Yay, root causes!
But you make the same error, I think. The non-mentoring are not significantly more informed about the process than the non-mentored. I mean, sure, they know more (in principle) about the system, but that doesn't mean they know more (or anything) about the process of introducing new people to the system. Knowing how != knowing how to teach how.
So you can ask them "why don't you mentor?" or "why don't you mentor effectively?" but their answers may not be enlightening.
My approach here would be to back up one step and address the actual problem rather than the proposed solution.
The actual problem, presumably, is inefficiency on the part of new employees. (That's a sweeping term, meant to cover employees who quit in frustration because they don't know what's going on, employees who are fired because they don't know what's going on, and employees who manage to stick around long enough to figure out what's going on but could have gotten more done with less effort if that time had been shortened.)
This discussion sorta assumes that the proper solution is assigning them mentors as individuals, and then seeks to overcome the obstacles to making that work. That might be true but has yet to be established.
One obvious alternative that comes to mind, for example, is zone coverage. That is, assign individuals the job of "answering questions about blah" and make that part of their job description. The person you talk to about how to finesse a grant application isn't the person you talk to about how to deal with your spouse being irate that you're never home. Post the list of resources and topics searchable by topic centrally, keep it up to date, make maintaining it (including dealing with "who do I talk to about X?" requests for nonlisted Xes and putting them up there) somebody's job.
There are other possibilities as well (though I have to say, having written the above, that I rather like it).
You're absolutely right. Although, to a non-trivial degree, "not enough mentoring" is a problem in its own right, for reasons of social and managerial pressure, whether or not it actually would solve the problems its intended to solve. So I'm kind of taking it as a given in my response to the survey.
Hopefully, if not knowing how to teach is a barrier, one of the answers to "why don't you mentor?" would be "I don't know how / I'm not comfortable doing it", but I might be giving senior scientists too much credit for self-awareness...
I like the zone coverage idea. If we get moving in a direction of asking deeper questions, I'll bring it up. The really hard part about that plan is "make it someone's job to create and maintain the information base". That requires both money and the social leverage to extract the information from the local experts.
Although, that still puts the onus on the person mentored to ask the questions, which is a problem. I think what we really need is a kind of mentor-yenta, who would occasionally drop in, watch what and how you were doing, ask some questions, and then, where necessary, drag the student off to a mentor, and say "you, ask him about X, Y, and Z; you, tell him about A, B, and C. Page me if you hit any roadblocks". The only hard part about that is finding someone who's an expert in evaluating competency and can keep track of a thousand employees and what they're good and bad at...
that still puts the onus on the person mentored to ask the questions, which is a problem.
True. But it turns a query-and-response interface into a menu-driven one, which is usually an advantage in terms of ease-of-introduction.
Mentor-yenta sounds like a fine plan. In corporate environments this is defacto your manager's job, though of course some are good at it and others aren't.
The only hard part about that is finding someone who's an expert in evaluating competency and can keep track of a thousand employees and what they're good and bad at...
They don't need to be experts at evaluating competency... in fact, that's the wrong approach. That evaluation expertise is distributed throughout your organization anyway, no need to try and replicate it centrally, which is good, because you'll fail if you try. What you need is someone who networks well, can talk to people to find out who everyone who already knows what's going on thinks the go-to person is for topic X, find out what the list of Xes ought to be.