You have two different problems to solve before coding:

1) How to express uncertainty/controversiality?

A common approach is relative standard deviation (mean/standard deviation). Or median/quartil. I'd not care too much about the differences.

2) How to combine average with uncertainty?

That depends on your application. Does it rate a book that people love or hate and discuss about? You should probably read it, too. Or is it a tool that performs badly in some situations but is good enough for the average job? If you can't anticipate your usage, you should stay away from it.

Keeping the two dimensions seperated (and plot it?) leaves that decision to the user. In the classical vendor selection he faces a multidimensional problem anyway, e. g. on price. Or you could report and sort on a different point of the distribution curve. Using average - 2 * sigma roughly means "90% of votes are better than X" (if you assume gaussion distribution)