Quantifying the dynamics of peak innovation in scientific careers

TL;DR

This study analyzes long careers in Computer Science and Physics, revealing that researchers tend to produce their most innovative work early on, with a peak year of high innovation often following a period of low productivity.

Contribution

It uncovers the timing of peak innovation in scientific careers and contrasts it with impact-driven productivity, highlighting a novel pattern across disciplines.

Findings

Most innovative publication occurs earlier than expected in a researcher's career.

Peak innovation year often follows a period of low productivity.

Researchers' peak innovation is associated with a long low productivity phase.

Abstract

We examine the innovation of researchers with long-lived careers in Computer Science and Physics. Despite the epistemological differences between such disciplines, we consistently find that a researcher's most innovative publication occurs earlier than expected if innovation were distributed at random across the sequence of publications in their career, and is accompanied by a peak year in which researchers publish other work which is more innovative than average. Through a series of linear models, we show that the innovation achieved by a researcher during their peak year is higher when it is preceded by a long period of low productivity. These findings are in stark contrast with the dynamics of academic impact, which researchers are incentivised to pursue through high productivity and incremental - less innovative - work by the currently prevalent paradigms of scientific evaluation.