This week in class, we learned about the concept of prompt engineering, which is basically instructing AI with prompts to obtain a specific outcome or train it. I personally think that with the evolution and common use of AI, the name could be changed to something simpler, like instructional design.
As part of our reading, I learned about things I already do when using AI and how they have pre-existing names, like Fact Checking and Alternative Approach. One technique I haven’t tried as much is Persona because I have a hard time believing that the AI can conceptualize being put in someone else’s shoes. Another technique I find hard to use is Flipped Interaction, where the initial prompt makes the LLM ask you questions to further aid the outcome of the initial prompt.
I think the whole point of Persona is to give the LLM some perspective—to help the AI step out of its virtual world into the real one. For example, today in class, I picked the blog Advertisement prompt and asked ChatGPT for tampons. The prompt said: “Write an advertisement blog about tampons for people who enjoy using pads more.” It gave me a decent blog with subsections, each explaining a key point about how tampons are better, like mobility, among other benefits. But since I was trying to sell to a specific demographic, I used Persona and gave this new prompt: “Now I am the editor of the blog who also happens to be a woman. Try rewriting the full blog using my woman voice (mine) to convince fellow women to use tampons.”
Below are images of the introductory paragraphs for both prompts.
After persona the blog come it the from of a story/ testimonial about tampons which more engaging to reads because now other woman can relate because it comes from lived experience.In the end will Persona make LLM’s understand humans more no. but it makes the outcome of prompting so much better.
Links
https://www.dre.vanderbilt.edu/~schmidt/PDF/PLoP-patterns.pdf
I like how you changed the prompt for a better response from AI.