A Survey of Personality, Persona, and Profile in Conversational Agents and Chatbots

TL;DR

This survey addresses how to represent and embody personality in neural conversational agents by defining Personality, Persona, and Profile and organizing models by underlying schemes. It catalogs 21 datasets and surveys embodiment methods, emphasizing memory-augmented and Transformer-based architectures, with a focus on explicit descriptive sentences and implicit cues derived from dialogue. Key contributions include a taxonomy linking personality schemes to modeling approaches, a detailed dataset inventory, and a synthesis of trends that highlights the prominence of Persona-Chat and Image-Chat as foundational resources. The work underscores evaluation challenges and advocates for richer data, prompting-based strategies, and hybrid symbolic/neural systems to advance personality in CAs for practical, engaging user interactions.

Abstract

We present a review of personality in neural conversational agents (CAs), also called chatbots. First, we define Personality, Persona, and Profile. We explain all personality schemes which have been used in CAs, and list models under the scheme(s) which they use. Second we describe 21 datasets which have been developed in recent CA personality research. Third, we define the methods used to embody personality in a CA, and review recent models using them. Fourth, we survey some relevant reviews on CAs, personality, and related topics. Finally, we draw conclusions and identify some research challenges for this important emerging field.