortcoming, the "recall problem," relates to the often post-hoc nature of diffusion studies. Rather than conducting one survey after an innovation has diffused and relying on potentially inaccurate recall, researchers should collect data over time, perhaps in field experiments, longitudinal panels, archives, or case studies with cross-checks.
A fourth shortcoming involves causality: "the pro-innovation bias in diffusion research, and the overwhelming reliance on correlational analysis of survey data, often led in the past to avoiding or ignoring the issue of causality among the variables of study." Field experiments are good for determining which variables are actually related causally.
Finally, little research has been done on the consequences of an innovation, such as how the benefits of an innovation are distributed and why. One negative consequence of innovations is the inequity bias: innovations tend to widen socioeconomic gaps, especially in developing countries where the difference between rich and poor can be extreme. In the 1970's, diffusions promoted the dominant paradigm of development (economic growth, capital-intensive technology, centralized planning, and "individual-blame" on the scale of a nation), but often at the expense of underprivileged. Change agents tend to favor wealthier clients, since they can afford to be interested. Now, development is defined more appropriately as "a widely participatory process of social change in a society intended to bring about both social and material advancement ... for the majority of people through their gaining greater control over their environment." Bordenave suggests addressing these research questions when doing diffusion research in developing nations, which also helps it overcome pro-innovation and individual-blame biases.
- What criteria guide the choice of innovations that are to be diffused: (1) promoting the public welfare, (2) increasing production of goods for export, (3) maintaining low prices for urban consumers, or (4) increasing profits for society's elites, such as large landowners and industrialists?
- What influence does society's social structure have on individual innovation-decisions?
- Are the technological innovations that are being diffused appropriate, well proven, and adequate for the stage of socioeconomic development of the nation?
- What are the likely consequences of the technological innovation in terms of employment and unemployment, migration of rural people to already overcrowded cities, and a more equitable distribution of individual incomes?
- Will the innovation widen or narrow socioeconomic gaps?
Chapter 4 discusses the entire lifecycle of an innovation, which may include the following stages: recognition of a problem (can be done by "lead users"), research (basic or applied, can be done by "lead users"), development (which can be driven by technological determinism - technology shapes society - or social constructionism - society shapes technology), commercialization, diffusion, adoption, and consequences. Most diffusion research starts with the first adopters, but more research should be done on this full life-cycle.