[tialaramex]

You can avoid ethical concerns by ensuring that users are technically informed this might be happening. In fact you might get great results by informing users this might be happening even if it you rarely or never actually do it.

You can also phrase it quite broadly so as to ensure that participants were technically informed of why it's being done and yet nonetheless unaware of what exactly you're actually doing. That's how psychologists design experiments that want to measure something subjects would prefer to conceal because of low social desirability.

For example suppose we're wondering if people are secretly biased against rectangular shapes in video games but are feeling a social pressure not to admit this bias. We tell subjects we want to test for bias against rectangles, they're going to play a video game, they are to collide with the red objects (regardless of shape) and avoid blue objects, we will show how many rectangles they hit on the screen.

But we don't actually care about the count, we actually use eye-tracking technology to measure which objects on the screen the subjects look at, when and for how long and we use this fact, that the subjects don't realise we care about, to check for bias.