Table 1 Reinforcement learning characteristics and their compatibility with valence inversion.

From: Sensory substitution reveals a manipulation bias

Reinforcement learning model characteristic

Additional info

Can it explain valence inversion?

Changes in baseline reward

Caused by deafening

No

Exploration bonus

For informative actions

No

 

For diverse actions

No

 

For simple actions

No

Manipulation bonus

For impact

Yes

Back to article page