Skip to main content

Table 1 Prominent features used in the feedback prediction model. The features are ordered by their correlation with the feedback score. Interestingly, the age of the agent is negatively correlated with the the feedback, meaning that younger agents receive better feedback. Unsurprisingly, the average feedback of the agent in previous calls is correlated with the feedback at the current call. In addition, the average ratio between agent messages and caller messages in previous calls is correlated with feedback, meaning that agents who generally write many messages are likely to receive a better feedback

From: Machine-learning based routing of callers in an Israeli mental health hotline

Feature

Correlation

Agent age

− 0.18

Agent average feedback

0.176

Agent average message ratio

0.16

Agent’s number of chats during the shift (before current chat)

− 0.145