Clinical Social Workers' Perceptions of Large Language Models in Practice: Resistance to Automation and Prospects for...

Area of Science:

• Clinical Social Work
• Artificial Intelligence in Healthcare
• Human-Computer Interaction

Background:

• Generative artificial intelligence (AI), particularly large language models (LLMs), is increasingly accessible.
• Understanding the integration of AI into clinical practice is crucial for social work.
• Clinical social workers' perspectives on AI tools are essential for responsible adoption.

Purpose of the Study:

• To explore clinical social workers' perceptions of generative AI (AI) and large language models (LLMs) in clinical practice.
• To identify factors influencing the perceived usefulness of LLMs in social work.
• To understand social workers' experiences and concerns regarding AI in client interactions.

Main Methods:

• Qualitative reflexive thematic analysis of interviews with 21 clinical social workers.
• Exploration of participants' experiences with LLMs, including ChatGPT, in a case consultation exercise.
• Analysis of perceptions following a client demonstration using ChatGPT.

Main Results:

• Two key themes emerged: factors enhancing perceived usefulness (administrative support, client engagement) and factors diminishing it (confidentiality, loss of nuance, empathy limitations).
• Practitioners use LLMs as idea generators but express concerns about information quality and the need for a human-centered approach.
• Adoption decisions are influenced by professional ethics and relational values, favoring augmentation over full automation.

Conclusions:

• Responsible AI implementation requires practitioner training and clear ethical guidelines for LLM integration.
• Ongoing evaluation is necessary to ensure AI tools enhance clinical practice without compromising the therapeutic relationship or social work values.
• A human-centered approach is vital when integrating AI into social work, prioritizing client well-being and therapeutic depth.