Writing Great Prompts for Child Welfare: Getting What You Need from Chatbots (Volume 3)
By now you’ve explored chatbots and probably used them successfully for some tasks. Now it’s time to supercharge your prompting skills. Effective prompt-writing not only reduces the number of prompts you need to write in order to get what you want from AI chatbots (such as Microsoft Copilot, ChatGPT, or Perplexity.ai ), but it also allows you discover new uses from AI that you may not have known about. Below we explore five practical steps will help you improve your prompts, which will save time and improve outcomes for children and families through the effective use of these tools.
1. Tell it what level of authority it should have
Remember that child welfare professionals have years of training that is reflected in how they create and present material. A well-trained chatbot can assume this level and type of knowledge quite well by drawing on the data used for its training. Your job is guide it toward this area of expertise so it knows how to give you the results you want.
2. Give it Context
Much of what we create has context and nuances that impact what we are trying to achieve. The output you are seeking from the chatbot might be impacted by a process that is unique to a particular state or it could be stemming from a court order that has important details that cannot be ignored. By providing this type of context, the chatbot will be able to pick up on details that you might not even realize if you were creating it from scratch.
3. Clearly Tell it what to produce
Are you creating a service plan? A court report? A training plan, a home study, or recruitment policy? Does it have bullet points, specific sections, or a particular format? By explaining exactly how you want the result to look, feel, and read, you will get a result that more closely aligns with what you want. This will save valuable time by reducing the amount of editing you need to do after you get the initial response from the chatbot.
4. State the Goal of the results
Who is your audience? Are you trying to convince or persuade someone of a particular intervention? Are you trying to create a resource that that needs to be learned quickly or that needs to use the language of a unique situation? Tell the chatbot what you want the outcome to be of the output you are creating with as much detail as you can. You’ll find that the chatbot “understands” a tremendous amount about how to create goal-driven content.
5. Tell it to ask you a specific number of follow-up questions.
This one may be surprising, but you will also be surprised at how much it helps to engage the chatbot in this type of conversation. Chatbots use models that rely on assumptions and it will fill these in (with its own probability-based assumptions) when they are not given explicitly. By asking the chatbot to clarify key points about your intentions you are exposing additional areas where the chatbot might make its own decisions instead of asking you. Those last few “context holes” could be the difference between several minutes or hours of editing. By narrowing the scope in this way you can also learn more about how these chatbots behave with certain types of material.
…
By giving attention to some of these key points, you will learn not only save yourself editing time, but you will get better and better at learning how these chatbots work. This will allow you to uncover more uses for these tools in the future as and you will see yourself develop into a more effective child welfare practitioner.
Reminder: Whenever you are using publicly available chatbots, never use private or proprietary information in your prompts such as PII, PHI, or any HIPAA-protected information.
Questions or comments about AI in child welfare? Please contact us at ai@fostercaretech.com.