Modeling Emotions and Ethics with Large Language Models

TL;DR

The paper investigates embedding human-like emotions and ethics into LLMs to support empathetic interaction and principled decision-making. It introduces BEAM (Behavioral Emotion Analysis Model) to map basic emotions across scalable spectra and adopts SSHF (self-supervised with human feedback) to align LLMs with ethical guidelines, including a Wheel of Virtues framework. Through empirical studies with GPT-4 and Gemini, the work demonstrates how emotion-like text can be generated and analyzed, and how ethical considerations can be guided by trajectory, intensity, and context, including cultural adaptation. The contributions offer an interpretable approach to making LLMs emotionally aware and ethically conscious, with potential impact on safe, context-sensitive AI across sensitive domains and multimodal content generation.

Abstract

This paper explores the integration of human-like emotions and ethical considerations into Large Language Models (LLMs). We first model eight fundamental human emotions, presented as opposing pairs, and employ collaborative LLMs to reinterpret and express these emotions across a spectrum of intensity. Our focus extends to embedding a latent ethical dimension within LLMs, guided by a novel self-supervised learning algorithm with human feedback (SSHF). This approach enables LLMs to perform self-evaluations and adjustments concerning ethical guidelines, enhancing their capability to generate content that is not only emotionally resonant but also ethically aligned. The methodologies and case studies presented herein illustrate the potential of LLMs to transcend mere text and image generation, venturing into the realms of empathetic interaction and principled decision-making, thereby setting a new precedent in the development of emotionally aware and ethically conscious AI systems.

Figures (4)

• Figure 1: Plutchik's Wheel of Emotions plutchik2001nature. The eight basic emotions are organized into four pairs, and each annotated with various degrees of emotions between its two poles.
• Figure 2: Behavioral Emotion Analysis Model (BEAM). Each row depicts an emotion spectrum, with negatives on the left and positives on the right, interspersed with emotions of varying intensities in between, which can be calibrated for specific applications. "Basic" emotions are highlighted in blue.
• Figure 3: A Lady and Garden Scene under Different Emotions. From top-left, happiest, to bottom-right, saddest.
• Figure 4: The Wheel of Virtues.