Human decision-making in a non-ergodic additive environment

Research output: Contribution to journalArticlepeer-review


Ergodicity Economics presents an intriguing perspective on the validity of the ergodic hypothesis within economic models and the influence of ergodicity breaking on human decision-making processes. Prior research has illuminated the impact of ergodicity breaking within multiplicative settings. However, the implications of ergodicity breaking within additive dynamics, especially in situations carrying a ‘risk of ruin’ or a complete loss, remain largely unexplored. In our research, we introduce the concept of ‘risk of ruin’ into our decision-making model to examine the effects of non-ergodicity in additive dynamics. Our theoretical framework and experiments show that human decision-makers are sensitive to non-ergodicity within purely additive dynamics. This sensitivity manifests itself in significantly different levels of risk aversion depending on the distance and associated likelihood of ruin. These findings underscore the critical role of time averages in human decisionmaking, suggesting that humans are less irrational than conventionally assumed in behavioural models rooted in expected values. Drawing on evidence from Ergodicity Economics, incorporating non-ergodicity has the potential to illuminate common trends in decision-making within compounding systems, like multiplicative growth dynamics. Our research underscores a similar potential for understanding decision-making patterns within additive dynamics when the risk of ruin is present.
Original languageEnglish
Article number20230544
Number of pages14
JournalProceedings of the Royal Society A: Mathematical, Physical & Engineering Sciences
Issue number2278
Publication statusPublished - 25 Oct 2023

Bibliographical note

Publisher Copyright:
© 2023 The Author(s).


  • ergodicity
  • human decision-making
  • Ruin


Dive into the research topics of 'Human decision-making in a non-ergodic additive environment'. Together they form a unique fingerprint.

Cite this