AdminGPT

You know that it had to be done, right?

With all the excitement, coverage, and concern about AI these days, I was simply curious if ChatGPT could do a better job at admin than real admin types could. (And by “better”, I mean better for us as the end users. After all, the real point of admin is to take work off our hands, not shove their work into ours.)

Trivial? Well, maybe … (But then Google also reveals that I’m not the first to come up with the term “AdminGPT”.)

I know that a lot more people are a lot more concerned about how, for example, students will use AI to cheat on their essays and the like. But, let’s be realistic about this. First, and most importantly, this kind of cheating will seriously reduce the number of excruciatingly awful term papers I have to suffer through each year. And, if you think about it, essays and term papers are the ultimate in open-book exams anyway. What’s to prevent the students from already cheating on them? You can buy these things online or have someone else write it for you for a fee. Hell, the whole expectation on our side is for some form of cheating insofar as the students should be recycling a bunch of other people’s ideas in the process. Is getting AI to do the research for this really that much different than asking Google or Wikipedia (or, from my golden, analog years, an encyclopaedia) for help? Quicker, faster, more efficient, and less effort, no question. But fundamentally different? Dunno …

So, with ChatGPT already showing to me that it could do a damn good job writing a sarcastic blog entry about admin (and good enough that I was duly impressed / unduly jealous as well as that my wife was fooled by it), I wanted to see how good it was at actually doing admin instead of just taking potshots at it. In particular how good is it at creating forms, an abomination against the laws of nature that I might have already griped about numerous times elsewhere in this blog?

The task? Create a work-safety assessment form asking about the specific hazards and specific precautions against corona in a university workplace setting.

The result? Et voilà …

1. Have you received training on how to prevent the spread of COVID-19 in your workplace?
   • Yes
   • No
2. If yes, please check any of the following topics you were trained on:
   • Proper hand hygiene
   • Proper cough and sneeze etiquette
   • Proper use of masks or face coverings
   • Physical distancing measures
   • Identifying symptoms of COVID-19
   • Reporting procedures for suspected or confirmed cases
   • Other (please specify):
3. Have you been provided with the necessary personal protective equipment (PPE) to protect yourself from COVID-19 exposure in your workplace?
   • Yes
   • No
4. If yes, please specify the PPE that you require:
5. Have you received training on how to properly use and maintain your PPE for COVID-19 safety?
   • Yes
   • No
6. Have you been informed of the policies and procedures for reporting COVID-19 symptoms or exposure in your workplace?
   • Yes
   • No
7. If yes, please describe the policies and procedures:
8. Have you been informed about the potential risks of COVID-19 in your workplace?
   • Yes
   • No
9. If yes, please check any of the following that apply:
   • Close contact with individuals who may be infected
   • Exposure to contaminated surfaces or objects
   • Inadequate ventilation or air circulation
   • Poor physical distancing measures
   • Inadequate cleaning and disinfection practices
   • Other (please specify):
10. Are there any areas in your workplace where physical distancing measures are difficult to maintain?
    • Yes
    • No
11. If yes, please specify the areas:
12. Are there any areas in your workplace where mask or face covering usage is difficult to maintain?
    • Yes
    • No
13. If yes, please specify the areas:
14. Have you observed any unsafe practices or conditions related to COVID-19 safety in your workplace?
    • Yes
    • No
15. If yes, please describe the unsafe practices or conditions:
16. Have you reported any unsafe practices or conditions related to COVID-19 safety to your supervisor?
    • Yes
    • No
17. If yes, were the concerns addressed promptly?
    • Yes
    • No
    • N/A
18. Have you been provided with the necessary resources to maintain proper hygiene and cleanliness in your workspace?
    • Yes
    • No
19. If yes, please specify the resources:
20. Have you been provided with any accommodations or alternative work arrangements to reduce your risk of COVID-19 exposure in your workplace (e.g. working from home, adjusted work schedules)?
    • Yes
    • No
21. If yes, please describe the accommodations or alternative work arrangements:
    Have you received any communication regarding COVID-19 safety measures and updates from your employer?
    • Yes
    • No
22. If yes, please specify the type of communication and frequency:
    Is there anything else you would like to share about COVID-19 safety measures in your workplace?

Thank you for taking the time to complete this form. Your responses will be used to evaluate COVID-19 hazards in the workplace and ensure that appropriate measures are taken to ensure the safety and well-being of all employees.

Now that’s pretty darn good, if not one heckuva lot more polite than the forms I’m used to filling out from the University of Not-Bielefeld. Getting thanked to fill out a form?! I always feel like the forms I’m filling out are more threatening than anything. And not just threatening in the direct sense of it simply being a form, but also with the more than implicit hint of “do this or else”. Granted, and despite some specific topics arguably being missing, the ChatGPT form is still a little longish, but we are talking admin here.

However, by comparison, the first version of the equivalent form from the University of Not-Bielefeld was nine pages long, complete with exclamation marks and surprise reverse checkboxes where suddenly “yes”, the default and correct answer for the last half page, meant “wrong answer”.

The real beauty of the ChatGPT form though and the reason why the University of Not-Bielefeld form was nine pages long comes down to a difference in perspective about who should be doing most of the admin work. As you can see, the ChatGPT form uses a lot of free-text fields, which are highly flexible but make it more difficult to pigeonhole the answers. In other words, someone has to read these answers and respond accordingly. By contrast, the university form (as well as its evil brethren) tended to be more rigid through being more specific. More pre-specified scenarios and so less post-specified thinking.

I can understand this, but only to a point. After all, the job description of admin has to include admin, shouldn’t it? It can’t really be a big surprise to anyone thinking about admin as a career that some paperwork might be involved sometimes. (The surprise instead to me is that anyone seriously thinks about it as a career choice.) My point is simply that they should be letting us do our work by not doing theirs …