The Writers Guild of America recently negotiated a contract that allows writers to take credit for writing they do based on ChatGPT and other AI language learning models, along with writing they revise if given writing produced by those AI tools. They own their writing with or without using the tools. I took this stance in writing the book Elements of Crime Patterns. I am the book's writer, even if AI greatly helped me. ChatGBT helped me immensely, but I had the concepts I wanted to present first and used AI as a tool to manifest them beyond what I could have done alone.
How did I use ChatGBT-4?
I gave these custom instructions to it:
ChatGBT asks: What would you like ChatGPT to know about you to provide better responses? I am writing a book on elements of crime patterns for law enforcement. The questions I ask are to help me write the book clearly and correctly. This is a reference book and needs to include as much detail as possible. ChatGBT asks: How would you like ChatGPT to respond? I want the book to make sense to police officers as well as researchers and criminal justice students. I do not want opinions. I need facts. I need detailed answers regarding the variables involved.
At first did not set up custom instructions and I was continually being warned about how I should not use the answers to commit crimes; similar dire statements were in each response. That grew very tiresome. These basic instructions to ChatGBT insured the subsequent quetioning started from that standpoint.
Once I got a better handle on it, AI gave a lot of meat to the bones I already had. I had picked all the categories I wanted in the book and added a few that came up as I used the tools.
(NOTE: Do NOT put your crime or investigative data into ChatGBT-4 or similar tools. Only use the AI tools available on your closed, secure systems - if you have them - and only with legitimate permissions.)
AI did not do much in the way of research for me - I knew what I wanted to write, but I needed to have the ability alone to come up with saying it the way I wanted to say it and retrieve it all. In a very real way, it corroborated my "theory" that we know a lot we do not articulate. Tacit knowledge is a genuine thing - it is what we know without words or speech. We can only hold a minimal number of conscious thoughts in our brains simultaneously, a tiny number compared to the thoughts we actively "know." Often, we do not know what we know because we have not consciously thought out the knowledge. We learn so quickly, and taking the time to acknowledge every bit of learning would waste time - our minds are designed to be efficient. We use mental shortcuts, called heuristics, to think quickly. AI helped me pinpoint what I wanted to say, slow down thinking, and say what we know in order to (hopefully) share a common understanding.
Before easily accessing AI tools, I had done much research (over-researched before my discovery that ChatGPT could help me), but much of the study could not verify the outcome I intended. ChatGPT gave me results I could prove with my genuine subject matter expertise since I knew what was true based on real-world observation and experience.
When AI did give me the information I did not recognize to be accurate, I researched it to verify it, and if I had doubts, I eliminated the results. It was like sifting, much continual querying to get what I wanted. Then reorganizing, formatting, revising... AI does not spit out an orderly book. Then, I put chapters into both Grammarly AI and MS Word Editor to check for plagiarism, which is very tedious. I was not very concerned with plagiarism because this book is primarily about facts rather than theories, and facts cannot be copyrighted, although the expression and presentation of facts can be copyrighted, as is my book.
In the second half of the book, I present 27 crime categories. I wanted to show crimes the way I might when I worked as a behavior specialist, where we used antecedents, behaviors, and consequences to show what actions were involved. It took many tries to get AI to finally spit its crimes out in a way that worked for me. As a law enforcement analyst, I had personally worked on 3/4 of the crimes in real life that are in the book, in both minor and significant ways, and taught courses on a few of the other crime categories. There are only 3 or 4 crimes that I had no direct experience with in my law enforcement analyst career but knew about and wanted to include.
I cited any research I referred to in the footnotes and read a lot more, but I could not find that much that supported the facts in the book. For example - concealment methods are written about in customs border protection cases, but there is no research on saying this is what they are and look for them that I could locate. I did not include my extensive bibliography in the book because I felt it falsely represented as if I used everything; I added the bibliography to this blog so that you may peruse it. It is on the right-hand side of the blog.
I used Scite.ai to help me with citations and Perplexity.ai to locate a few extra resources. I tried other AI resources that I did not find helpful enough to use.
I decided to self-publish after first submitting a proposal to a publisher (I took it back while waiting for the review process) for several reasons. I wanted the book to be affordable and to keep it more practitioner-based than academic. In August, I found an excellent tool called Atticus that allowed me to format the book for Amazon publishing, and it made that part of the task almost easy. I would not recommend doing what I did - every single thing by myself, including editing. (One reason I decided not to index it besides not thinking it necessary.) The good thing about how I did it is that I can easily correct things and upload new versions that can be printed as the book whenever I want to do so. I have a lot of control.
I could risk doing it all myself because I have nothing to lose in that I am retired and have no career plans. There is nothing to gain, either, because I want nothing for myself - I want to stay retired. The truth is, this book would not leave me alone, and now I have it out of my head, once I let enough people know about the book, I can go back to my idyllic retirement life—another reason why you should not copy me.
Using AI to write is a wonderful way to feel like you have superpowers. As a former member and short-term President of the Police Futurists International, I have an inclination to want to think about the future. It was a pleasure to use these tools of the future and show what can be done in the present via this book. I encourage everyone to experiment with it!
And I hope you find Elements of Crime Patterns useful.
Read:
Want to Save Your Job From A.I.? Hollywood Screenwriters Just Showed You How by Adam Seth Litwin, associate professor of industrial and labor relations at Cornell University, published in The New York Times on September 29, 2023.
No comments:
Post a Comment