Welcome to Codidact Meta!
Codidact Meta is the meta-discussion site for the Codidact community network and the Codidact software. Whether you have bug reports or feature requests, support questions or rule discussions that touch the whole network – this is the site for you.
Should we start displaying the score of a post instead of the raw votes?
Currently, when viewing a post, Codidact will show you the raw votes on a post, with the breakdown into upvotes and downvotes:
There's been some feedback that this is a bit too much to show, especially coming from platforms like Stack Exchange where they generally just show the aggregate score of upvotes and downvotes as one number (with the option to expand the votes to see the split). We decided to show both counts automatically to better show when there's controversy.
However, we now also have another option. We have a method for scoring posts that assigns a score between 0 and 1 to each post.
Perhaps instead of showing the raw votes on each post, we should instead show the post score (e.g. 0.81363... or 0.3793...), rounded to the nearest two or three decimal places (so that it would show as 0.937 or 0.276), with the raw votes available on request, perhaps either on click or in the tools menu.
This would take people a bit of time to get used to, but it might be worth that initial adjustment time, since this... is our scoring system and we want people to be familiar with it quickly.
This has the added benefit of making it much clearer why answers are sorted the way they are by displaying their score (that's currently computed without being displayed) for everyone to see. The raw votes matter less than the computed score.
If we show raw Wilson score I think we're going to see a lot of confusion and questions -- "is 0.65 good?" "what does it …
3y ago
I agree that on a list of questions, one clear indicator of fitness is most helpful. On a post's own page, it might make …
3y ago
All the proposals so far are missing what people really want to know, which are two orthogonal metrics: How good/bad …
3y ago
instead of - Absolutely not! Showing separate + and - votes is a good thing. Or as they say, that's not a bug, it's a …
3y ago
In my view displaying fractional numbers representing an unintuitive measure would be even worse than displaying two int …
4mo ago
5 answers
If we show raw Wilson score I think we're going to see a lot of confusion and questions -- "is 0.65 good?" "what does it take to get to 0.8?" etc. Even if we explain that it's a value from 0 to 1 indicating some quality measure, I don't think it's all that meaningful to show that number to most people -- and certainly not as the only indication.
We should consider context. I think it's very important to show the raw votes for answers to questions. The sorting conveys relative ranking, so we don't need to show meters or scores or anything else there -- just the votes. I'm not averse to making the scores available, but they shouldn't be primary because the page already conveys this information in a more intuitive way.
Why is it important to show the raw vote counts? Because ranking, and scores were we to show them, give a sense, but if you're considering which of these answers on EE to follow to modify your phone, you really ought to know that while that answer has lots of upvotes it also has lots of downvotes. That should prompt you to look more closely. Maybe the downvoters are wrong or misunderstood, or maybe they know one missed step and you'll brick your expensive phone.
For top-level posts (questions or articles), though, our priorities might be different. Is it actually valuable to see question scores on the question list at all -- either raw scores or a meter or both? What are we trying to accomplish by showing this? If we answer that, I think we'll be in a better position to decide what to show there.
Finally, on the question page I talked about answers but not the question. Since we're showing raw votes for all the answers, I think consistency demands that we show them for the question too. And just as we're not showing graphical meters for answers (because we don't need to), I wouldn't show one for the question on the question page.
A final thought about meters on the question page: when we were kicking ideas around on wireframes, someone suggested that the meter should take other factors into account, for example that a question is closed. If the function of the meter is to highlight questions that are likely to be most fruitful for a busy skimmer, then we should take such factors into account -- at which point the meter isn't just a reflection of score. If, on the other hand, the meter is just a reflection of score, then what are we trying to accomplish by showing it? (The answer might be multi-modal communication, but I'd rather ask and find out than guess.)
I agree that on a list of questions, one clear indicator of fitness is most helpful. On a post's own page, it might make more sense to also show the up- and down-vote counts.
Given that the "score," here, does not directly represent anything anyone can count, but rather is the result of plugging the up- and down-votes into a not-fully-intuitive continuous function, I think that showing the literal number that comes out of that function would be more confusing than helpful. Instead, I suggest representing it with a Likert scale, with a Help topic that provides both the formula and broad qualitative interpretations of the various scores. For example:
-
↓↓ (Wilson score 0 - 20%): The community has expressed a clear consensus that this is not a helpful post.
-
↓ (Wilson score 20% - 40%): There is some indication from the community that this is not a helpful post.
-
↕ (Wilson score 40% - 60%): The community hasn't expressed a clear consensus regarding how helpful this post is.
-
↑ (Wilson score 60% - 80%): There is some indication from the community that this is a helpful post.
-
↑↑ (Wilson score 80 - 100%): The community has expressed a clear consensus that this is a helpful post.
In place of the ASCII arrow glyphs in this example, some variation of a traditional five-level signal-strength symbol could be used, such as:
All the proposals so far are missing what people really want to know, which are two orthogonal metrics:
- How good/bad the crowd thinks the answer is.
- How heavily the crowd has weighed in.
The first tells you how much to believe the answer, and the second how much to believe the first point. Put another way, you want to see a score and how big the error band is for that score.
So lets show that.
The obvious formula for #1 is (up votes) / (total votes). I'd multiply that by 100 then round to the nearest integer. Values from 0-100 are easier to explain than 0.0 to 1.0. You don't really need to know the difference between 98.2 and 98.4, so keep it simple.
There are more choices how to present the confidence, but the total number of votes is a really simple value. We could get into probability, gaussian distribution assumptions, standard deviations, and the like, but too many people won't understand that and just tune out.
Examples
+5, -3: Score 63, votes 8
+20, -15: Score 57, votes 35
+0, -0: Votes 0
+0, -3: Score 0, votes 3
I would display the 0-100 score most prominently, probably without a label, then the number of votes below that in smaller font with a label.
0 comment threads
instead of - Absolutely not!
Showing separate + and - votes is a good thing. Or as they say, that's not a bug, it's a feature.
If you want to show the score (whether integer net score or Wilson # or whatever) in addition to the + and - values, that's fine with me. But the + and - values are themselves very useful, and anyone who wants a simple net score can easily figure that out.
To answer the "redundant" part, how about always showing the non-zero parts. Examples:
- +2, -1 - show both (as now)
- +2, -0 - show: +2 (-0 is implied)
- +0, -1 - show: -1 (+0 is implied)
- +0, -0 - show either: 0 (unsigned, single value) or some "No votes yet" indication.
That cuts down on the clutter for the (hopefully common) situation where a post is all + or all -, while keeping things 100% clear when there is a mixture.
In my view displaying fractional numbers representing an unintuitive measure would be even worse than displaying two integers representing the up and down counts.
I would be in favor of displaying a single overall count as it is on SE, and indicating the rest by the background color of the overall count, e.g.
- Green: Almost no down-votes
- Orange: Mixed votes
- Red: Negative votes Perhaps the color could represent the Wilson score. The details of the up and down votes (and maybe even the raw Wilson score) could be displayed while hovering above the over count.
Edit: To make the coloring more friendly one of these pallets could be used in addition to shading, e.g. (In this example I assume that there was total of 100 or 101 votes, the difference of up and down votes is displayed):
V1
V2
V3
1 comment thread