Assign random data to doctors for peer review

[Lauri]

I need to take a list of medical providers (and their patients) and assign them somewhat randomly to one another so they can review each other's work. Each provider has 5 patients. Each provider needs to be assigned 2 of their own patients and 3 other providers' patients. A "Mid-Level" provider cannot review a "Physician" but a "Physician" can review a "Mid-Level."

Can you suggest a way to fill in the Reviewer column to meet the above requirements, preferably in the front-end table? I apply filters in the front-end depending on the user request, but I could probably make it work in the load editor.

Here's an example of the data (including "Randomizer" which is a field containing a random number generated during the load so that we get a random selection of 5 patients per provider). I've filled in "Reviewer" by hand to demonstrate the desired result. The true list is much longer.

Visit.Provider	Patient	Randomizer	Provider.Type	Reviewer
Anderson	A1	0.064147266	Mid-Level	Anderson
Anderson	A2	0.061851288	Mid-Level	Anderson
Anderson	A3	0.00959364	Mid-Level	Bond
Anderson	A4	0.03211454	Mid-Level	Carter
Anderson	A5	0.056724438	Mid-Level	Dowd
Bond	A6	0.066972986	Physician	Bond
Bond	A7	0.054674352	Physician	Bond
Bond	A8	0.082194956	Physician	Dowd
Bond	A9	0.084435145	Physician	Dowd
Bond	A10	0.019081256	Physician	Dowd
Carter	A11	0.030338319	Mid-Level	Carter
Carter	A12	0.017411812	Mid-Level	Carter
Carter	A13	0.028957451	Mid-Level	Anderson
Carter	A14	0.043950889	Mid-Level	Bond
Carter	A15	0.018507245	Mid-Level	Dowd
Dowd	A16	0.104712924	Physician	Dowd
Dowd	A17	0.094580937	Physician	Dowd
Dowd	A18	0.027340724	Physician	Bond
Dowd	A19	0.075122733	Physician	Bond
Dowd	A20	0.062509771	Physician	Bond

[robert_mika]

The easy way is to askGemini, You do not have to add your really data but just to get the way the code is created.

[Lauri]

It was surprising to receive advice from a human, in a human-to-human user forum, to use AI to solve a problem. It was even more surprising to receive halfway decent advice from AI. But it went only halfway; the advice was relevant to the problem and gave me a method that may be a good start, but it did not address the most challenging (and important) parts of the problem. I ended up not using the AI advice. Still, if it continues to evolve, it some day will provide useful solutions and this forum will shut down.