1timspalding
Lucy and Chris are working on new work page. My current job is to write new functions to get reviews and sort them.
So, my question to you is: How do you want reviews sorted? What's important to you? At present there are two options—recent and thumbs—and the default is recent. We are going to come up with something a bit more nuanced than "most recent first."
Detail: First, let me say that we are not taking anything away. You will be able to:
1. See reviews by recentness
2. See reviews by thumbs
3. Filter by language
We are adding:
4. See reviews with a given star rating
You will also be able to make 1-3 "stick"—so if you always only want to see the newest reviews, fine. But the new default order is going to be neither one or another of these, but a calculated mix, based on several factors—let's call it a "heat metric." It will therefore present a mix of new and much-praised reviews, so as to avoid domnation by either a bunch of so-so new reviews, or a bunch of reviews from 2006.
As currently conceived the heat metric will involve:
1. Recentness of the review
2. Thumbs
3. Flags
4. Member activity
1. "Review recentness" is obvious.
2. "Thumbs" is the number of up-thumbs the review got. I'm going to make these "decay" over time somewhat, so that a recent thumbs-up is worth more than an old one.
3. "Flags" is going to a pretty heavy negative, but many up-votes will cancel them out, because some people misused flags.
4. "Member activity" is a metric of whether you are around all the time, or wrote a review back in 2010 and left. I think that—all things being equal—LibraryThing should front the work of current, active members.
So, my question to you is: How do you want reviews sorted? What's important to you? At present there are two options—recent and thumbs—and the default is recent. We are going to come up with something a bit more nuanced than "most recent first."
Detail: First, let me say that we are not taking anything away. You will be able to:
1. See reviews by recentness
2. See reviews by thumbs
3. Filter by language
We are adding:
4. See reviews with a given star rating
You will also be able to make 1-3 "stick"—so if you always only want to see the newest reviews, fine. But the new default order is going to be neither one or another of these, but a calculated mix, based on several factors—let's call it a "heat metric." It will therefore present a mix of new and much-praised reviews, so as to avoid domnation by either a bunch of so-so new reviews, or a bunch of reviews from 2006.
As currently conceived the heat metric will involve:
1. Recentness of the review
2. Thumbs
3. Flags
4. Member activity
1. "Review recentness" is obvious.
2. "Thumbs" is the number of up-thumbs the review got. I'm going to make these "decay" over time somewhat, so that a recent thumbs-up is worth more than an old one.
3. "Flags" is going to a pretty heavy negative, but many up-votes will cancel them out, because some people misused flags.
4. "Member activity" is a metric of whether you are around all the time, or wrote a review back in 2010 and left. I think that—all things being equal—LibraryThing should front the work of current, active members.
2timspalding
Oh, and reviews by friends and people you follow will be strongly boosted upwards for you.
3AnnieMod
See reviews by people in my friend and interesting lists at the top of the list?
When there are a lot of reviews, I tend to look for reviews from people I had met in LT and whose reviews had made me notice them.
In a field of 5-10, I can load all and just look. When there are 20+, that will make the section really useful to me.
Other from that, I really don’t care about the order one way or another (as long as it is filterable by language).
Edit: cross posting with >2 timspalding:. I’d say that they should be able to be at the top, not just strongly boosted. :)
When there are a lot of reviews, I tend to look for reviews from people I had met in LT and whose reviews had made me notice them.
In a field of 5-10, I can load all and just look. When there are 20+, that will make the section really useful to me.
Other from that, I really don’t care about the order one way or another (as long as it is filterable by language).
Edit: cross posting with >2 timspalding:. I’d say that they should be able to be at the top, not just strongly boosted. :)
4bnielsen
Ah, good news!
I usually look at other users' reviews after having completed one of my own to see if I overlooked something obvious. So I'll go for lengthy reviews rather than short.
I usually look at other users' reviews after having completed one of my own to see if I overlooked something obvious. So I'll go for lengthy reviews rather than short.
6bnielsen
>5 lilithcat: :-)
While we are at it, maybe some way of correcting review language? (I'm thinking of say reviews in English, but with their language marked as French.) I'm sure there's a thread about it.
