Skip to main content
. 2024 Sep 11;8(1):159–177. doi: 10.5334/cpsy.117

Figure 6.

Illustration of inability to perform inverse RL in high softmax temp environments

DoM(0) belief update in high SoftMax temperature environment. We depict the updated DoM(0) beliefs against senders with different DoM level (row indicate sender’s DoM level, column indicate sender’s type) averaged across 20 different simulations. Due to the noisy behaviour of the senders, the DoM(0) finds it hard to identify the sender’s correct beliefs from its actions, even when it interacts with an adaptively matched sender (DoM(–1), top row). When interacting with the higher DoM sender, the receiver is still deceived, but with a lower certainty.