** Before using AI for any legal questions or research, consult with a lawyer first to avoid waiving confidentiality or privilege protection when using AI. Be sure to read our blog post on protecting privilege with AI use in litigation here: https://www.gilbertsondavis.com/chatgpt-do-i-have-a-case-5-uk-and-us-cases-on-ai-and-privilege-in-litigation/ **
Here are 5 tips to keep in mind to improve the quality of the AI results when prompting AI for complex questions:
1. Keep it simple
- Give short, clear, specific, and direct instructions to AI to improve the chances AI will execute the task you intended
- Split complex issues and instructions into separate prompts to allow the AI to use more context window per instruction
- Iterate on the AI content by asking questions, correcting incorrect or misguided statements, challenging assumptions, and refining and expanding upon the result
- Describe the results you want so the AI has a clear objective
2. Manage the context
- Keep the information you give to AI focused on what is relevant to the immediate task
- Provide more background information to assist the AI to find a better match for your question or task
- Describe and define the scope of information the AI will reference
- Include a ‘skill’ prompt that has been used before and has been shown to work
3. Fact-check with a clear reference
- Always check for and correct any incorrect facts or assertions
- Ask for the source and where specifically the information came from with a page and paragraph reference for review
- Review the AI results with the source side-by-side to compare
- Check the AI’s assumptions – when you do not provide all of the information, the AI tends to make assumptions, and those assumptions may well be false, leading to flawed results
4. AI can assist but should not decide
- Use AI to assist – AI is better suited to help you search for and organize information since these tasks can be performed within the context provided
- Avoid letting the AI decide – using AI to evaluate or make decisions is risky because the correct result may require further information, facts, or reasoning beyond the scope of the information you provided to the AI or the AI’s capability
5. Audit the results
- Always review the AI content thoroughly
- Refrain from sending AI content to another person if it has not been vetted, as the recipient tends to figure that the information they receive is error-free
- Consult with a professional for advice and review
How Can We Assist you?
Our lawyers have expertise and experience in litigation and technology matters and can assist you in resolving your legal issues in a timely and cost-effective manner. Please contact us for an initial consultation through our Request Consultation Form, email to info@gilbertsondavis.com, or by telephone (416) 979-2020.
Brief informational summaries about insurance litigation, commercial litigation and family law litigation matters in the courts of Ontario and Canada are periodically published on our website. Please note that our website content is for informational purposes only, and should not be construed or relied upon to provide legal advice. If you require legal advice, please request an initial consultation with Gilbertson Davis using the Request Consultation Form on this webpage or by contacting our Intake Coordinator on (416) 979-2020, ext. 233 (both subject to the Terms of Use described on our Contact page).
Comments & Opinions by Gilbertson Davis lawyers and staff on its Blog, or in media interviews, appearances or publications, or in professional publications, are personal to them, and do not necessarily represent the opinions of the Firm or anyone at the Firm other than the individual expressing those comments or opinions.
About the Author
