ISO is best. There's no debate there. From a data science perspective, YYYY/MM/DD is the only reasonable choice.
But most of the time you're using dates, you're only concerned with the month and day. That's the very reason we don't use ISO in our daily lives. If you started every mention of a date with the year, people would think you're a crazy person, or a time traveler, or perhaps a recently-awakened coma patient. There's just no need to begin with the year. Next Wednesday, 2024 December 18.
If you exclude the year, then the choice is month/day or day/month. Between the two, month/day is far more useful for the same reasons ISO is best. If I need both the month and the day, then I want the month first. The only time I would want the day first is if the month doesn't matter, and I can omit the month in that case. Giving me the day first and then the month forces me to wait for the month and then remember the day. It's inefficient transfer of information. If you exclude the year, MM/DD is objectively, if only marginally, better than DD/MM.
But then why would anyone use MM/DD/(YY)YY? Because we're already using MM/DD.