Social Skill Training with Large Language Models

People rely on social skills like conflict resolution to communicate effectively and to thrive in both work and personal life. However, practice environments for social skills are typically out of reach for most people. How can we make social skill training more available, accessible, and inviting? Drawing upon interdisciplinary research from communication and psychology, this perspective paper identifies social skill barriers to enter specialized fields. Then we present a solution that leverages large language models for social skill training via a generic framework. Our AI Partner, AI Mentor framework merges experiential learning with realistic practice and tailored feedback. This work ultimately calls for cross-disciplinary innovation to address the broader implications for workforce development and social equality.

Introduction. People need both general and domain-specific skills to succeed in home and work life (Dean, 2017). Specialized workers need not only technical competence, but also field-specific soft skills that extend broader social skill-sets. For example, mental health counselors use active listening (Nemec et al., 2017), a skill for building trust and empathy (DeVito, 2019; Ramsey and Sohi, 1997). Similarly, business managers employ conflict resolution (De Dreu et al., 2001) and strengthen team bonds (DeVito, 2019) with specialized strategies (Lipsky et al., 2003). Learning these skills may involve passive observation, trial-and-error, or explicit instruction, but ultimately, a learner will need deliberate practice (Giddens and Griffiths, 2006), as social skills are inherently interactive. Learning environments for social skills can be inaccessible, especially when training is offered by experts in formal programs, which are expensive, time-consuming, and limited in availability.

Discussion / Conclusion. This perspective paper examines a widespread challenge: mastering essential social skills for both personal and professional success. Opportunities to practice these skills in a safe learning environment are limited, especially for underprivileged groups. We show how LLMs can help create environments where everyone can practice social skills via our proposed AI Partner and AI Mentor framework. Here, the AI Partner offers a risk-free practice environment, while the AI Mentor provides knowledgeable tailored advice. Below we highlight a few take-aways that illustrate how this approach can reshape social skills learning moving forward. Firstly, utilizing LLMs on APAM requires addressing multiple technical challenges, such as enhancing the simulation of AI partners to exhibit a consistent, plausible and instructive personality, and building AI mentors to have context awareness, domain expertise and feedback efficiency. Secondly, deploying LLM based social skill training systems has the potential to amplify limitations such as hallucinations and biases, thus our APAM framework offers a roadmap for how to use LLMs for social skill training by breaking safe usage into a continuum dependent on current capabilities.