The fast-feedback trap in global innovation Feedback horizons shape what science, capital, and policy can see, reward, and scale

This is a preprint version of the manuscript. It has not been peer reviewed, has not been accepted for publication, and should not be cited as a peer-reviewed article. The manuscript is being shared to establish scholarly priority and to enable academic feedback. No media promotion, press release, or public publicity campaign is associated with this preprint. Modern innovation systems allocate attention before long-term social value is known. They use early proxies: papers, preliminary data, patents, benchmarks, prototypes, trial endpoints, investment rounds, and deliverables. This is rational when problems return fast, clean, decision-relevant feedback; it becomes a selection error when value is delayed, diffuse, preventive, or counterfactual. Building on the late-blooming insight that early performance and eventual excellence can have different predictors, this Analysis shifts the unit of analysis from individuals to problems. We synthesize 465 source records and 865 coded evidence records across health, climate, artificial intelligence, disaster preparedness, energy, antimicrobial resistance, food systems, and research assessment. The evidence does not yield a pooled effect size or universal causal law. It is consistent with a problem-level selection mechanism: feedback horizon shapes how severity becomes fundable, publishable, investable, and governable. The remedy is temporal pluralism in evaluation: match measurement windows to the feedback structure of the problem. This framework explains why severe problems become institutionally selectable only when their feedback horizon is shortened by benchmarks, surrogate endpoints, mandates, shocks, or credible milestones.