What do the Amazon Star Product Ratings Mean? Can we get Transparency?

Well they are just an average of all the individual ratings, no?   Well that approach would be very manipulable, and you see right away that they place more weight to people that have bought the product.  Turns out it's much more than that.   It's machine learning and lots more that is proprietary and un-revealed.  So not transparent.   And in a complex context.   Which makes you wonder what transparency for this example of machine learning means.  Even if we were given the exact set of algorithms used, what use would they be?   Depends also on the data used to create the learn the algorithms.   And how we as customers plan to use the ratings.   So can we really be transparent in AI?  Any more that a human can be transparent?  Quote below made me think this, good article:

What do Amazon's Star Ratings Really Mean?  In Wired   By Louise Matsakis

" ..... Starting in 2015, Amazon began weighting stars using a proprietary machine-learning model. Some reviews now count more than others in the total average, based on factors like how recent they are and whether they come from “verified” purchasers (meaning Amazon could confirm the reviewer actually bought the item they claimed to love or hate). David Bryant, an Amazon seller who also blogs about the company, believes Amazon may also take into consideration factors like the age of the reviewer’s account and the average star rating they usually leave. “There appears to be some discount applied to reviewers who predominantly leave negative reviews,” he says. .... " 

Better Recommendation Engines

Researchers Devise Better Recommendation Algorithm 
Improved recommendation algorithm should work especially well when ratings data are “sparse.”

MIT News  By Larry Hardesty

Researchers at the Massachusetts Institute of Technology (MIT) have developed a new recommendation algorithm based on a theoretical analytic framework using cosine similarity, which they say should work better than current algorithms. The researchers note the algorithm should be especially effective when ratings data is "sparse." Sparse data means there may be so little overlap between users' ratings that cosine similarity is rendered meaningless, making it necessary to aggregate the data of many users. MIT professor Devavrat Shah says the framework assumes the relative weight a user assigns to ratings remains the same, and each user's function is running on the same set of features. Shah notes this yields sufficient consistency to extrapolate statistical inferences about the probability that one user's ratings will predict another's. The team used the framework to demonstrate that, in instances of sparse data, their "neighbor's-neighbor" algorithm should return more accurate predictions than any known algorithm.... " 

Amazon, Data and Store Operation

An interesting thought.   In the past rating was also done by knowledgeable employees in the store.  Now just organize it better.   My own retail experience makes it seem obvious.    But probably late.

Will Amazon’s use of data transform how retailers operate stores?    by Tom Ryan   in Retailwire:

How does Amazon Books differ from Barnes & Noble as well as the many book chains that have bitten the dust in recent years? Discovery and data.

That’s what Jennifer Cast, VP of Amazon Books, told a group of journalists Tuesday in a tour of Amazon Books’ first New York City store at Shops at Columbus Circle, its seventh store. (See photos on our Facebook page…)

The primary way Amazon is linking data to discovery is Amazon.com ratings. All books on display (except some best-sellers and new books) have ratings above four so customers “know these are great books.”

The ratings are showcased in special “feature” displays throughout the store. The most common shows ratings above 4.5 or 4.8 in themed sections (non-fiction, cooking, etc.).  ....   " 

I see the term 'Blended retailing' being used.   Good further thoughts in the discussion.

China Rating Everyone

A plan to organize Chinese society by rating everyone.  In the CACM: " ... Imagine a world where a  government monitors everything you do, amasses huge amounts of data on almost every interaction you make, and awards you a single score that measures how "trustworthy" you are. ... " 

Beware of Rating People Online

Not usually a follower of this, but happened on this bitingly vicious satirical look at the online 'ratings' game for people and businesses.   Exaggerated, as all satire should be.  Has a glaring hole in its logic, as pointed out in an excellent The Verge review below.  Still makes its point well.  Plausible that we might head in this direction.  Have seen systems proposed that derived personal reputation. Could make any consumer technologist squirm.   First episode of 'Black Mirror' on Netflix.

Black Mirror's third season opens with a vicious take on social media by Tasha Robinson