Brain-Inspired Affective Empathy Computational Model and Its Application on Altruistic Rescue Task

Abstract

Affective empathy is an indispensable ability for humans and other species' harmonious social lives, motivating altruistic behavior, such as consolation and aid-giving. How to build an affective empathy computational model has attracted extensive attention in recent years. Most affective empathy models focus on the recognition and simulation of facial expressions or emotional speech of humans, namely Affective Computing. However, these studies lack the guidance of neural mechanisms of affective empathy. From a neuroscience perspective, affective empathy is formed gradually during the individual development process: experiencing own emotion-forming the corresponding Mirror Neuron System (MNS)-understanding the emotions of others through the mirror mechanism. Inspired by this neural mechanism, we constructed a brain-inspired affective empathy computational model, this model contains two submodels: (1) We designed an Artificial Pain Model inspired by the Free Energy Principle (FEP) to the simulate pain generation process in living organisms. (2) We build an affective empathy spiking neural network (AE-SNN) that simulates the mirror mechanism of MNS and has self-other differentiation ability. We apply the brain-inspired affective empathy computational model to the pain empathy and altruistic rescue task to achieve the rescue of companions by intelligent agents. To the best of our knowledge, our study is the first one to reproduce the emergence process of mirror neurons and anti-mirror neurons in the SNN field. Compared with traditional affective empathy computational models, our model is more biologically plausible, and it provides a new perspective for achieving artificial affective empathy, which has special potential for the social robots field in the future.

Keywords: Artificial Pain; affective empathy; altruistic behavior; mirror neuron system; self-awareness; spiking neural network.

Figure 1

The neural mechanism of affective empathy. The brain areas contain: Emotion Cortex, Motor Cortex, and Perception Cortex. All the blue arrows represent the process of experiencing own emotions, and all the orange arrows represent the process of empathizing with others. The blue dashed arrow represents the process of action execution and action re-afference, and the orange dashed arrow represents perceiving other's overt action. There are mirror neurons (MN) in the Motor Cortex.

Figure 2

The architecture of the brain-inspired affective empathy computational model. The two dotted boxes represent two submodels: The Artificial Pain Model and the affective empathy spiking neural network (AE-SNN). Each circle represents a neuron population. The orange arrows indicate excitatory connections, and the blue arrow indicates inhibitory connections.

Figure 3

The spike sequence of the three module neurons and the STDP training process between the Motor neuron and the Perception neuron. The firing of the Emotion neuron triggers the firing of the corresponding Motor neuron. As a result of action re-afference, the Perception neuron subsequently fires. There is a 200 ms delay from the firing of the Motor neuron to the firing of the Perception neuron. The blue curved arrows indicate the causality of the three modules, and the blue straight arrows represent the STDP training process.

Figure 4

The intrinsic motivation of altruistic behavior. (A) Self-pain relief. (B) Altruistic behaviors that relieve others' pain.

Figure 5

The setups of the grid world. The green pentagon represents the agent. The direction pointed by the tip of the pentagon is the direction of the agent's next action. The gray area is walls, the danger zone is on the (A), and the safety zone is on the (B). The yellow circle is the switch. The black circle in the danger zone is a dangerous object.

Figure 6

The random exploration process of agentA. The agentA explores the environment and uses the Artificial Pain Model to detect its own internal state, completing the training of the AE-SNN. (A,B) Show that agentA collides with the dangerous object during the exploration. (C,D) Show that agentA generates a motor damage and is detected by the Artificial Pain Model; agentA is in the pain state and turnsred. (E,F) Show that the switch being touched, establishing a pathway from the danger zone to the safety zone; agentA then enters the safety zone and returns to the normal state.

Figure 7

The value of Free Energy. The X-axis represents the number of steps and Y-axis represents the value of Free Energy. Point (A) indicates that the agentA collides with a dangerous object at 19 steps, and the damage generates; Point (B) indicates that the agent enters the safe zone at 82 steps and recovers to its normal state.

Figure 8

The spikes of the neurons in the Emotion Module (A), Motor Module (B), and Perception Module (C) in the training phase of AE-SNN. The X-axis represents the firing time, and each unit represents 100 ms. Y-axis represents the index of the neurons.

Figure 9

The synaptic weights from the Perception neurons to the Motor neurons. (A) The change of synaptic weights. The X-axis represents the training epochs and Y-axis represents the value of weights. (B) The trained value of synaptic weights. The X-axis represents the index of Perception neurons and Y-axis represents the index of Motor neurons.

Figure 10

The altruistic rescue process. The agentA through AE-SNN empathizes with agentB and carries out rescue behavior. (A) Shows that agentA with affective empathy ability is placed in the safe zone and agentB (green pentagon with black border) is placed in the danger zone for exploration. (B) Shows that agentB inevitably collides with a dangerous object, causing the pain state and turning red; agentA generates pain empathy through AE-SNN. (C) Shows that agentA learns to find the switch. (D) Shows that the switch is touched, agentB enters the safe zone and relieves the pain state.

Figure 11

The spikes of the neurons in the Perception Module (A), Motor Module (B), and Emotion Module (C) in the altruistic rescue task phase. The X-axis represents the firing time and each unit represents 100 ms. Y-axis represents the index of the neurons.