HCC class lecture 24 comments John Canny 42005

HCC class lecture 24 comments John Canny 4/20/05

Administrivia

Diffusion The mechanism whereby innovations propagate in a society. Diffusion happens on social networks. It depends on both internal factors (attitudes and behavior of the actors) and structural factors; i. e. the structure of the network.

Diffusion is not “automatic”. Many famous ideas did not diffuse quickly, or not at all. Examples: Boiling water for sanitation: Failed in a Peruvian village because it contravened custom. Citrus fruits for scurvy: Took many, many years in spite of apparently strong evidence. Dvorak keyboard: Clearly superior to QWERTY but inertia seems to prevent adoption.

Elements of Diffusion The innovation: an idea, thing or practice that is perceived as new. Communication Channels: Means or media through which information about the innovation propagates. Time: Diffusion is a dynamic process. A Social System: We are often interested in diffusion within particular social groups. Social networks are central also.

Adoption categories Innovator/Early Adopter: The first 10 -15% of adopters. Early Majority: The remainder of the first half of adopters (15 to 50 th percentile). Late Majority: The second half of adopters (ignoring laggards) Laggards: The last 10 -15% of adopters.

Innovations Technological innovations: Hardware: Presence and diffusion are obvious. Physical form dictates minimum costs, but various strategies can mitigate this: “shaver-and-blades” strategy in printers, projectors, cell phones, DVRs. Software: Includes ideas: Marxism, religions, news, rumors, “just-in-time” manufacturing, etc. Hard to track, adoption is often unclear.

Innovations Perceived advantage: Is important. It includes prestige as well as objective advantage. Compatibility: Must be compatible with existing practice and values. Complexity: Perceived difficulty and “work” to use it. Trialability: Easy “on-ramp” allows users to try it and dispel some of their apprehensions. Observability: Can others see the results and assess them?

Communication channels Mass media: People learn about innovations through mass media, but most will not adopt (innovators may). Scientific evidence: May be more convincing, but still not enough (consumers seem to know a lot about how science is done). Interpersonal channels: The most important channel for most people. Adoption for many happens when they know others whom they trust have adopted.

Heterophily and Homophily: the degree to which individuals who interact are similar in beliefs, education, social status. Heterophily: the degree of difference in these attributes. Innovations move rapidly in homophilous networks. Users trust their peers, and see their own social role in terms of technologies that peers adopt. However, Heterophily is essential for diffusion throughout a large social group. The group will contain many homophilous groups, and the adoption categories (innovator etc. ) are often in different groups.

Innovation/Decision Knowledge: Persuasion: Decision: Implementation: Confirmation:

Innovation/Decision Knowledge: User learns about innovation. Persuasion: Decision: Implementation: Confirmation:

Innovation/Decision Knowledge: User learns about innovation. Persuasion: User forms favorable attitude. Decision: Implementation: Confirmation:

Innovation/Decision Knowledge: User learns about innovation. Persuasion: User forms favorable attitude. Decision: User decides to adopt. Implementation: Confirmation:

Innovation/Decision Knowledge: User learns about innovation. Persuasion: User forms favorable attitude. Decision: User decides to adopt. Implementation: User actually puts innovation to use. Confirmation:

Innovation/Decision Knowledge: User learns about innovation. Persuasion: User forms favorable attitude. Decision: User decides to adopt. Implementation: User actually puts innovation to use. Confirmation: User seeks reinforcement of their decision.

Social Systems Social structure: Includes formal social structures such as bureaucracies, government hierarchies, caste systems, understood social roles. These structures involve influence/power relationships and exchange. Communication structure: How people actually communicate in the network. This is less studied in traditional diffusion theory, but is now much more active because electronic networks are observable.

Opinion leaders and change agents Opinion leaders: Individuals who influence others behavior: tribal elders, sages, columnists, activists. Typically follow community norms, rather than lead them. Change agents: Individuals who attempt to influence the system toward change. Often they are outsiders. A good strategy for them is to marshall the influence of opinion leaders, since their own influence is likely to be limited.

Consequences Desirable/Undesirable: Poor implementation may cause the innovation to fail or backfire (e. g. deforestation/desertification). Direct/Indirect: Innovations have many second-order effects that may be hard to track. Most technical innovations are more accessible to the rich and welleducated and tend to increase social stratification. Anticipated/Unanticipated: Indirect affects may be anticipated or not. Many are impossible to predict, especially since technology adoption often changes social practices in the long term.

Network effects The key to successful diffusion is the conversion of opinion leaders. Once a critical mass of opinion leaders convert, the innovation usually takes off rapidly. Densely-connected graphs or cliques are very important. Converting a few members in a clique makes the innovation seem popular because of individuals who have several neighbors in the clique.

Two-step flow model Researchers originally believed that the mass media influenced individual’s opinions directly (the “hypodermic” needle model). This model was gradually displaced by the “two-step flow” model where ideas flow first from the media to opinion leaders and then to the general public. But the opinion leaders may exert their own or the community’s views first.

Measuring Opinion leadership Sociometric: Ask users who influenced them. Informant ratings: Ask a smaller but well-informed set of users about opinion leaders. Self-designation: Ask the potential leaders themselves if they believe they have influence. Observation: Direct measurement of communication and adoption actions in a network. The most reliable, but traditionally expensive method.

Opinion leaders External communication: Opinion leaders have greater exposure to external communication (media, network). Accessibility: They are accessible and participate extensively in social activities. Status: They generally have high social status. Innovativeness: To be trusted as authorities about innovation, they must innovate regularly.

Other network actors The actors in a diffusion network need not be people. They can also be: Corporations: e. g. offering of broadband services, data services, GPS among telcos. States: Legislation, taxes, social programs etc. follow a particular adoption hierarchy among US states. Nations: Similar phenomena are manifest among european states, e. g. adoption of the euro.

Strength of weak ties. Mark Granovetter showed that weak ties (often not direct or weak direct ties) are very important for certain decisions, like taking a new job. These weak ties are often heterophilous, allowing the actor to step across traditional boundaries. Weak ties do not support much influence, but allow crucial information to flow rapidly.

Individual thresholds Network models allow for explanations of adoption in terms of personal adoption thresholds: the fraction of neighbors who have adopted before the actor is persuaded to adopt. Adoption thresholds vary. They are lowest for innovators, higher for early majority etc. The threshold model explains the “snowballing” effect of adoption.

Discussion Topics T 1: Think about how opinion leadership could be measured with the social network methods we already discussed. T 2: What do you think homophily and heterophily mean as graph properties? i. e. how could you compute them? T 3: Where does diffusion fit among Vygotsky’s genetic domains?