Yes but their conclusion is "well our hypothesis is probably still right, we just didn't look hard enough" rather than "maybe it doesn't have a significant effect".

Well yes, that's just decent Bayesian rigor.

Previous studies have set a prior of x% confidence, you see evidence to the contrary, you update to some x-y%.

Given the volume of research on sleep, it probably takes more evidence, even if it's your own study to throw that out.

Are there previous studies that have shown smartphone use impairs cognitive ability?

You can look at that statement in a couple of different ways. Yes, experimental bias is one of them. The other is saying: we didn't see this effect, so we need to do longer term studies to see if it does exist.

Regardless, that statement is a good thing. It acknowledges a social bias towards the effect of smartphones. It doesn't give room for people to imply a result based upon that bias. On the surface, at least, it doesn't indicate that data was fudged to reach a particular result.

"Might mean" is a far cry from "our hypothesis is probably still right." This type of speculation is commonplace -- even expected -- in the discussion section of an article. Not much different than the lame duck "further research is needed to..."

Why should the null hypothesis be preferred?