Got same results. Perhaps they’re recommending the books you need, not the books you want.
I put in Mortimer Adler’s “How to Read a Book”, and it pointed me to a list of ‘Books and Reading’.
Wow; it never occurred to me.
The fundamental failure of these recommendation algorithms is that they assume that if I read and enjoyed a book, I would want to read another book like it next. When I would most likely want to read something entirely different.
What should I read next? Probably something from the ginormous piles of stuff I’ve already purchased. Tsundoku 4 Lyfe, yo.
[quote=“ChuckV, post:6, topic:50854, full:true”]I put in The Story of O and the results consisted entirely of Anne Rice, with the exceptions of what appears to be a Christian romance novel and Kahlil Gibran’s The Prophet. Weird.[/quote]Y’know, she’s written some freaky sh!t, to put it mildly. Very mildly.
I put in Neal Stephenson and got Neal Stephenson back. I put in Margaret Atwood and got Margaret Atwood back.
I know of the Sleeping Beauty trilogy. It was just odd that the list was so heavily weighted towards her work, and the two exceptions were the icing on the cake.
James Chiles, Inviting disaster - lessons from the edge of technology, nothing. Added it in so it now shows, but with no suggestions.
Richard Poisel, electronic warfare anything, nothing.
Robert McShea, Test and evaluation of aircraft avionics and weapon systems, nothing.
Vincent Dunn, Collapse of burning buildings, “Sorry, there were no results - this is probably because your search edition doesn’t appear in many users’ booklists.”.
Barbara Moran, The day we lost the H-bomb, nothing.
Richard Thompson, Crystal clear - about WW2 crystal oscillators history, nothing. Name query offers instead Jane Heller, Crystal clear, which is something entirely different.
Bee Wilson, Swindled - the dark history of food fraud, nothing.
Trevor Kletz, What went wrong, you can guess for yourself.
Charles Harper, Handbook of ceramics, glasses and diamonds, drumrolls please, NOTHING.
And that’s the lighter stuff.
I was interested to learn that one of the first ten recommendations for Ralph Ellison’s Invisible Man is The Book of Five Rings. Now I can’t stop thinking of the crossover fanfiction possibilities.
A lot of these recommendation engines are only as good as the people that use them. If a book is only in one user’s book list, it will recommend the other books in that user’s book list. I never liked whatshouldireadnext for this reason. I’ve used Your Next Read a handful of times, and found it much better. Yes, it’s still going to recommend the same type of book (ie if you put in a John Scalzi book, it’s going to recommend other fun sci fi stuff), but really, how else should it work? “You like this particular genre of sci fi, here’s some more similar stuff”. It’s helped me find authors that I’d never really read much of in the past, when I’m looking for something in the same genre. If I’m looking for variety, I’ll head to some of the “best books of the year” lists in other genres, rather than trying to get a recommendation engine to recommend something to me based on one book that I give it.
I think I can see the logic there: Ralph Ellison’s “The Invisible Man” -> “Hey, H.G. Wells also wrote a book called The Invisible Man.” -> “In The Hobbit there’s a ring that turns you into an invisible man, that’s kind of close.” -> What’s better than one ring? Five! So my recommendation is: “The Book of the Five Rings”
I think I’ve seen (not read, necessarily) half a dozen ‘Netflix thinks I’m gay’ articles over the years. I do think their bottleneck is the smallish, mostly-mediocre catalog, not the algorithm. But surely there must be some room for improvement.
They had this talk by the Dalai Lama up there and I like His Holiness so I watched it. I’m pretty sure the algorithm then somehow assumed, “you are a complete fucking moron who will swallow any pseudo-science, woo crap we feed you.”
Right because Dalai Lama = The Secret.
Martin Luther King = Madea
I’ve had better luck on Netflix mining the ever more specific genre classification buckets.
“dark animated sci fi foreign horror comedic TV series”
For algorithms of this type, maybe they could ask why we liked a particular feature instead of just entering star counts.
Sadly, many people who DO find The Secret to be worthy if anything beyond a garbage bin are also very into the Dalai Lama. Likewise MLK and Madea. And the only thing these recommendation algorithms go on is “Well, many other people that liked X also liked Y, so if you like X, chances are that you like Y too”. Unfortunately, it is so very, very wrong a lot of the time. What I wish Netflix had was a “Stop showing me bullshit like this” button, but I try to emulate that by rating the silly garbage stuff 1 star, hoping that Netflix will get the point.
Yes, it’s not so off in one sense, but it just seems like a random guess more than not. I can’t remember what movie it was I watched, but it was like the jackpot of recommendations. There were a ton of great, interesting movies it pulled up. I wish that they were all like that. I think the more niche a movie is the more it is likely to find something truly interesting as a result of you watching it, whereas, oh “You watched Good Morning Vietnam?” you might also like “Mrs. Doubtfire.” I guess it’s not narrowing down things too much to watch a movie so many others have, and did you watch it because of Robin Williams, because of Vietnam, because you like things that got an Oscar. I know if I were to rate all the movies I watched it’d probably do better, but I just don’t feel like feeding so much information into their system. It’s enough they know what I watch.
I wouldn’t be surprised, with today’s emergence of conversation-agent grade AI already used in e.g. advertising or social data mining, if instead of manual rating you soon got a choice of talking about the movie with a bot, in a videochat-like session, and sharing your impression about the movie in more details and more dimensions than one-to-five stars can ever contain.
They that can give up essential privacy to obtain better movie recommendations deserve neither privacy nor better movie recommendations.
This topic was automatically closed after 5 days. New replies are no longer allowed.