Fi doms tend to enter a full brain listening mode (and INFPs in particular, can sustain this mode for a long period of time, while ISFPs tend to listen deeply for a bit, and then move into action).
As far as brain regions, Fi-doms show high activity in T3 (precise speaker) and T4 (Intuitive Listener), F8 (Grounded Believer). INFPs show more activity in F7 (Imaginative Mimic) while ISFPs show more activity in T6 (Purposeful Futurists).
Subjectively, it feels to me like when I listen, I'm not just paying attention to surface content, I'm trying to build up a whole model of the other person's perspective, what's important to them, and what's behind their words. It sometimes feels like it's a kind of "mental simulator," that's trying to run a crude simulation of the other person's perspective and emotional state.
This also makes sense when one considers that each person's Fi values system is different. Unlike the other judging functions, there's no way to directly translate and evaluate objectively (via logic or shared social constructs). Hence, the only thing to do is to try to enter into the other person's perspective, and try to figure out how what they are saying makes sense to them. I found it puzzling as a kid since others wouldn't do the same for me... so it felt very much like a one-way street that way.
Nardi did observe that one's an Fi dom makes up their mind, they may no longer be willing to listen, so isn't like Fi-doms are the understanding saints all the time. One example he was an INFP having a discussion with an ESTJ on gun control. The INFP stated in a full brain, deeply listening state for a full ten minutes (Nardi was astounded). Then, the INFP made up her mind and that was that.
And, while it is possible to use alternate brain regions to do things (and, of course, a subset of people have different brain region arrangements, particularly some left-handers), generally the process is less efficient. Nardi had people do math problems. Although some people used other parts of the brain to do math (verbal centers, for example), they were much slower and less accurate than the people who used the typical brain region for the task.
And again... I like Nardi a lot, but I do take his findings with a few grains of salt. They haven't been peer reviewed or replicated, but he comes across as someone who believes he is honestly researching and reporting what he finds.