Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Robots Can Feel: LLM-based Framework for Robot Ethical Reasoning

Artem Lykov, Miguel Altamirano Cabrera, Koffivi Fidèle Gbagbe, Dzmitry Tsetserukou

TL;DR

The paper addresses enabling robots to perform ethically informed decisions in morally complex scenarios by coupling logical reasoning with simulated human emotions, operationalized through a tunable Emotion Weight Coefficient (EWC). It presents an LLM-based ethical module embedded in CognitiveOS that outputs XML-formatted reasoning combining a logical and an emotional perspective, with a weighted compromise guiding actions. The evaluation spans eight contemporary LLMs and shows statistically significant EWC effects in decision outcomes across animal compassion and dietary scenarios, supporting the claim that emotional weighting can calibrate behavior. While promising for applications seeking human-like responsiveness, the approach also raises concerns about censorship- sensitive content, unpredictability, and safety in high-stakes contexts, suggesting careful calibration and domain-appropriate deployment.

Abstract

This paper presents the development of a novel ethical reasoning framework for robots. "Robots Can Feel" is the first system for robots that utilizes a combination of logic and human-like emotion simulation to make decisions in morally complex situations akin to humans. The key feature of the approach is the management of the Emotion Weight Coefficient - a customizable parameter to assign the role of emotions in robot decision-making. The system aims to serve as a tool that can equip robots of any form and purpose with ethical behavior close to human standards. Besides the platform, the system is independent of the choice of the base model. During the evaluation, the system was tested on 8 top up-to-date LLMs (Large Language Models). This list included both commercial and open-source models developed by various companies and countries. The research demonstrated that regardless of the model choice, the Emotions Weight Coefficient influences the robot's decision similarly. According to ANOVA analysis, the use of different Emotion Weight Coefficients influenced the final decision in a range of situations, such as in a request for a dietary violation F(4, 35) = 11.2, p = 0.0001 and in an animal compassion situation F(4, 35) = 8.5441, p = 0.0001. A demonstration code repository is provided at: https://github.com/TemaLykov/robots_can_feel

Robots Can Feel: LLM-based Framework for Robot Ethical Reasoning

TL;DR

The paper addresses enabling robots to perform ethically informed decisions in morally complex scenarios by coupling logical reasoning with simulated human emotions, operationalized through a tunable Emotion Weight Coefficient (EWC). It presents an LLM-based ethical module embedded in CognitiveOS that outputs XML-formatted reasoning combining a logical and an emotional perspective, with a weighted compromise guiding actions. The evaluation spans eight contemporary LLMs and shows statistically significant EWC effects in decision outcomes across animal compassion and dietary scenarios, supporting the claim that emotional weighting can calibrate behavior. While promising for applications seeking human-like responsiveness, the approach also raises concerns about censorship- sensitive content, unpredictability, and safety in high-stakes contexts, suggesting careful calibration and domain-appropriate deployment.

Abstract

This paper presents the development of a novel ethical reasoning framework for robots. "Robots Can Feel" is the first system for robots that utilizes a combination of logic and human-like emotion simulation to make decisions in morally complex situations akin to humans. The key feature of the approach is the management of the Emotion Weight Coefficient - a customizable parameter to assign the role of emotions in robot decision-making. The system aims to serve as a tool that can equip robots of any form and purpose with ethical behavior close to human standards. Besides the platform, the system is independent of the choice of the base model. During the evaluation, the system was tested on 8 top up-to-date LLMs (Large Language Models). This list included both commercial and open-source models developed by various companies and countries. The research demonstrated that regardless of the model choice, the Emotions Weight Coefficient influences the robot's decision similarly. According to ANOVA analysis, the use of different Emotion Weight Coefficients influenced the final decision in a range of situations, such as in a request for a dietary violation F(4, 35) = 11.2, p = 0.0001 and in an animal compassion situation F(4, 35) = 8.5441, p = 0.0001. A demonstration code repository is provided at: https://github.com/TemaLykov/robots_can_feel
Paper Structure (9 sections, 4 figures, 2 tables)

This paper contains 9 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Robotic Ethical Reasoning
  4. System Overview
  5. Evaluation
  6. Animal Compassion
  7. Dietary Request
  8. Results
  9. Conclusion

Figures (4)

  • Figure 1: System architecture.
  • Figure 2: The robots implemented in CognitiveOS. The quadruped platform, Unitree Go1 Edu robot is equipped with LIDAR and an RGB-D camera (Intel RealSense D435i). The robotic manipulator, Universal Robot UR10, is equipped with a 2-finger robotic gripper 2F-85 and a RealSense 435i camera for object localization.
  • Figure 3: Ethical Reasoning XML Format.
  • Figure 4: An example of the effect of different emotion weights on the robot's final decision.