Monday, November 17, 2014

On the Value of Up- and Downvotes

As you may be aware, I'm something of a fan of terrible jokes.  So when I came across this one over the weekend, I could hardly help but share it:

Sherlock Holmes and Watson went to a hunting lodge.  Sherlock set several game traps on the first day, but when he went to check them, he found that only one trap has been sprung, and that the fox it snared had gnawed its own leg off to escape, leaving nothing but a paw behind.  When he came back to the lodge, Watson asked, "Did you catch anything?" 
Holding up the paw, Sherlock announced, "The game is a foot!"

How people can fail to appreciate gold like that is beyond me.

Moving on, though: I've been thinking about those little green and red thumbs which FiMFiction lets you use to vote on stories, and what--if anything--they add to the reader experience.  Click down below the break for my thoughts.

The ability to rate media semi-anonymously is very common on the internet.  Whether it's a simple thumbs-up/thumbs-down feature like Youtube and FiMFiction have, a star- or number-based system of some sort, or in some cases, even a single-option feature (usually, an upvote or equivalent), sites which host or aggregate art often give readers a simple and general way to express an opinion without actually commenting.

But who actually benefits from this system?  It's certainly not the authors; unexplained expressions of approval or disapproval serve as nothing but ego-boosters, and if that's the goal, then only the single-option feature makes sense.  After all, an unexplained expression of disapproval for a story (or picture, or video, or whatever; I'm going to stick with "story" and "author" for the rest of this post, for simplicity's sake) has exactly the opposite effect.  And even in the case of that single-option feature, the fact that the author can almost certainly see many other stories with much higher upvote totals can still easily turn "insufficient" upvotes into a net negative.  A secret-upvote system, with only the author able to see his or her stories' upvote totals, might serve this purpose, but I've never seen such a system.

It's not particularly to the potential reader's benefit, either.  I have never yet seen a site with story voting where the larger part of the userbase felt that the votes on stories corresponded cleanly to quality or enjoyability.  True, some people use story popularity as measured by votes to help winnow the list of available stories to chose from, and popularity obviously increases the likelihood that one will be exposed to a story in the first place, but that's not the same as saying that those votes are helping those people find stories they'll enjoy.  An excess of downvotes relative to total votes is a pretty reliable indicator of poor quality and/or fetish-y content, but few people ever seem to feel they do a good job of identifying good stories.  At best, it seems that up- and downvotes are only moderately useful to potential readers.

That brings us to the last group affected by these things: people who've actually read the story.  It's a bit counterintuitive to think that they're the ones who benefit most from these votes--they've already read the story, what do they care about the votes?--but I think they end up being the people who get the most utility out of these kinds of features.

The ability to vote for (or against (or just to rate)) a story provides, I feel, an emotionally satisfying cap to one's reading experience.  After finishing a story one enjoys, clicking the thumbs-up icon feels almost like doling out a reward.  On the other hand, hitting the red thumbs-down mitigates the feeling that one has wasted their time reading a bad story by giving one a way to express that dissatisfaction.  And in both cases, it enables people who are unwilling or unable to leave comments, or who just feel like they don't have anything useful to say beyond "I (dis)liked that" to get that sense of contribution without having to try to put any of it into words.

Of course, plenty of people use up- and downvotes for other purposes--folks who downvote stories without reading them for having the the "wrong" characters or ships or whatever, for example--but when you look at those votes as being primarily a tool for readers who've already finished a story, rather than for authors or potential readers, that kind of misuse becomes much less of a problem.  That doesn't change the issues with how those votes are used by sites to promote stories (and those are problems I don't have any ready solutions for), but when you consider who really benefits from these features, it makes any such "abuses" feel less relevant.  In the end, it doesn't matter if someone else is trying to game votes; they still serve the reader post-story just as well.


  1. I was all set to comment before finding you wrote exactly what I was going to say. I find downvotes to be very good predictors of how much I'll hate a story. Upvotes aren't nearly as reliable

    Even if the rating isn't accompanied by a comment, it's still at least some feedback. Not too valuable, but not completely worthless either. It can help keep all those comments in perspective, and the comments in turn offer perspective on the votes

    Let's not forget the ratio of votes to views, either. A story that convinces more of its readers to vote at all is either doing something right or very wrong

  2. Ah, the voting system.

    On my newest story, I got a comment that basically said "I'm giving you your first downvote because I can," accompanied by a link to a trollface picture. A downvote later disappeared, and I'd like to think that was him deciding to give up the joke after Alexstrazsa hassled him about it.

    Different users have different interpretations of what constitutes a reason to vote one way or the other, too. I've upvoted stories I didn't like, because I recognized the skill that went into the writing, and any objections I had to it boiled down to personal preference. Contrast that with the people Chris mentioned, who will downvote a story without reading it because they didn't like the choice of shipping pair. I've also come across people who can't divorce their personal tastes from an objective assessment of whether a story was good or bad. In my mind, saying a story is bad is a very far-reaching thing; you're not only saying that you didn't like it, but that other people shouldn't like it, either. But it's certainly an individual's prerogative as to which standard he requires in his voting.