While we are at it, maybe some way of correcting review language? (I'm thinking of say reviews in English, but with their language marked as French.) I'm sure there's a thread about it.
7waltzmn
Speaking just for myself, all of this sounds good. And I mostly agree with AnnieMod that this matters a lot more for things with a lot of reviews than for things with just one or two.
But the one other thing that I would add (at least as an option for the heat metric) is length. Since I'm looking mostly at non-fiction and trying to find out if the book is (a) reliable and (b) readable, I have little use for reviews of less than 25 words -- they don't tell me anything useful. I'd like to be able to promote longer reviews. Others, from what I've heard, prefer shorter reviews.
Admittedly we sometimes get 5,000 word reviews that are pure bloat, but I can pass over those if I must. It's much harder to skip over twenty or thirty too-short-to-be-useful reviews.
But the one other thing that I would add (at least as an option for the heat metric) is length. Since I'm looking mostly at non-fiction and trying to find out if the book is (a) reliable and (b) readable, I have little use for reviews of less than 25 words -- they don't tell me anything useful. I'd like to be able to promote longer reviews. Others, from what I've heard, prefer shorter reviews.
Admittedly we sometimes get 5,000 word reviews that are pure bloat, but I can pass over those if I must. It's much harder to skip over twenty or thirty too-short-to-be-useful reviews.
8MarthaJeanne
I am usually looking for short reviews for a book I might want to read, perhaps 25 - 200 words. It would be helpful to have a way to ignore the 1 - 10 word reviews and the ones that go on and on and on.
>7 waltzmn: Given a choice between scrolling past several very short ones or one loooong one, I'd rather see the short ones.
>7 waltzmn: Given a choice between scrolling past several very short ones or one loooong one, I'd rather see the short ones.
9Maddz
>7 waltzmn: Unless, of course, all those reviews are variations on 'don't bother reading this - it's rubbish'. (It would be useful to know why it's rubbish...)
10Charon07
The new sorting methods sound great, and I’m interested to see the new “heat metric” sorting. In addition to the selectable sorting methods 1-4, I’d also be interested in a way to sort contacts’ reviews to the top.
11waltzmn
>8 MarthaJeanne: I am usually looking for short reviews for a book I might want to read, perhaps 25 - 200 words. It would be helpful to have a way to ignore the 1 - 10 word reviews and the ones that go on and on and on.
That was the main point of my suggestion. I personally would actually like to see (roughly) 100-2000 word reviews. Reviews of fewer than 25 words are likely to be useless.
I probably won't want to see a 10,000 word review -- my first choice would be for those of moderate length -- but there aren't a lot of 10,000 word reviews, and there are a lot of 1-10 word reviews. If the choice is to scroll past one 10,000 word review to get to the 100-2000 word reviews, I'd rather do that than scroll past ten pages of 1-10 word reviews.
The ideal answer would probably be for each of us to have a preferred length (you 25-200 words, me 100-2000, someone else perhaps 1-25), but that makes the heat algorithm awfully complicated. Just putting preference on "longer" versus "shorter" is easier to implement.
>9 Maddz: Unless, of course, all those reviews are variations on 'don't bother reading this - it's rubbish'. (It would be useful to know why it's rubbish...)
Yes, well, that's why you need more than ten words. Which is my point, though I evidently didn't express it well. :-) A review that short might as well just be a star rating.
I must say that, while I have (some) interest in overall star ratings, I don't have much finding reviews associated with a particular star rating. I know there are people who want to read the best and worst reviews of a product, but that strikes me as a lot more applicable to microwave ovens than to books.
That was the main point of my suggestion. I personally would actually like to see (roughly) 100-2000 word reviews. Reviews of fewer than 25 words are likely to be useless.
I probably won't want to see a 10,000 word review -- my first choice would be for those of moderate length -- but there aren't a lot of 10,000 word reviews, and there are a lot of 1-10 word reviews. If the choice is to scroll past one 10,000 word review to get to the 100-2000 word reviews, I'd rather do that than scroll past ten pages of 1-10 word reviews.
