# The revival of the Baldwin Effect

## TL;DR

This paper revisits the Baldwin effect, clarifies previous misconceptions, and demonstrates through simulations that learning can significantly increase the likelihood of behaviors becoming innate over evolutionary time.

## Contribution

It provides a corrected interpretation of prior work and empirically demonstrates the Baldwin effect's occurrence in computational models with quantitative estimates.

## Key findings

- Learning can increase fixation probability of behaviors by several orders of magnitude.
- The Baldwin effect can be observed in silico under certain parameter conditions.
- Previous claims were misinterpreted; the effect is distinct from just learning ability selection.

## Abstract

The idea that a genetically fixed behavior evolved from the once differential learning ability of individuals that performed the behavior is known as the Baldwin effect. A highly influential paper [Hinton G.E., Nowlan S.J., 1987. How learning can guide evolution. Complex Syst. 1, 495-502] claimed that this effect can be observed in silico, but here we argue that what was actually shown is that the learning ability is easily selected for. Then we demonstrate the Baldwin effect to happen in the in silico scenario by estimating the probability and waiting times for the learned behavior to become innate. Depending on parameter values, we find that learning can increase the chance of fixation of the learned behavior by several orders of magnitude compared with the non-learning situation.