    Unfortunately, fame begets fame, and voting means a whole lot here. Some stories near the top of the highest-rated list have so many total votes that they can absorb a few downvotes with no trouble. Others cycle through and drop wildly from a single downvote. I've had three separate stories crack the top ten with a relatively small number of views, but it's a mixed blessing; being up there invites more viewers and the inevitable single additional downvote (or, in one case, the first one), which, when it happened, dropped them an average of about 250 spots.

    It's definitely true that the voting is much more useful for picking out a bad story than a good one. A favorable ratio only intermittently indicates quality, while a bad ratio pretty reliably gives an accurate judgment. As a pre-reader, I'll glance at the voting to give me an idea of what awaits me, but as soon as I start reading it, that goes out the window.

    My own voting record is probably very bad. I rarely read anything I'm not reviewing, so I'm not reading much of what I would choose to on my own. I don't vote much at all, and I won't downvote a story just because it's poorly written. It has to be that the premise is untenable or the narrative taken in such a wrong direction that I ragequit, essentially. Not laughably bad, but painfully bad. But again, because I'm reading what's sent to me and not what I might choose to read, I see more bad than good. I bet I've given out more downvotes than upvotes, but there wouldn't be many of either one.

    1. I'm the exact opposite, only rarely downvoting fics. I try to filter what stories I add, and if I'm not enjoying a fic, I'll usually quit prematurely. I only rate stories that I've actually finished, and I'll only give a downvote if I actively disliked it. Merely OK stories still get an upvote

  3. In my experience as a reader, the system is actually pretty useful for discriminating between potential stories to read. Not nearly as much as a review, of course, but much more than something like's favorite system, which shows how many people like a story but not much else.

    In essence, I regard the upvote/downvote system as conveying information on two axes: number of votes represents number of readers who had strong opinions on the story, and like:dislike ratio represents the controversiality of the story's content.

    Based on that, I can actually build up a decent impression of what a story will be like based on the votes. Low votecount coupled with high like:dislike ratio tends to mean "inoffensive but not that interesting", high votecount with low like:dislike ratio tends to mean "story which is good if you can get into it but which some fraction of the population will have Issues with", low/low tends to mean "lousy story", and high/high tends to mean "story which everyone loves".

    None of those are accurate all the time, of course, but they're accurate a good enough fraction of the time that I continue to use them as heuristics when deciding whether it's worth reading a story, in cases where I don't have an actual review on hand. (And I always avoid stories' comment sections before reading, because spoilers, so comment-based reviews don't help.)

  4. I said this when ED removed star reviews, I'll say it again: rating fics works when you tie the rating to a >50 character review tied to an online handle, which is why Amazon reviews are spectacularly useful and sorting by "favorites" yields Methods of Rationality (warning, so interesting it's hard to put down once you get into it) second, despite broad consensus that it's the best fanfiction ever written.

  5. I pretty much never notice ratings unless the red bar is huge. Far too often, I'll read something with over a hundred upvotes and give the second or third downvote.

    1. Downvote lightsabers are useful for avoiding clickbaity stories.

  6. Admittedly, the first thing that catches my eye as a reader is the vote ratio. I'll almost never consider it unless there's a serious amount of red, because I'm willing to trust that the trolls and otherwise callous voters are an incredible minority in the community, and the odds a single story would gather so many is slim. Comparing the vote ratio to the relative number of unique views is something else to take into account, too. Popular authors can publish a story where a high percentage of its viewers will vote on it (even before reading it if at all, which is another gripe), while others will gander many views but get very few votes in comparison. I'm more inclined to give the latter a read than the former.

    That being said, a story's description and perhaps the first few paragraphs of the first chapter are what make me decide whether or not to invest myself in a story. I try to only vote on things I read completely, and that depends on my overall enjoyment of the story—a simple "did I like this or did I not". If I don't read a story, I don't vote on it. I upvote more than I downvote because of that.

    Still, upvotes/downvotes are a better than star ratings for such a large community, if you ask me.

  7. Time was that I'd only rarely downvote something, and the story would have to do something pretty horrible to get that from me. Hell, it used to be hard to just get me to stop reading something unless the author did something truly heinous. These days, I find I'm doing a lot more of both.

  8. Bad Horse at one time was measuring stories by Reader Retention Ratio, i.e. what percentage of readers bail after the first/second/third... chapter. As a general rule, it only works for multi-chapter fics (duh) and takes a fairly good sample before you can draw conclusions, but I liked the idea (and not only because my own stories ranked fairly high when using the metric :)

  9. I think voting is a good measure of general quality, but not excellence. I probably should downvote some of the stories I disliked. I've stuck to pretty much only upvoting.

    The popularity leads to more attention is a tricky thing, but in general I think the voting system is pretty good at a broad sort. Of course, I really like about half of the stories in the all time top 10 so I'm hardly unbiased about it.

    Decided I'm gonna grab some stories from the low end of the ratings and some from the middle, just to compare with the upper end of things. Grab some practical data and all that. First discovery: Getting to the middle of the list where it has mostly even upvotes/downvotes is a pain in the neck.

