Not on a micro-level, no, but on a platform level irrational customers can be dealt with. If they consistently rank low across all businesses you just normalize their votes or remove them as outliers. If they vote randomly it will simply average out (both for the user and for the businesses with more than 10 reviews). If they have a particular consistent bias (e.g. against a particular ethnic group, or particular kind of food) clustering could help (if you can label the clusters, I have some ideas). You just have to work on this, but I don't know if anyone in power cares enough?

Also consider that customers themselves might have (non-algorithmic) reputation that will decide if/how they can affect the reputation of others. Pagerank, if you will.

The entire field is crying out for big data, but then drowns in big dollars.

The problem isn't technical. The problem is the lack of stable, profitable business models around reviews that are long term compatible with being honest.

I think there are compromises, such as doing away with a star rating review system, and only showing positive reviews.

The unfortunate side effect is you never end up learning which companies to stay away from, because a system like that only focuses on the good, but you need a system that benefits both buyers and sellers.