I've run a lab experiment (article in MagPi, the Rapsberry Pi magazine) in the past that had some rather neat rotation in the reaction vessel. We did this with angled hose fittings, one inlet and one outlet, and a small pump. The angle is important to get good rotary flow, i.e. the hose fittings should be as close as possible to tangential to the round tank. We used a miniature water pump run at the lowest voltage that it worked at reliably (5V on a nominally 12V pump), and the pump introduced no bubbles. I tried some cheap pumps from eBay. They were fine to start with but not robust enough.
This shows what happened when we added dye (methylene blue) to the system. The water is flowing in at the top of the image, out at the bottom. The flow isn't much, but it gets a good rotary motion going.

The beaker is about 7cm/3" in diameter here. Note the bubbles come from squirting in the dye with a syringe from out of shot.
To upgrade this for your use I have a few ideas, which may be comined, and coudl easily be controlled with some simple electronics of a Raspberry Pi or Arduino:
- Run the pump intermittently at random
- Run it off different voltages to get fast/slow movement. In our case we had to run really hard to get bubbles, unless there was a slight leak admitting air.
- Use a reversible pump, and reverse it (ours wasn't reversible, some of the cheap ones are - I've just looked on AliExpress). Reversing the flow wont give a mirror image of the movement on the surface.
- Have one outlet, multiple inlets, multiple pumps (or valves), with T pieces as required for the hose. In this case you'd probably want one tangential inlet, and one radial, so you could, for example, start it swimming round in circles, switch inlets, then it would dart across to the other side. A tangential inlet in the opposite direction could also be good.
- An inlet in the base might give some interesting effects if run briefly (even better if you can detect when the duck is over the inlet and turn it on, which could be done with a camera and OpenCV)
If you can't fabricate the connections like this, you can try sealing the hose straight through the wall and finishing with a 90° elbow to get the tangential motion. You can even run the hose in from above with an elbow. Depending on any decoration you have around the pond, it might be possible to conceal this.
You'll want to experiment with the water depth, so put the inlets high enough that the duck will float if they're at the surface, but low enough that you can raise the water above them by a few cm.
If your pond is square, you'll get a less circular flow, and there's a risk it will get stuck in the corners. But you can get a tangential flow more easily as that's square to the walls. Using 3 T pieces to inject flow at all 4 corners should ensure it keeps
going in the same direction, probably sucking the water from the middle of the base if you can.
I think your 6:1 pond:duck ratio is the minimum worth going for; 10:1 would be better I think. You might also get good effects with multiple ducks.
Here are a couple of plumbing ideas, to combine, modify etc..

Pumps marked * would be good to have reversible, with appropriate drive electronics. Arrows indicate the direction of flow at in-/outlets that I think should be better if you can't reverse the pumps