Let’s finish up with what philosophers call “the question of meaning.” If we are to believe in a “fair reputation system” we have to ask: what, grasshopper, does a one-star (or a five-star) review really mean?
Yelp’s view (and TripAdvisor’s, and Netflix’s, and Amazon’s, and Airbnb’s, and eBay’s … you get the picture) is that a rating is an indication of individual preference, and that aggregating these preferences can be used to recommend the experience to others. If “reputation” is the social distillation of others’ opinions about ourselves, then these rankings are a measure of “reputation”, distilled multiple times into a single number.
But just because that’s what the aggregators would like a rating to mean doesn’t make it so. There is a world of context behind each click that may not fit into any algorithm’s assumptions. In Botto Bistro’s case, a one-star review is a gesture of support, a way to thumb your nose at an impersonal system, and a quick path to cheap pizza. It’s a dramatic example of how one “one-star review” may have a completely different meaning from another “one-star review” even though they are recorded in Yelp’s systems in the same way.
It’s tempting to think that if only “fake” reviews could be removed, then the remaining ones would indicate preference, but there’s little to back up such a claim. In other contexts (like eBay or Airbnb) a five-star review is more often a courtesy than a rating. Less dramatically, in the case of Netflix, one outcome of the Netflix prize was that reviews of one movie are heavily primed by what other movies we have recently reviewed (so that “objectivity” is tainted).
And maybe a “fake review” should be seen as a commitment on the part of the restaurant owner, an investment in branding that indicates an intent to stay in the marketplace, a sign that the restaurant can, in fact, be trusted — just like any other investment in branding.
Ratings do not always mean what data scientists would like them to mean. Human expressions are protean, and no attempt by Yelp or others to pin them down will succeed completely. There is something postmodern about ratings, just as any other human expression. We imbue them with meaning and intent outside of what the system wants, and there is no objective fact behind it all that is the “true” rating. Il n’y a pas de hors-click.
Reputation is a multi-faceted, qualitative concept. It has been pushed through a meat-grinder by digital reputation systems and has come out the other side homogenized, devoid of texture, but easier to digest. There’s nothing inherently wrong with reputation systems by themselves, but giving them too much authority or influence will inevitably throw up bad incentives on the part of the system owner and those taking part in the reviews. Before we declare “fake reviews” to be a crime or an obvious case of bad morals, we at least need to demand some accountability from reputation site owners.