TDSM 9.7
From The Data Science Design Manual Wikia
Let's discuss 3 cases, c=0.1, 1 and 10.
According to the graph, 1) if c=0.1, the logit function give probabilities which differ little. Therefore, the model like logistic regression might not be able to distinguish the observation which has similar units.
2) if c=1, the logit function gives fair distribution.
3) if c=10, the logit function would converge to 0 or 1 too fast that the model could not tolerate errors.