Building Local, Open, and Carbon-Friendly AI Solutions for Social Work Research

Social work researchers increasingly confront vast collections of unstructured data containing valuable insights but requiring significant processing resources. While cloud-based large language models (LLMs) offer powerful tools for natural language tasks, they present serious ethical concerns regarding data privacy, security, ongoing costs, and environmental impact. This workshop introduces local LLMs as a compelling alternative that prioritizes security, privacy, and sustainability for social work research. Running directly on researchers' hardware, these models eliminate the need to transmit sensitive data to external servers, ensuring participant confidentiality while significantly reducing the carbon footprint associated with cloud computing.

We will demonstrate how local LLMs enhance research productivity through powerful natural language capabilities. These models excel at extracting key information from diverse text data sources, identifying meaningful patterns, and surfacing insights across various research contexts. Their summarization abilities transform lengthy documents into concise, actionable reports while preserving essential content. Local LLMs efficiently process open-ended survey responses, converting unstructured qualitative feedback into organized findings without exposing sensitive information to external systems. They analyze social media content to reveal community sentiment and emerging trends, offering researchers real-time insights into public discourse. The workshop will also introduce visual analysis capabilities, expanding research possibilities beyond text to include image-based data sources increasingly common in contemporary social work research.

The workshop delivers a practical implementation roadmap covering four essential components:

1. Hardware Selection: Clear guidelines for choosing appropriate computing resources based on your research needs and budget constraints, with options suitable for standard research environments rather than requiring specialized infrastructure.

2. Software Solutions: Overview of accessible platforms that streamline the installation and operation of local LLMs, eliminating technical barriers to adoption.

3. Hands-On Examples: Step-by-step demonstrations showing how to prepare text data for analysis and conduct various analytical procedures using local LLMs.

4. Inclusive Access Methods: We will feature no-code and code-based solutions, ensuring the workshop content is accessible regardless of technical background.

Participants will leave equipped with the knowledge and skills to implement local LLMs in their research workflows. This approach addresses ethical concerns around data privacy and promotes environmental responsibility by reducing computational energy consumption. By embracing these open-source, locally-deployed solutions, social work researchers can enhance the rigor and efficiency of their work while maintaining the highest standards of ethical practice and data stewardship.