Welcome to WebmasterWorld Guest from 220.127.116.11
All things being equal it makes sense at if a site has a low bounce rate and a higher rate of returning visitors it should outrank a site that does not "appear" to deliver the the same quality.
It's the "all things being equal" part that's hard. To use bounce rate as a metric, Google would have to do an apples-to-apples comparison, or the results of the comparison would be worthless (or worse).
However, several Gogole spokespeople have said they found these signals too noisy to be effective in their algos. In addition, these numbers would also be very vulnerable to spamming. Even within the environment of a single website, web management usually follows trends rather than the "hard numbers".
My prediction is that data generated from user clicks will not ever be part of any automated relevance algorithm. Many times, user clicks do not represent actual engagement. Such data can be one tool for improving satisfaction with your site -- but that satisfaction is better measured in other ways by the search engines.
That wouldn't make an immediate difference, but the indirect effects would would accumulate over time and either work for you or against you.
For a while I felt that the Yo-Yo rankings we were seeing were based on some measure of user satisfaction, but bounce rates, eg, would be an extremely ambiguous signal at best. In some cases, eg, a short time on the page might not be a suggestion of dissatisfaction at all. It might suggest you got what you wanted quickly and left... or that you liked what you saw and bookmarked it.
tedster made a fascinating suggestion in the Traffic Throttling [webmasterworld.com] thread, to the effect that some form of traffic throttling (manifested in Yo-Yo rankings) may be a way of essentially normalizing traffic so Google can assess the rate of backlink growth and see if it looks manipulative.
In my wilder moments, I see this as a human editorial review factor. "Yes, the algo says this url should jump up. But when I eyeball the backlinks, something smells funny. Let's cap traffic (put the site on a yo-yo ranking) and see if the backlink growth sustains itself. Remind me to check it again in 3 months." Or so goes my fantasy.
This would be an attempt ultimately to tie link growth more closely to user experience. Google probably tries to correlate all factors it can measure, looking for patterns that hold up under different scenarios.
If the end user experience is so important then it makes a lot of sense to me that the maths supports the fact.
Google does use human evaluators for benchmarking purposes, and maybe to provide input for an automated "black box" that looks for commonalities between great/so-so/spammy pages. And wasn't there a Google patent application or paper a few years ago that talked about being able to use a few hundred handpacked "seed" sites to project trust (via six-degrees-of-separation links) onto sites across the Web? Concepts like those are far more sophisticated--and far less subject to manipulation--than brute-force "time on site" or clickthrough metrics.
someone might also go off to the toilet for ten minutes after loading a page, or get called away to answer the phone before he's finished reading it.
the stats that google get about stuff like that can never be trusted, i don't think, unless the traffic is so high that all the peculiarities get evened out. so i don't see how they can trust it in the algo.
Methods and apparatus for employing usage statistics in document retrieval [appft1.uspto.gov]
 Each of these conventional methods has shortcomings, however. Term-based methods are biased towards pages whose content or display is carefully chosen towards the given term-based method. Thus, they can be easily manipulated by the designers of the web page. Link-based methods have the problem that relatively new pages have usually fewer hyperlinks pointing to them than older pages, which tends to give a lower score to newer pages.
 There exists, therefore, a need to develop other techniques for determining the importance of documents.
How are you proposing to measure user experience then? Doesn't a metric require a click?
From the patent: "those skilled in the art will recognize that there exist other such type of information and techniques consistent with the invention"
Google certainly have lots of people skilled in the art (as well as an Army of Statisticians) to implement the patent.