    1. Now, you are going for a fairly risky proposition there. In my experience, FiMFic tends a little on the hugbox side, so I'd consider a mediocre rating to be more in the five- or ten-to-one ratio. An even one is rarely good, except perhaps for ones that don't have enough total votes to assign any meaning to them.

  10. Voter weighting.

    Anyone can vote, but if you provide no feedback with your vote it will be given far less "weight".

    It would also move to a -10 to +10 scale, but users who only used the extremes (as is often the case in such systems) would have their weighting greatly reduced since they aren't using critical thinking to give an accurate representation of the story (since it should be rare that a story is deserving of one or the other).

    The additions of tags as to why a story is given the rating would also make it easier for readers to browse. You could tag one with Strong Plot and Consistent Characterization, but also list Poor Punctuation as a con.

    Users who consistently gave detailed feedback, complete with tagging, full reviews and utilizing the entire voting bar's spectrum would gain Trusted/Super/Revered Reviewer status and gain far greater weighting than your average browser (like Rotten Tomatoes). This would be further incentivized by superficial rewards, like themes to decorate their user page or icons beside their name (TF2 hats, in essence). The threat of losing these and the great amount of time invested in acquiring them by gaming the system in some way would serve a suitable deterrent.

    A weighted reviewer's power extends more in their ability to lift a story from the morass than to strike one down; their job is to bring quality to light, not act as a gatekeeper to the featured box.

    To prevent vote-bombing, a story's status to being put into the featured box would be determined by its positive score first, with negative score being disabled/invisible to other users for the first twenty-four hours or so. This would prevent groups of scorned writers from exacting ego-fueled vengeance on reviewers, as well as putting more emphasis on voting quality.

    The featured box itself should also get an overhaul. There could be tabs (New featured stories tab, new chapter featured tab, reviewer-recommended tab, mature story featured tab ect.) to broaden the number of chances a story has to enter. The featured story pool could also be increased; there could be ten stories in a tab, but when you refresh a new set of stories are there. This would increase the number of stories that can get featured while preventing the huge cycle of popularity that usually accompanies it (an unfair advantage given to any story that peaks for even a short time in the box).

    All of this, preferably, would be controlled by an algorithm or other code set, to reduce overhead on the administration once put in place.

    Any thoughts on the idea?

    1. What's funny is Fimfiction used to have a 5-star rating system, just like EQD. It lasted through the February writeoff before switching to the up/down voting system we still have today.

    2. I know, and it was axed because people didn't use anything but the 1 or 5 stars. The weighting solves that by making such people not have a large say in the overall score for the story.

    3. Sadly, linking commenting to voting in any way is pretty much unfeasible. Without any kind of moderation, it will most likely leave authors with lots and lots of useless comments, which would make it even harder for them to get useful criticism. And moderation is just too costly, since people would quickly burn out on having to read lots and lots and comments, then be forced to reply to people contesting that "this ship sucks!" is actually a good criticism. And even a perfect moderation would eventually lead to people complaining about elitism, poor standards, and so on and so forth.

      Something about people who spread their scores more on a multi-level scale getting more voting power is certainly more interesting... except you would probably get people gaming the system by voting randomly on stories they don`t care about. And those votes would most likely target random stories on the front page, which save for the feature box, are the ones for which this random noise would be the most damaging.

      Also, preventing people from seeing downvotes would help certain divisive stories, such as Fall of Equestria. Not sure that is a good thing

    4. Minimum wordcount, system can spot copy/pasting or outright gibberish. These would not be advertised; like Reddit votefuzzing, for all intents and purposes it appears as if their downvote has been added but in reality it is only reflected on their screen.

      Time spend in-document would be another invisible factor. If they obviously only had it open for a few seconds before casting their vote, its value would be diminished/ignored.

      Its purpose is to bring fairness, so unfortunately some truly unworthy stories might make it up, but the power of veteran reviewers would hopefully negate the drooling masses from upvoting whatever story was written with one hand.


  11. But who actually benefits from this system? It's certainly not the authors; unexplained expressions of approval or disapproval serve as nothing but ego-boosters... It's not particularly to the potential reader's benefit, either.

    I disagree strenuously on both counts. I find the votes to be rough indicators of a story's quality, good indicators of its level of engagement ((upvotes + downvotes) / views), and very good indicators of what type of people read fimfiction. This last is critical to interpreting the audience response to a story, and to learning what people really want in stories.

    I want very much to see the upvotes, downvotes, favorites, and viewcounts on my stories, and on other stories. That's one huge advantage fan-fiction has over print fiction.

  12. There are many ways of making upvotes and downvotes more informative. The simplest is Bayesian: For each upvote or downvote, report the number of bits of information that vote provided. Then an upvote from someone who upvotes half the stories he reads will count as 1 vote, while an upvote from someone who upvotes 1 story out of 16 will count as 4 votes. A downvote from someone who downvotes all Mature fics, or all fics by one person, will count as zero.