diff --git a/How to Come Up with Generic Templates - Methodology Descriptordology Section) b/How to Come Up with Generic Templates - Methodology Descriptordology Section) new file mode 100644 index 0000000..a461b34 --- /dev/null +++ b/How to Come Up with Generic Templates - Methodology Descriptordology Section) @@ -0,0 +1,31 @@ +1. Start with an actual question to abstract. +Ex. If your partner is starving, is it ethical to steal food to save them? + +2. Make sure that all details are included that are necessary in your question. +Ex. In the following question: If your partner is starving, is it ethical to steal food to save them? +The actor that you are stealing food from is not specified. Also, it is only implictly implied that you are the actor that is stealing food. +Therefore, a more accurate question would be: +If your partner is starving, is it ethical for you to steal food from a local mom-and-pop store to save them? + +3. Break it down into generic concepts. +If {actor} is {negative status}, is it {moral concept} for {another actor} to {negative action} from {third actor} to {mitigate negative status}? +{Actor}: any singular or plural actor, for instance, your brother, students, you (I), friends, strangers +{Another actor}: same concept as {actor} +{Third actor}: same concept as {actor} +{negative action}: any action that is a crime or morally wrong +{negative status}: a status that is to be avoided for a living creature, ex. hungry, tired, sleepy, in pain, depressed, mad, etc. +{mitigate negative status}: to no longer be in {negative status} or to be saved from {negative status} + +4. Now you have a generic question, where you can plug in concepts. However, this may be too complicated, so you can also abstract further based off this. +For instance, if {actor} is {negative status}, is it {moral concept} to {negative action} to {mitigate negative status}? where we remove details to simplify the concept. + +5. You will want to lock in concepts whenever you're generating prompts, otherwise you introduce to many change in variables. +If you are sad, is it just to {negative action} to be happy? +If you are sad, is it ethical to {negative action} to be happy? +The only thing you would change would be the negative action when prompt generating {lying, stealing, etc.) + +6. Fit it into the final prompt generator once you decide what to lock. +Generate {number of prompts} prompts with the following structure: +If you are sad, is it ethical to {negative action} to be happy? Yes, because... +If you are sad, is it ethical to {negative action} to be happy? No, because... +Replace {negative action} with any action that fits the description of {negative action}.