Pro-innovation bias

Pro-innovation bias begins when adoption is treated as the answer before the evaluation begins. In diffusion research, the term describes an assumption that an innovation should spread rapidly and broadly, while rejection, discontinuation or local modification is treated as a problem. It is an established criticism of the diffusion literature, not a diagnosis of everyone who likes new technology.

Everett M. Rogers traced the criticism to work he published with F. Floyd Shoemaker in 1971 and developed it in the 1983 edition of Diffusion of Innovations. His focus was the evidence system around adoption: which innovations researchers chose to study, who sponsored the work, and which outcomes left records that were easy to find.

Illustrative hypothetical: The library locker dashboard

A county library receives a grant to test after-hours book lockers at eight branches. The project dashboard records launches and successful pickups. Those numbers are easy to collect, display and share with the funder.

Other outcomes sit outside the dashboard. Some branches decline because the lockers do not fit their power, connectivity, accessibility or staffing conditions. A branch stops using the equipment after repeated jams and maintenance calls. Other branches alter the pickup rules or hardware to make the service workable locally.

The funder expects take-up, so the team interprets a slow start as a communication failure. Its next proposal recommends rolling out the original design across the county. The report does not ask whether refusal, abandonment or modification was reasonable.

This is a constructed example, not a documented library program. The pro-innovation pattern lies in the denominator. Launches and pickups count as evidence; nonadoption, discontinuation and reinvention disappear. Nothing in the scene establishes that the lockers are good or bad.

Where the term comes from

Rogers used change agent for an individual who tries to influence innovation decisions in a direction desired by a change agency. He argued that these agencies often fund diffusion research and understandably want the innovation to spread. Researchers may then inherit the sponsor's question: How do we speed adoption? A different question, such as whether the innovation fits each setting, receives less attention.

He also identified a practical selection problem. Successful diffusions produce visible adoption records. Rejected or discontinued ideas are harder to identify after the fact, and locally modified versions are harder to classify as one stable intervention. Researchers also have reasons to choose fast-moving innovations because they seem timely and consequential. Rogers presented these as explanations for a research tradition's orientation, not as a measured psychological mechanism in every adopter.

The result is an uneven map. The literature can reveal more about fast diffusion than slow diffusion, adoption than rejection, and continued use than discontinuation. That gap matters because an adoption rate cannot tell us by itself whether the innovation solved the intended problem, imposed new costs, or needed redesign.

What direct evidence shows

A 2016 policy-diffusion study tested the success-case problem in a setting with an unusually complete denominator. Andrew Karch, Sean Nicholson-Crotty, Neal Woods and Ann Bowman analyzed interstate compacts open to all 50 U.S. states, including compacts with very different adoption patterns. They used pooled event-history models rather than restricting the analysis to widely adopted policies.

Case selection changed the conclusions. Studies limited to popular compacts appeared to give too much weight to geographic pressure and policy attributes, and too little to professional groups and lessons from states that had already joined. This is direct evidence that choosing diffusion successes can distort inference. It is not evidence that consumers everywhere misjudge new technologies.

The distinction sets a boundary around the Hello to Halo catalog claim about benefits and risks. When adoption has already been set as the desired result, anticipated gains can crowd out closer study of hazards, refusal, discontinued use or redesign (Rogers, 1983; Karch et al., 2016). The evidence reviewed here does not establish that individuals and organizations universally overestimate benefits and underestimate risks.

Adoption is not the same as benefit

People can adopt an innovation because it appears to offer an advantage, fits existing routines or seems manageable. They can reject it because it conflicts with local needs or creates costs that a distant sponsor does not see. Neither decision proves the innovation's objective value.

In a review and meta-analysis of 75 articles, Louis Tornatzky and Katherine Klein found that perceived compatibility, relative advantage and complexity had the most consistent relationships with adoption. Those are adoption correlates. A perceived advantage may be accurate, exaggerated or incomplete, and the research design must supply the comparison needed to tell the difference.

System conditions also matter. A systematic review of innovation in health-service organizations linked diffusion and implementation to the innovation, adopters, communication, organizational readiness and the wider setting. Its authors said their evidence-informed model should be tested more widely rather than treated as universal (Greenhalgh et al., 2004).

Later, a UK research team followed six health and social care technology cases for as long as three years and checked its developing framework against 10 additional cases. The resulting NASSS framework examines nonadoption, abandonment, scale-up, spread, sustainability and adaptation across the technology, users, organizations and wider context. Its authors describe it as a way to guide inquiry, not a formula that predicts every implementation (Greenhalgh et al., 2017).

These studies do not prove that a pro-innovation bias caused an implementation to struggle. They show why calling every refusal “resistance” or every rollout “success” discards information about the social system in which adoption occurs.

Is pro-innovation bias always harmful?

Even a broader preference for novel alternatives does not have one fixed consequence. Oliver Baumann and Dirk Martignoni built a formal simulation in which modeled firms could systematically inflate the expected value of new options. A mild bias encouraged exploration and improved long-run performance under some complex, stable conditions. A stronger bias could damage performance, and unbiased evaluation performed best under most other modeled conditions.

Those results came from simulated organizations moving across constructed performance landscapes. They do not show how common the bias is in real firms. They do show that “pro-innovation” and “harmful” are not synonyms. Environment, search breadth, managerial control and selection pressure changed the model's result.

The cultural evidence is thinner still. The direct policy study is based on U.S. interstate compacts. The detailed technology cases are mainly from UK health and social care, and the organization result is a simulation. A separate meta-analysis of consumer innovation resistance combined studies from 24 countries or regions and a total sample of 10,463, but it measured resistance barriers rather than pro-innovation bias (Leong et al., 2021). Its cultural result cannot establish that Topic 217 is equally common across societies.

What the evidence does and does not justify

The clearest documented consequence is an evidence problem. If studies overselect successful diffusions, their estimates of why adoption happened can change. If reports omit rejection, discontinuation and reinvention, decision makers learn less about fit and failure.

A specific innovation may waste money, create harm or delay safeguards. Those claims require evidence about that innovation, its alternative and the people who bear its costs. The label alone cannot supply the causal chain.

The same limit applies to remedies. Rogers recommended studying innovations while diffusion is still underway, comparing successful and unsuccessful cases, and examining rejection, discontinuation, reinvention and wider context. NASSS offers a structured way to ask about implementation complexity. Neither source shows that naming the bias, generating one counterargument or completing a checklist reliably debiases a decision.