Observability
Observability measures how visible the results of adoption are to potential adopters watching others. Rogers found that high observability accelerates diffusion because it allows prospective adopters to evaluate the innovation through others' experience, without needing to conduct their own trials. It is particularly important for crossing from early to mainstream adoption, where the majority relies heavily on signals from peers. The AI transition has introduced a new form of observability — viral demonstration — that amplifies the innovation's best-case performance while concealing typical performance, producing inflated expectations that Rogers warned produce higher rates of disappointment and discontinuation.
In the AI Story
Traditional observability operates through local observation. The farmer sees the neighbor's fields; the physician notices colleagues' patients improving; the teacher observes another classroom. This kind of observation is continuous, contextualized, and includes failures alongside successes.
AI observability operates differently. Developers post demonstrations on X showing spectacular AI-assisted builds. Writers share polished outputs generated in minutes. Executives circulate case studies of dramatic productivity gains. What is visible is the highlight reel; what is invisible is the failed prompts, the corrections, the hours of iteration, the outputs quietly deleted because they did not work.
This selective observability produces the viral demonstration phenomenon: innovations appear more capable and more reliably capable than they are. Potential adopters form expectations calibrated to the highlight reel rather than to typical performance, setting up disappointment when their own adoption fails to replicate the visible successes.
A deeper observability problem is structural. When every adopter reinvents the tool for her own context — see reinvention — the visible output conceals the invisible process. The potential adopter sees what was produced but cannot see how, which means observability of the output does not translate into observability of the process required to produce it.
Origin
Rogers derived observability from empirical findings that innovations whose effects were visible to others (dairy practices producing visibly healthier cattle, agricultural methods producing visibly better yields) diffused faster than innovations whose effects were embedded in processes or outcomes not easily seen from outside.
The AI case introduces complications Rogers did not anticipate: viral amplification of best-case performance, the invisibility of the reinvention process, and algorithmic curation that shapes what becomes observable.
Key Ideas
Local vs. viral observability. Traditional observation is continuous and contextualized; viral observation is selective and decontextualized.
Highlight reel bias. Social media amplifies spectacular successes while hiding typical performance.
Output vs. process observability. AI outputs are highly visible; the reinvention process that produced them is not.
Inflated expectations. Selective observability calibrates potential adopters to best cases, producing disappointment during their own adoption.
Appears in the Orange Pill Cycle
Further reading
- Rogers, Diffusion of Innovations (2003), Chapter 6
- Neil Postman, Amusing Ourselves to Death (1985)
- Shoshana Zuboff, The Age of Surveillance Capitalism (2019)