The ideal answer would probably be for each of us to have a preferred length (you 25-200 words, me 100-2000, someone else perhaps 1-25), but that makes the heat algorithm awfully complicated. Just putting preference on "longer" versus "shorter" is easier to implement.
>9 Maddz: Unless, of course, all those reviews are variations on 'don't bother reading this - it's rubbish'. (It would be useful to know why it's rubbish...)
Yes, well, that's why you need more than ten words. Which is my point, though I evidently didn't express it well. :-) A review that short might as well just be a star rating.
I must say that, while I have (some) interest in overall star ratings, I don't have much finding reviews associated with a particular star rating. I know there are people who want to read the best and worst reviews of a product, but that strikes me as a lot more applicable to microwave ovens than to books.
12paradoxosalpha
>11 waltzmn: I know there are people who want to read the best and worst reviews of a product, but that strikes me as a lot more applicable to microwave ovens than to books.
Agreed. And I don't have an overall interest in star ratings. The fact that LibraryThing has never even offered an explicit definition of the meaning of star ratings means that star data is pretty much useless, as far as I am concerned.
Agreed. And I don't have an overall interest in star ratings. The fact that LibraryThing has never even offered an explicit definition of the meaning of star ratings means that star data is pretty much useless, as far as I am concerned.
13lilithcat
>12 paradoxosalpha:
The fact that LibraryThing has never even offered an explicit definition of the meaning of star ratings means that star data is pretty much useless
Of course, LT can't. Star ratings are purely subjective. I've read reviews that raved about a book and gave it 3 stars, and others that panned a book and gave it 3 stars. A lot of people put on their profiles how they use the ratings, and they're all over the map.
I suppose LT could give a definition based on someone's (Tim's?) idea of what they ought to mean, but how would LT ever enforce that?
The fact that LibraryThing has never even offered an explicit definition of the meaning of star ratings means that star data is pretty much useless
Of course, LT can't. Star ratings are purely subjective. I've read reviews that raved about a book and gave it 3 stars, and others that panned a book and gave it 3 stars. A lot of people put on their profiles how they use the ratings, and they're all over the map.
I suppose LT could give a definition based on someone's (Tim's?) idea of what they ought to mean, but how would LT ever enforce that?
14paradoxosalpha
>13 lilithcat:
It doesn't need to be enforceable, but when there isn't even a suggestion, it makes the data into garbage. You might assume that 1 star is "bad" and 5 stars is "great," but even that totally rough assumption isn't supported by any guideline from the site. Moreover what aspect of the book is being rated? Does a "high" rating indicate
reader satisfaction?
recommendation for others?
anticipation (for an unread book)?
appetite for rereading?
There could be lots of other meanings, and in the current environment, any user is justified in adopting them for their own purposes. But that makes the social utility of star data vanishingly small.
And as long as I'm arguing with you,
>5 lilithcat: Could you also shove reviews by the author down to the bottom pit of hell?
Authors are entitled to have opinions about their own books and to use the review field to make comments about them. I do, and I don't care if you think those reviews should go to hell. They even get thumbs.
Authors who try to "game the system" to promote their books through posted reviews typically won't admit to being the author anyway. They will use sock puppet accounts.
It doesn't need to be enforceable, but when there isn't even a suggestion, it makes the data into garbage. You might assume that 1 star is "bad" and 5 stars is "great," but even that totally rough assumption isn't supported by any guideline from the site. Moreover what aspect of the book is being rated? Does a "high" rating indicate
reader satisfaction?
recommendation for others?
anticipation (for an unread book)?
appetite for rereading?
There could be lots of other meanings, and in the current environment, any user is justified in adopting them for their own purposes. But that makes the social utility of star data vanishingly small.
And as long as I'm arguing with you,
>5 lilithcat: Could you also shove reviews by the author down to the bottom pit of hell?
Authors are entitled to have opinions about their own books and to use the review field to make comments about them. I do, and I don't care if you think those reviews should go to hell. They even get thumbs.
Authors who try to "game the system" to promote their books through posted reviews typically won't admit to being the author anyway. They will use sock puppet accounts.
