interesting. my interest is less about breadth, and more about the idea that anyone can become an "instant expert" if exposed to deep deep education on a topic. for instance, how would our views on nuclear power look if we all had to watch this video before casting our vote on politicians with a range of views on nuclear power?
in addition to the issue of merely having "more information", the jury context pulls people away from their political in-group and social media bubbles or echo chambers. this makes them more psychologically capable of wrestling with their own identity, e.g. accepting the reality of climate change even though your in-group and information/media bubble is skeptical. serving on a jury creates a sense of civic responsibility. your "tribe" becomes that group of community peers more so than your fellow fox news or democracy now watchers.
diversity itself isn't even so much the point to me. it's just a consequence of using random selection, the point of which is merely to make sure the group's overarching ideological center of balance is roughly approximate to that of the electorate itself. if you could turn a slider that pulls all the members toward the center, while preserving the statistical accuracy with respect to the overall position of the two groups in issue space, i doubt you'd get much difference in outcomes. and you might even have some improvements in efficiency insofar as a group of ideologically aligned centrists might get along better and have a better chance of embracing the truly optimal positions, which by definition should be at the political centroid in the first place. a kind of related video.
you could look at it the opposite way. right now we have a very diverse election turnout, because people all across the political spectrum vote. but the ideological center of balance is distorted, because there are demographic turnout disparities. i.e. the people who show up to vote are older/whiter/wealthier/etc. than the electorate as a whole.
that said, this is purely an academic argument, because there's no way to directly select a statistically valid sample of the population (as in, same rough ideological center of balance) without a random sample, which is necessarily going to be diverse. and i accept it's quite possible the benefits of diversity outweigh the potential for in-fighting between the communist atheist on the jury and the free market religious conservative on the the jury.
but it's an important consideration, because if the juries vote in single-member districts, or if they use multi-winner voting methods that aren't proportional (e.g. plain score voting), then you'll get highly competent elected officials who closely match the average views of the voters they serve (well, what those voters actually want more so than what they think they want, given they've deliberated). do we want the juries using proportional representation tho, so they elect e.g. city councils that are diverse? a whole other can of worms, but probably one that will come up when we put our first election by jury proposal on the ballot.
* i think we should definitely highlight the fact that demographic turnout disparities are solved with sortition/EBJ, because you effectively get compulsory universal voting, stochastically speaking.