Embracing LLMs for Responsible Innovation in Learning Design

Generative AI and Large Language Models (LLMs) such as ChatGPT are among the most groundbreaking technologies to emerge in the past several years. At Socratic Arts , we recognize the potential of LLMs such as ChatGPT to revolutionize learning design and we're dedicated to pursuing innovation while upholding responsible practices.

Socratic Arts has a long history of pioneering AI research, beginning at universities including Northwestern, Yale, and the University of Texas at Austin. Today, we are excited to explore the many ways we can incorporate generative AI into our work, ranging from design acceleration to tool integration, AI-driven simulations, and the development of LLM-based Socratic tutors for learners.

Additionally, we also feel it’s imperative to take a strong ethical stance and protect the proprietary data and interests of our clients while also honoring and enhancing the uniquely human creative talents of instructional designers.

We want to harness the power of LLMs like ChatGPT as valuable tools without compromising the integrity of instructional content, and we set forth a few guiding principles we follow in our work.

LLMs as Collaborative Accelerators

LLMs such as ChatGPT are powerful tools that can enhance our productivity as instructional designers and content developers. While they can provide valuable insights and initial drafts, human expertise and perspective are essential in shaping the content to align with design goals and learner needs. We believe it’s valuable to adopt LLMs and generative AI tools as accelerators and collaborators, but they can’t replace human creativity, insight, design expertise, and judgment.

Here's how we embrace this technology:

• Prompt Engineering: We use basic and more advanced approaches to engineering prompts to increase ChatGPT’s power. At Socratic Arts, we are conducting internal research, leading workshops, and speaking at conferences to help further the field.
• Prompt Sharing: We've built, and will evolve, our own internal prompt library to codify our best practices and leverage expertise across the organization.
• Rapid Research: We use LLMs to educate ourselves on a given subject at lightning speed. Fact-checking is essential because all LLMs are prone to hallucination.
• Design Acceleration: We harness LLMs to accelerate brainstorming and ideation. Defining constraints and other design parameters helps us develop an array of targeted solutions.
• Content Collaboration: We engage LLMs and other generative AI tools as collaborators, not as independent creators. LLMs are great for quickly surveying the general point-of-view on a given topic or creating rough first outlines.
• Socratic Tutoring and Role Play: With the right prompts, ChatGPT 4 can act like a mini tutor or role play partner and help learners practice skills in very specific contexts. As secure LLMs become more available, we are excited about the possibilities this type of interaction may have for our clients.
• Tool Integration: We prioritize the integration of LLMs or other generative AI technologies into our production tool suite. At Socratic Arts, we are building generative AI into our own proprietary tools, and excited to partner with other companies providing useful tools as well.

Responsible Use of LLMs

While LLMS such as ChatGPT are a valuable resource, they have limitations. To ensure responsible innovation in learning design, we use the following guidelines:

1. Client Policy: Ask clients if they have an internal policy about LLM and generative AI use. Discuss how this impacts our work together.
2. Data Privacy: Never upload proprietary information to an open source LLM like ChatGPT 3.5 and 4 since they do not guarantee data privacy. Only upload proprietary client data to a secure LLM or other generative AI with written permission from client.
3. Hallucination: LLMs like ChatGPT may occasionally provide false information and references. Fact-check everything. Use SMEs to validate content for accuracy.
4. Human Revision: Treat LLM output as a first draft.
5. Copyright Caution: Be mindful of copyright issues. Though individual users are not likely to be liable, ChatGPT and other generative AI tools are often trained on large amounts of copyrighted material without permission. We recommend not asking ChatGPT to create output in the style of a copyrighted author or publication.
6. Future Regulations: Expect evolving legal frameworks around Generative AI and LLM use in the coming years.

If you have any concerns or questions regarding the use of LLMs or generative AI in your projects, please don't hesitate to reach out to us.

References and Additional Resources:

• OpenAI's Advice on Educational Use of ChatGPT
• Congressional Research Service Assessment on Copyright and LLMs

Please note that the legal landscape regarding AI and copyright is evolving, and we will be updating our stance, as need necessitates.

?