15norabelle414
I would rather have a straightforward toggle that moves my connections' reviews to the top than an inscrutable algorithm. (Which isn't to say that I wouldn't use an algorithmic list, just that I like having a more orderly option.)
16elenchus
>2 timspalding:
>3 AnnieMod:
>15 norabelle414:
In the absence of a filter ("toggle") for Friends and Interesting accounts, I fully support the suggestions in >2 timspalding: & >3 AnnieMod: to boost those accounts in my feed.
Generally I support the other proposed features in >1 timspalding:, as well, but the Friends and Followed functionality is more important to me than these others.
>3 AnnieMod:
>15 norabelle414:
In the absence of a filter ("toggle") for Friends and Interesting accounts, I fully support the suggestions in >2 timspalding: & >3 AnnieMod: to boost those accounts in my feed.
Generally I support the other proposed features in >1 timspalding:, as well, but the Friends and Followed functionality is more important to me than these others.
17GraceCollection
I would be interested in 'filtering' reviews by libraries which are similar to mine. Not everyone cares about someone who has a lot of the same books they do (which is a valid point; I still have in my library books which I think are utter garbage but present a learning opportunity, and someone else who retains those same books as favourites probably has different opinions than me on more than just those particular titles) which is why I think it should be a filter option and not part of the new algorithm.
18SandraArdnas
Please, please have 'friends and interesting libraries' reviews appear at the top. It's one of few reasons I still visit Goodreads for books with a lot of reviews. It would also make these categories far more useful and I'd again start paying attention to it and adding people whose reviews I appreciate.
19timspalding
>12 paradoxosalpha: Agreed. And I don't have an overall interest in star ratings. The fact that LibraryThing has never even offered an explicit definition of the meaning of star ratings means that star data is pretty much useless, as far as I am concerned.
I'm half with you. I personally don't find them useful. When you get deep into the data you realize that most books get a 4, and the standard deviation is low. Meanwhile you get oddnesses like how Marlowe gets higher than Shakespeare, or how series always get better. (Both are about selection bias.)
That said, my opinion doesn't matter. Even on LibraryThing, people like stars.
Also, FWIW, one-star reviews are interesting and/or funny. Sometimes it's worth it to listen to the counter-argument to a book.
As others are saying, the notion of LibraryThing telling people what stars should mean doesn't fly for many reasons.
I'm half with you. I personally don't find them useful. When you get deep into the data you realize that most books get a 4, and the standard deviation is low. Meanwhile you get oddnesses like how Marlowe gets higher than Shakespeare, or how series always get better. (Both are about selection bias.)
That said, my opinion doesn't matter. Even on LibraryThing, people like stars.
Also, FWIW, one-star reviews are interesting and/or funny. Sometimes it's worth it to listen to the counter-argument to a book.
As others are saying, the notion of LibraryThing telling people what stars should mean doesn't fly for many reasons.
20paradoxosalpha
>19 timspalding:
Well, I will continue to neglect stars on LT, and I hope that they will not be covertly figured into any display choices.
Edited to add: The basic principle of quantifying an unspecified quality is deeply flawed.
Well, I will continue to neglect stars on LT, and I hope that they will not be covertly figured into any display choices.
Edited to add: The basic principle of quantifying an unspecified quality is deeply flawed.
21krazy4katz
>12 paradoxosalpha: I confess that even my own star ratings are not reliable. Depends what else I have read recently and liked or didn't like as well as my mood for the genre by the time I finish. Sometimes I go back and change my stars. However I still want them in the reviews. The difference between a 1-2 star rating and a 4-5 star rating is probably meaningful most of the time. Of course context is important. If a review only has stars, I ignore it.
22SandraArdnas
>20 paradoxosalpha: I understood that star ratings are not planned as a part of algorithmic calculation, but rather a filter of its own, to see only 1 star ratings for instance, which is useful when you want to check what very negative or very positive reviews have to say, or the middle of the road ones for that matter. I sometimes want to check more negative reviews for highly rated books. Sometimes it's just a case of unsuitable book for said reader, but sometimes people have insightful objections that help decide whether it's something to pile on my TBR.
